Integration of Gene Dosage and Gene Expression in Non-
Small Cell Lung Cancer, Identification of HSP90 as
Marie ¨lle I. Gallegos Ruiz1, Karijn Floor1, Paul Roepman2, Jose ´ A. Rodriguez1, Gerrit A. Meijer3,
Wolter J. Mooi3, Ewa Jassem4, Jacek Niklinski5, Thomas Muley6, Nico van Zandwijk7, Egbert F. Smit8,
Kristin Beebe9, Len Neckers9, Bauke Ylstra3., Giuseppe Giaccone1.*
1Department of Medical Oncology, Vrije Universiteit Medisch Centrum, Amsterdam, The Netherlands, 2Agendia BV, Amsterdam, The Netherlands, 3Department of
Pathology, Vrije Universiteit Medisch Centrum, Amsterdam, The Netherlands, 4University of Gdansk, Gdansk, Poland, 5University of Bialystok, Bialystok, Poland,
6Thoraxklinik Heidelberg, University of Heidelberg, Heidelberg, Germany, 7Netherlands Cancer Institute, Amsterdam, The Netherlands, 8Department of Pulmonary
Diseases, Vrije Universiteit Medisch Centrum, Amsterdam, The Netherlands, 9Urologic Oncology Branch, Center for Cancer Research, National Cancer Institute, Bethesda,
Maryland, United States of America
Background: Lung cancer causes approximately 1.2 million deaths per year worldwide, and non-small cell lung cancer
(NSCLC) represents 85% of all lung cancers. Understanding the molecular events in non-small cell lung cancer (NSCLC) is
essential to improve early diagnosis and treatment for this disease.
Methodology and Principal Findings: In an attempt to identify novel NSCLC related genes, we performed a genome-wide
screening of chromosomal copy number changes affecting gene expression using microarray based comparative genomic
hybridization and gene expression arrays on 32 radically resected tumor samples from stage I and II NSCLC patients. An
integrative analysis tool was applied to determine whether chromosomal copy number affects gene expression. We
identified a deletion on 14q32.2-33 as a common alteration in NSCLC (44%), which significantly influenced gene expression
for HSP90, residing on 14q32. This deletion was correlated with better overall survival (P=0.008), survival was also longer in
patients whose tumors had low expression levels of HSP90. We extended the analysis to three independent validation sets
of NSCLC patients, and confirmed low HSP90 expression to be related with longer overall survival (P=0.003, P=0.07 and
P=0.04). Furthermore, in vitro treatment with an HSP90 inhibitor had potent antiproliferative activity in NSCLC cell lines.
Conclusions: We suggest that targeting HSP90 will have clinical impact for NSCLC patients.
Citation: Gallegos Ruiz MI, Floor K, Roepman P, Rodriguez JA, Meijer GA, et al. (2008) Integration of Gene Dosage and Gene Expression in Non-Small Cell Lung
Cancer, Identification of HSP90 as Potential Target. PLoS ONE 3(3): e0001722. doi:10.1371/journal.pone.0001722
Editor: Christoph Plass, Ohio State University, United States of America
Received December 28, 2007; Accepted February 4, 2008; Published March 5, 2008
Copyright: ? 2008 Gallegos Ruiz et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing Interests: Paul Roepman is employed by Agendia B.V.
* E-mail: email@example.com
. These authors contributed equally to this work.
Lung cancer is the leading cause of cancer deaths worldwide
, and non-small cell lung cancer (NSCLC) represents 85% of
lung cancers. A better understanding of the molecular events
underlying the development and progression of the disease may
contribute to improve clinical management of NSCLC patients. A
number of genes, e.g. P53, RAS, P16 and EGFR, have been
shown to be altered in NSCLC . Given the heterogeneous and
complex nature of this tumor type, it is likely that many genes
driving NSCLC tumorigenesis have yet to be identified.
Chromosomal aberrations are thought to be critical events in
human tumorigenesis, and several genomic regions frequently
harboring DNA gains (3q, 5p, 7q, 8q, 11q and 16p) and losses (3p,
4q, 5q, 6q, 8p 9p and 13q, 17q) have been identified in NSCLC
patients . Using array based comparative genomic hybridiza-
tion (aCGH) and gene expression microarrays, DNA copy number
changes and gene expression can be measured throughout the
whole genome of tumor cells. By combining the data from these
analyses, it is possible to obtain an integrated genome wide view of
gene dosage aberrations and their effect on gene expression, which
might help in identifying genes important in NSCLC .
In the present study, we have performed an integrative analysis
of chromosomal copy number and gene expression on radically
resected tumor samples from 32 NSCLC patients. Two new
algorithms, ‘CGH call’ and ‘ACE-it’, were applied to
analyze the data. We identified a deletion on chromosome region
14q32.2-33 in 44% of NSCLC patients. This deletion was related
with improved patient survival, and was associated with decreased
expression of HSP90, a molecular chaperone for several
oncoproteins that is being explored as a novel target in anticancer
therapy. Low HSP90 expression was correlated with improved
survival in the 32 NSCLC patients analyzed initially. Further
analysis of three independent sets of NSCLC patients confirmed a
PLoS ONE | www.plosone.org1 2008 | Volume 3 | Issue 3 | e1722
significant association between patient survival and HSP90
expression. In addition, in vitro experiments show NSCLC cell
lines to be extremely sensitive to the HSP90 inhibitor 17-AAG.
Our data suggest and important role for HSP90 in NSCLC.
Patients and samples
The test set consisted of radically resected tumor specimens of
32 early stage NSCLC patients. Three patients had a survival time
of less than 30 days and were considered postoperative deaths.
Therefore these three patients are not included in the survival
analyses. Patients had a median follow up of 86 months (range 0.4-
135.5). Verbal informed consent had been obtained from all
patients and handling of samples was in accordance with protocols
approved by the ethical board ‘‘subcommissie voor de ethiek van
het mensgebonden onderzoek’’ from the VU University Medical
Center in Amsterdam.
The first validation set consisted of 140 radically resected
NSCLC patients from the European lung cancer consortium.
Patients had a median follow up of 35 months. All patients
included had had no prior malignancy, pathological tumor stage 1
or 2 (T1-2), node stage 0+1 (N0-1), no distant metastasis (M0) at
time of operation, and no residual disease after resection (R0).
None of these patients received (neo)adjuvant chemo- or
radiotherapy. The second validation consisted of 111 early stage
NSCLC patients from Bild et al. . The third validation set
consisted of the publicly available ‘‘datasets 1 and 2’’ from Lu et al.
 and contained 54 early stage NSCLC patients. A full
description of patient characteristics of all four patient sets is
provided in Table 1.
Isolation of genomic DNA and array Comparative
Cryo-sections of frozen tissue samples, flanking the sections used
for RNA and DNA isolation, were verified by the study pathologist
(WM) to contain at least 50% of tumor cells. Genomic DNA was
extracted from each sample using Trizol following manufacturer
instructions (Life Technologies, Breda, The Netherlands). DNA
labeling and hybridization on CGH 30K oligonucleotide micro-
arrays was performed as described by van den IJssel et al .
RNA isolation and gene expression micro arrays
RNA isolation and cDNA labeling followed standard protocols.
Hybridization was performed on Agilent platform according to
standard procedures described by the manufacturer and elsewhere
For array CGH, spot analysis and quality control were
performed using BlueFuse version 3.2 (BlueGenome, Cambridge,
UK). Breakpoints, gains, losses and amplifications were detected
using the algorithm CGH call. This algorithm converts raw
log2ratios to absolute measures of ‘‘loss’’, ‘‘normal’’, ‘‘gain’’ or
‘‘amplification’’ by applying a segmentation algorithm combined
with a probability mixture model . In order to statistically test
whether gene expression was affected by gene dosage we applied
Table 1. Clinical characteristics of test and validation patient sets
Validation set 1
Validation set 2
Validation set 3
n (%)n (%)n (%)n (%)
Male 24 (75) 105(75)24(44)
Female8 (25)35 (25)29 (54)
Adenocarcinoma13 (41) 43 (31)58 (52)14 (26)
Squamous Cell Carcinoma15 (47)78 (56)53(48) 36 (67)
Large Cell Carcinoma3 (9)7 (5)0 (0)4 (7)
Others1 (3) 12(8)0(0)0 (0)
Smoking status n/an/a
Never0 (0)7 (5)
Former16 (50) 57(41)
Current 11 (34)65 (46)
Unknown5 (16)11 (8)
IA14(44) 25 (18)30 (27) 47 (87)
IB9 (28) 68 (48)27 (24)7 (13)
IIA2 (6)5 (4)0 (0)0 (0)
IIB7 (22) 42(30)0 (0)0 (0)
median[range] median[range] median [range]median[range]
Age at diagnosis-years 67 22–786437–79n/a6648–81
Percentages that do not reach 100% indicate missing data; n/a=information not publicly available
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org22008 | Volume 3 | Issue 3 | e1722
an array CGH expression integration tool, ACE-it, in which the
called array CGH and normalized log10 ratios for expression
arrays were used as input data. ACE-it uses the one-sided
Wilcoxon rank sum statistics to test which chromosomal copy
number aberrations recurrently affect RNA expression. Calculated
p-values are adjusted for multiple testing using the Benjamini
Hochberg method . ACE-it only tests genes that meet the
criteria of contamination and balance, which are controlled
through a threshold on the number of samples. Here the threshold
was set at a fixed default setting of 9 samples, meaning that only
those chromosomal positions were taken into account that had at
least 9 samples in one CGH calling status and no more than 9 in
the other status. The entire array CGH data set of the test series
(n=32) is available at the GEO database (http://www.ncbi.nlm.
nih.gov/projects/geo, accession number GSE7878). The gene
expression data of both the test set (n=32) and validation set 1
(n=140) for the 359 genes identified by ACE-it is available at
the Array Express database (www.ebi.ac.uk/aerep/login, acces-
sion number Array design: A-MEXP-749, Experiment data:
Gene expression data of validation set 2 was obtained using
Affymetrix Hu133plus2 chips (GEO accession number GSE3141).
The mean value of the MAS5 calculated signal intensities of four
probesets detecting HSP90AA1 was used in our calculations.
The third validation set contained gene expression data from
Affymetrix Hu95 and Hu133 chips (GEO accession number
GSE6253). The Hu95 chip contained one probeset detecting
HSP90AA1 and the Hu133 chip contained four probesets, of
which the mean value of the RMA calculated signal intensities was
used in our calculations.
An univariate cox regression analysis was performed to
investigate relation of gene expression values with survival time.
Survival curves were constructed using the Kaplan Meier method
and differences in overall were evaluated using the log-rank test. In
the test set, three patients with less than 30 days survival time were
excluded from the survival analysis, as their death was considered
surgical mortality. To determine the independent effects of HSP90
expression, histologic subtype, tumor stage, age and gender, a
multivariate cox regression analysis was performed. A P value of
less than 0.05 was considered statistically significant.
Multiplex Ligation dependent Probe Amplification
For Multiplex Ligation
(MLPA), the subtelomere probe set P070 (MRC Holland,
Amsterdam, The Netherlands) containing a probe located within
the band 14q32.33 (region 104874216 to 105070384) was used.
MLPA was performed according to manufacturer’s instructions
using 100ng DNA as input. DNA isolated from blood of a pool of
healthy donors was used as reference sample. Probe signals were
normalized by dividing the peak area of chromosome 14q by the
peak area of chromosome 14p. MLPA generated 14q/14p ratios
are plotted against the mean normalized log2ratios of the oligos in
area 104787271-105071522 from the array (total of 11 oligos).
This region covers the region of the MLPA probe.
In order to validate expression values obtained via gene
expression arrays, we performed quantitative real-time PCR using
TaqmanH technology and the ABI PRISM 7500 Sequence
Detection System instrument equipped with the SDS version
1.3.0 software (Applied Biosystems, Foster City, CA, USA).
Forward and reverse primers and probes were designed
and produced byApplied
(Hs00743767_sH), and for the endogenous control gene GUSB
Table 2. Chromosomal regions with gains and losses present
in .20% of NSCLC patients analyzed
Start EndStart End
3 95285925 1267490423q11.2 3q21.2
3 1303584021992538463q21.3 3q29
5 237445 503848055p15.33 5q11.1
7 8160729 269121477p21.37p15.2
7 102223012 1235765277q22.1 7q31.32
7 129933159137018487 7q32.27q33
Start EndStart End
1 7942876218763 1p36.331p36.31
2 2184931562203216492q35 2q35
2239090261242663692 2q37.3 2q37.3
3 4668709352531946 3p21.31 3p21.1
4 76846 97557094p16.34p16.1
5 175887337 1768797115q35.2 5q35.3
7 2497672725138 7p22.37p22.3
7 4452136944915245 7p13 7p13
10 132822632135257796 10q26.3 10q26.3
16 374984794625 16p13.316p13.3
2236290852 36537184 22q13.1 22q13.1
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org32008 | Volume 3 | Issue 3 | e1722
(Hs00939626_mi). PCR was carried out in a 25-ml reaction
volume that contained 50 ng of cDNA, 16 TaqMan Universal
PCR Master Mix, and the primer and probe sets for HSP90AA1
and GUSB. Each sample was analyzed in duplicate, and the
average threshold cycle (Ct) values of each sample for GUSB, were
subtracted from the average Ct values for HSP90AA1. Taqman-
generated DCt values were log transformed to expression values
and plotted against the Dlog2ratios between GUSB and
HSP90AA1 from the expression array.
Cell growth inhibition studies
Growth inhibition following in vitro exposure to the clinically
used Hsp90 inhibitor 17-AAG (InvivoGen, San Diego, CA) was
examined in 4 EGFR wild type NSCLC cell lines (H460, H157,
H441 and A549, obtained from Drs. P. Dennis and F. Kaye, NCI,
Bethesda, MD). Briefly, 105cells were seeded in 6-well tissue
culture plates (Sigma-Aldrich, St. Louis, MO), allowed to adhere,
and then continuously exposed to various concentrations of 17-
AAG (0, 10, 30, 100, 300, 1000 nM) for 24, 48, or 72 hours. At
each time point, cells were detached from the wells, incubated with
trypan blue, and viable (trypan blue-excluding) cell number was
determined in triplicate using a hemacytometer. The mean cell
number with standard error bars is shown at each time point. In
addition, the IC50 (drug concentration at which 50% growth
inhibition is obtained) of 17-AAG at 72 hours was determined for
each cell line.
Gene dosage-related gene expression changes in NSCLC
Chromosomal aberrations were abundant in the 32 NSCLC
patients analyzed. In order to identify breakpoints of gains and
losses we applied the algorithm CGH call . In Table 2 the
chromosomal regions in which gains or losses were present in at
least 20% of patients are listed. The statistical tool ACE-it  was
used to determine whether gene copy number affected gene
expression. A total of 359 transcripts turned out to be significantly
affected by copy number. In Figure 1 the areas of affected genes
are indicated for 32 NSCLC patients, shown in green (gained
regions) or red (lost regions).
An univariate cox regression survival analysis was performed for
expression of all 359 genes that were identified to be influenced by
copy number (see supplementary table S1). After multiple testing
correction (using the Benjamini Hochberg method) none of 359
genes remained significant for survival on this small patient set
(n=32). The top list of genes correlated with survival (ranked on
raw p-values) contained mainly genes located on chromosomes 3
and 5 gained regions. We also observed one gene in the top-20 list,
HSP90AA1, located on chromosome 14. The HSP90AA1 gene
(generally referred to as HSP90), located on 14q32.2, was the only
gene in this region with significantly reduced expression in patients
affected by loss of this region (P=0.05). These observations
prompted us to investigate this locus in more detail.
Figure 1. Percentage of called gains and losses and their effect on gene expression in 32 NSCLC patients. Summary plot for called gains
and losses in 32 resected NSCLC patients with DNA copy number changes indicated in grey. Positive values indicate the percentage of samples found
with a gain. Negative values indicate the percentage of samples harboring a loss at the specified chromosome location. Genes in specified regions
affected by copy number gain are indicated in green and genes affected by copy number loss are indicated in red. A selection of affected genes is
indicated. The full list of 359 affected transcripts can be found in supplementary table S1.
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org4 2008 | Volume 3 | Issue 3 | e1722
Genomic aberrations on 14q and HSP90AA1 gene
expression correlation with survival
We investigated the correlation of the recurrent tight deletion at
region 14q32.2-33, with survival. Interestingly, patients in whom
this region was deleted had an improved overall survival (OS)
compared to patients harboring a normal gene dosage at this locus
(5 year OS 69% vs. 41%, P=0.004-Figure 2A).
In order to investigate HSP90 expression in relation with
survival, we divided the 32 patients into two groups based on their
survival status two years after tumor resection. About half of the
patients were categorized as ‘‘high-risk’’ and half as ‘‘low-risk’’. To
identify whether both risk groups could be discriminated using
gene expression of HSP90, we constructed Kaplan Meier survival
estimates for two equally large groups with ‘‘low’’ and ‘‘normal’’,
based on the median HSP90 expression. Low expression of
HSP90 was associated with better overall survival (5 year OS 70%
vs. 40%, P=0.13–Figure 2B) although differences between the
two groups were initially not significant due probably to a
combination of a low magnitude of the difference and a low
number of patients analyzed.
Technical validation of 14q deletion and HSP90
To validate the deletion observed in region 14q32.2-33 we
performed multiplex ligation dependent probe amplification 
(MLPA) analysis using a subtelomere probeset containing a probe
in region 14q32.33. The correlation coefficient (R2) between the
loss of this area detected by array CGH and by MLPA was 0.439.
To validate the HSP90 expression data obtained with micro-
arrays we performed quantitative RT PCR using the TaqmanH
technology. A good correlation between the expression of HSP90
measured with the two different techniques was observed
Validation of HSP90 expression and relation with survival
in independent patient series
To further investigate the association between low HSP90
expression and NSCLC patient prognosis, we used three
independent validation sets of NSCLC patients. In all three
patient sets, the ‘‘low-risk’’/‘‘high-risk’’ patient distribution across
patient cohorts was approximately two-thirds vs. one-third.
Therefore, we used the 33-percentile of HSP90 expression as cut
off for separation of patients with ‘‘normal’’ (i.e. high-risk) and
‘‘low’’ (i.e. low-risk) expression. For all three validation sets, low
expression of HSP90 was correlated with improved overall
survival. This correlation was significant for the first and third
validation sets (P=0.003 and P=0.04), and borderline significant
for the second set (P=0.07) (Figure 3). Multivariate analysis
revealed that HSP90 prognostic value was independent from the
stage, histologic subtype, age and gender of patients (Table 3).
Validation of Hsp90 as a viable molecular target in a
panel of NSCLC cell lines
Hsp90 is a molecular chaperone that stabilizes several
oncoproteins, including EGFR, and constitutes a novel potential
target for anticancer therapy. The data above showed that Hsp90
expression level is a prognostic indicator of long-term survival in a
large series of NSCLC patients, and suggest that Hsp90 inhibitors
may have broader utility in this disease than previously
recognized. To examine this possibility, we tested whether
pharmacologic inhibition of Hsp90 function would impact the in
vitro cell growth of a panel of NSCLC cell lines. The growth of
Figure 2. Loss of 14q32.2-32.33 chromosomal region and HSP90 expression in relation with survival. (A) Kaplan-Meier curves for overall
survival are shown for 29 patients in relation to gene dosage in chromosome region 14q32.33. (B) Overall survival for 29 patients in relation to HSP90
expression. Low expression was defined as expression lower than the median of the total 32 samples, ‘‘normal’’ expression was defined as higher
than the median of 32 samples. Three of 32 patients included in the analysis of gene dosage and expression were excluded from the survival analysis
because of a survival time of less than 30 days.
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org5 2008 | Volume 3 | Issue 3 | e1722
these cell lines, all bearing wild-type EGFR, was profoundly
inhibited, in a dose- and time-dependent manner, by sub-
micromolar concentrations of 17-AAG, an Hsp90 inhibitor 
currently in phase II clinical trial (Figure 4). At 72 hours, the IC50
was below 50 nM 17-AAG in all cases, and higher drug
concentrations consistently resulted in marked cytotoxicity.
Based on array-CGH data in NSCLC, it has been shown that
multiple molecular carcinogenesis pathways exist that are most
likely related to gender and smoking habits . Furthermore, it
was shown that there is a large overlap between aberrations
observed in the adenocarcinoma and squamous cell carcinoma
subtype, except for 3q gains which seem to be more specific for the
squamous cell carcinoma subtype . Various gene expression
signatures have been correlated to survival of NSCLC patients
[17–19]. In addition, molecular studies have allowed the
development of personalized treatment approaches in several
tumor types [20–22]. However, it is likely that many cancer-
related target genes have not been identified yet. In this regard,
integrated genome wide screening of copy number changes and
gene expression using microarrays has been recently carried out in
various tumor types to identify genes whose expression is affected
by gene dosage [4,23–26]. These studies aim to identify novel
cancer-related genes and to define novel biomarkers for response
or prognostic signatures. In both NSCLC and ductal pancreatic
cancer, two focal amplifications of 8p12 and 20q11 have been
studied in detail leading to two candidate genes (WHSC1L1 and
TPX2) important in these diseases 
We here performed an integrated genome wide screening of
gene copy number changes and gene expression in 32 radically
resected NSCLC patients, in order to identify novel NSCLC-
related genes. By using ACE-it, a novel informatics tool for
integration of gene dosage and gene expression data, we identified
359 transcripts to be significantly affected by copy number. A cox
survival analysis on all 359 genes revealed no significant relation
after multiple testing. The top list of genes related to survival
mainly included genes residing on gained regions 3q and 5p.
These regions cover many genes and to pinpoint the gene of most
importance is a challenging task. Further investigations should
elucidate the importance of these genes in relation to NSCLC, in
particular of those in the top list of correlation with survival such as
SLC45A2, WDR70 and NIPBL (see supplementary table S1). In
this paper we focused on the recurrent deletion on chromosome 14
and the gene HSP90, which was also in the top list of relation with
survival and not previously investigated in detail. Deletion of
region 14q32.2-33 was correlated with improved survival, further
suggesting that it may contain one or more genes related to
NSCLC progression. Deletion of this region has been previously
described by one group reporting genomic aberrations in NSCLC,
but was not investigated in further detail . Out of the 109
genes mapping to the 14q32.2-33 region, HSP90 was the only
gene with significantly lower expression in patients harboring the
14q32.2-33 deletion. In the initial series of 29 patients (3 patients
excluded from survival analysis), we observed improved survival in
patients with lower levels of HSP90. The association between
HSP90 expression levels and NSCLC patient prognosis was
confirmed to be significant in three independent validation sets of
NSCLC patients. Multivariate analysis including stage, histology,
age and gender showed that HSP90 remained independently
related to survival.
A critical issue in defining ‘‘low’’ and ‘‘normal’’ expression is the
choice of an appropriate cut off value. In the initial analyses we
used the median of expression ratios as cut off between ‘‘low’’
and ‘‘normal’’ expression, since the low-risk and high-risk
separation of patients was equal. However, in the validation
Table 3. Multivariate cox analysis of HSP90 expression and
Validation set 1 Validation set 2Validation set 3
HR 95% CI
value HR 95% CI
value HR 95% CI
HSP900.41 0.22–0.750.004 0.52 0.27–0.960.036 0.24 0.05–1.070.061
STAGE1.47 1.11–1.950.008 n/a 0.51 0.13–1.940.320
HISTOLOGY 0.84 0.59–1.190.320 0.71 0.41–1.210.210 0.48 0.18–1.290.140
AGE0.99 0.95–1.030.550 n/a1.00 0.94–1.070.900
GENDER0.26 0.09–0.720.010 n/a 1.23 0.45–3.340.690
HR=Hazard Ratio; n/a=data not publicly available or incomplete
Figure 3. HSP90 expression and survival in three validation sets of NSCLC patients. Overall survival for (A) 140 patients with NSCLC,
validation set 1 (B) 111 NSCLC patients, validation set 2 and (C) 54 patients with NSCLC, validation set 3. The cut off for distinction between low and
‘‘normal’’ expression was based on the 33-percentile of expression values.
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org6 2008 | Volume 3 | Issue 3 | e1722
sets the low-risk and high-risk survival groups were not equally
balanced (two thirds versus one third). Consequently the cut off
values used in these data sets was not the median value, but the 33-
In this study, using a genome wide integrative analysis of gene
copy number and expression we were able to identify expression of
HSP90 as an important gene in early stage NSCLC patients.
HSP90 is a chaperone protein involved in the stabilization of
multiple oncoproteins such as EGFR, Her-2 and Akt . Recent
work has shown that HSP90 plays a role in maintaining the active
conformation of EGFR and in particular EGFR mutants [29,30].
We show here that several NSCLC cell lines bearing wild-type
EGFR are sensitive to HSP90 inhibition, indicating that the
inhibitory effect of 17-AAG can not be solely attributed to mutant
EGFR. HSP90 has been recently recognized as a potential cancer
therapeutic target and investigations of HSP90 inhibitors are
ongoing [31,32]. In this regard, glioblastoma cells overexpressing
EGFR, but resistant to inhibition by EGFR kinase inhibitors, were
sensitive to HSP90 inhibition . Therefore, while Hsp90 may
be required to stabilize over-expressed or mutated EGFR, our
data support a more wide-ranging and complex role for Hsp90 in
mediating NSCLC growth and survival. The low nanomolar
sensitivity observed in the 4 cell lines tested in our experiments, is
in agreement with other published reports of highly 17-AAG
sensitive tumor cell lines [34–36]. These concentrations are readily
achievable in patients for prolonged time periods using current
scheduling and dosing regimens .
In summary, the observation that HSP90 expression level is a
prognostic factor for NSCLC patient survival (independent of
EGFR mutational status), coupled with the extreme sensitivity of
EGFR wild type NSCLC cells to the Hsp90 inhibitor 17-AAG,
suggests that Hsp90 inhibitors may have greater clinical utility in
NSCLC than has been previously considered and warrants further
investigation of the dependence of other proto-oncogenes on this
chaperone protein in NSCLC.
Found at: doi:10.1371/journal.pone.0001722.s001 (0.03 MB
We thank the European lung cancer consortium for providing microarray
gene expression data of the 140 independent NSCLC patients; Netherlands
Cancer Institute: Sjaak Burgers, Tony van de Velde; Medical University of
Gdansk: Jacek Jassem, Amelia Szymanowska, Marcin Skrzypski, Barbara
Szostakiewicz, Witold Rzyman; Medical University of Bialystok: Jerzy
Figure 4. Sensitivity of a panel of wild type EGFR-expressing NSCLC cell lines to the Hsp90 inhibitor 17-AAG. (A–D) time- and dose-
dependent inhibition of the in vitro growth of H460, H157, H441, and A549 NSCLC cell lines following exposure to 17-AAG. Cells were seeded at 105/
well, and viable cell number was determined on subsequent days as described in Methods. 17-AAG concentrations at (H441) or above (A549, H460 &
H157) 30 nM uniformly resulted in time-dependent loss of cell viability. The IC50 value of 17-AAG (continuous exposure for 72 h) for each cell line is
as follows: H460=30 nM, H157=15 nM, H441=8 nM, and A549=20 nM.
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org7 2008 | Volume 3 | Issue 3 | e1722
Laudanski, Miroslaw Kozlowski, Lech Chyczewski; Thoraxklinik Heidel- Download full-text
berg: Michael Meister, Philipp A. Schnabel, Henrik Dienemann and Hans
Hoffman. We also give our special thanks to Sjoerd Vosse (Department of
Pathology, VU University Medical Center, Amsterdam, The Netherlands)
for helpful bioinformatics discussions and to Hans Gille (Department of
Clinical Genetics, VU University Medical Center, Amsterdam, The
Netherlands) for performing MLPA analysis.
Conceived and designed the experiments: BY GM GG JR MG. Performed
the experiments: KF KB MG. Analyzed the data: PR WM MG.
Contributed reagents/materials/analysis tools: EJ JN TM Nv ES LN.
Wrote the paper: BY PR GM GG JR ES LN MG. Other: Designed the
1. Jemal A, Siegel R, Ward E, Murray T, Xu JQ, et al. (2006) Cancer statistics,
2006. Ca-A Cancer Journal for Clinicians 56: 106–130.
2. Sekido Y, Fong KM, Minna JD (2003) Molecular genetics of lung cancer.
Annual Review of Medicine-Selected Topics in the Clinical Sciences 54: 73–87.
3. Balsara BR, Testa JR (2002) Chromosomal imbalances in human lung cancer.
Oncogene 21: 6877–6883.
4. Chin K, DeVries S, Fridlyand J, Spellman PT, Roydasgupta R, et al. (2006)
Genomic and transcriptional aberrations linked to breast cancer pathophysiol-
ogies. Cancer Cell 10: 529–541.
5. van de Wiel MA, Kim KI, Vosse SJ, van Wieringen WN, Wilting SM, et al.
(2007) CGHcall: calling aberrations for array CGH tumor profiles. Bioinfor-
matics 23: 892–894.
6. van Wieringen WN, Belien JAM, Vosse SJ, Achame EM, et al. (2006) ACE-it: a
tool for genome-wide integration of gene dosage and RNA expression data.
Bioinformatics 22: 1919–1920.
7. Bild AH, Yao G, Chang JT, Wang Q, Potti A, et al. (2006) Oncogenic pathway
signatures in human cancers as a guide to targeted therapies. Nature 439:
8. Lu Y, Lemon W, Liu PY, Yi YJ, Morrison C, et al. (2006) A gene expression
signature predicts survival of patients with stage I non-small cell lung cancer.
Plos Medicine 3: 2229–2243.
9. van den IJssel P, Tijssen M, Chin SF, Eijk P, Carvalho B, et al. (2005) Human
and mouse oligonucleotide-based array CGH. Nucleic Acids Research 33.
10. Glas AM, Floore A, Delahaye LJMJ, Witteveen AT, Pover RCF, et al. (2006)
Converting a breast cancer microarray signature into a high-throughput
diagnostic test. Bmc Genomics 7.
11. van de Wiel MA, Kim KI, Vosse SJ, van Wieringen WN, Wilting SM, et al.
(2007) CGHcall: calling aberrations for array CGH tumor profiles. Bioinfor-
matics 23: 892–894.
12. Benjamini Y, Hochberg Y (1995) Controlling the False Discovery Rate-A
Practical and Powerful Approach to Multiple Testing. Journal of the Royal
Statistical Society Series B-Methodological 57: 289–300.
13. Schouten JP, McElgunn CJ, Waaijer R, Zwijnenburg D, Diepvens F, et al.
(2002) Relative quantification of 40 nucleic acid sequences by multiplex ligation-
dependent probe amplification. Nucleic Acids Research 30: e57.
14. Schulte TW, Neckers LM (1998) The benzoquinone ansamycin 17-allylamino-
17-demethoxygeldanamycin binds to HSP90 and shares important biologic
activities with geldanamycin. Cancer Chemotherapy and Pharmacology 42:
15. Shibata T, Uryu S, Kokubu A, Hosoda F, Ohki M, et al. (2005) Genetic
classification of lung adenocarcinoma based on array-based comparative
genomic hybridization analysis: Its association with clinicopathologic features.
Clinical Cancer Research 11: 6177–6185.
16. Tonon G, Wong KK, Maulik G, Brennan C, Feng B, et al. (2005) High-
resolution genomic profiles of human lung cancer. Proceedings of the National
Academy of Sciences of the United States of America 102: 9625–9630.
17. Beer DG, Kardia SLR, Huang CC, Giordano TJ, Levin AM, et al. (2002) Gene-
expression profiles predict survival of patients with lung adenocarcinoma.
Nature Medicine 8: 816–824.
18. Wigle DA, Jurisica I, Radulovich N, Pintilie M, Rossant J, et al. (2002)
Molecular profiling of non-small cell lung cancer and correlation with disease-
free survival. Cancer Research 62: 3005–3008.
19. Chen HY, Yu SL, Chen CH, Chang GC, Chen CY, et al. (2007) A five-gene
signature and clinical outcome in non-small-cell lung cancer. New England
Journal of Medicine 356: 11–20.
20. Lynch TJ, Bell DW, Sordella R, Gurubhagavatula S, Okimoto RA, et al. (2004)
Activating mutations in the epidermal growth factor receptor underlying
responsiveness of non-small-cell lung cancer to gefitinib. New England Journal of
Medicine 350: 2129–2139.
21. Pao W, Wang TY, Riely GJ, Miller VA, Pan QL, et al. (2005) KRAS mutations
and primary resistance of lung adenocarcinomas to gefitinib or erlotinib. Plos
Medicine 2: 57–61.
22. Slamon D, Pegram M (2001) Rationale for trastuzumab (Herceptin) in adjuvant
breast cancer trials. Seminars in Oncology 28: 13–19.
23. Balsara BR, Sonoda G, duManoir S, Siegfried JM, Gabrielson E, et al. (1997)
Comparative genomic hybridization analysis detects frequent, often high-level,
overrepresentation of DNA sequences at 3q, 5p, 7p, and 8q in human non-small
cell lung carcinomas. Cancer Research 57: 2116–2120.
24. Pollack JR, Sorlie T, Perou CM, Rees CA, Jeffrey SS, et al. (2002) Microarray
analysis reveals a major direct role of DNA copy number alteration in the
transcriptional program of human breast tumors. Proceedings of the National
Academy of Sciences of the United States of America 99: 12963–12968.
25. Redon R, Muller D, Caulee K, Wanherdrick K, Abecassis J, et al. (2001) A
simple specific pattern of chromosomal aberrations at early stages of head and
neck squamous cell carcinomas: PIK3CA but not p63 gene as a likely target of
3q26-qter gains. Cancer Research 61: 4122–4129.
26. Tay ST, Leong SH, Yu K, Aggarwal A, Tan SY, et al. (2003) A combined
comparative genomic hybridization and expression microarray analysis of gastric
cancer reveals novel molecular subtypes. Cancer Research 63: 3309–3316.
27. Choi JS, Zheng LT, Ha E, Lim YJ, Kim YH, et al. (2006) Comparative genomic
hybridization array analysis and real-time PCR reveals genomic copy number
alteration for lung adenocarcinomas. Lung 184: 355–362.
28. Burrows F, Zhang H, Kamal A (2004) Hsp90 activation and cell cycle
regulation. Cell Cycle 3: 1530–1536.
29. Yang SC, Qu S, Perez-Tores M, Sawai A, Rosen N, et al. (2006) Association
with HSP90 inhibits Cbl-mediated down-regulation of mutant epidermal growth
factor receptors. Cancer Research 66: 6990–6997.
30. Shimamura T, Lowell AM, Engelman JA, Shapiro GI (2005) Epidermal growth
factor receptors harboring kinase domain mutations associate with the heat
shock protein 90 chaperone and are destabilized following exposure to
geldanamycins. Cancer Research 65: 6401–6408.
31. Cullman SB, Whitesell L (2006) Heat shock protein 90: A unique chemother-
apeutic target. Seminars in Oncology 33: 457–465.
32. Zhang H, Burrows F (2004) Targeting multiple signal transduction pathways
through inhibition of Hsp90. Journal of Molecular Medicine-Jmm 82: 488–499.
33. Thomas CY, Stallings-Mann M, Zang H, Yang YC, Burrows F, et al. (2003)
Glioblastoma (GBM) cells that overexpress epidermal growth factor receptors
(EGFR) are sensitive to HSP90 but not EGFR kinase inhibitors. Clinical Cancer
Research 9: 6098S.
34. Braga-Basaria M, Hardy E, Gottfried R, Burman KD, Saji M, et al. (2004) 17-
allylamino-17-demethoxygeldanamycin activity against thyroid cancer cell lines
correlates with heat shock protein 90 levels. Journal of Clinical Endocrinology
and Metabolism 89: 2982–2988.
35. Yang J, Yang JM, Iannone M, Shih WJ, Lin Y, et al. (2001) Disruption of the
EF-2 kinase/Hsp90 protein complex: A possible mechanism to inhibit
glioblastoma by geldanamycin. Cancer Research 61: 4010–4016.
36. Zsebik B, Citri A, Isola J, Yarden Y, Szollosi J, et al. (2006) Hsp90 inhibitor 17-
AAG reduces ErbB2 levels and inhibits proliferation of the trastuzumab resistant
breast tumor cell line JIMT-1. Immunology Letters 104: 146–155.
37. Xu WP, Neckers L (2007) Targeting the molecular chaperone heat shock protein
90 provides a multifaceted effect on diverse cell signaling pathways of cancer
cells. Clinical Cancer Research 13: 1625–1629.
HSP90 in Lung Cancer
PLoS ONE | www.plosone.org8 2008 | Volume 3 | Issue 3 | e1722