Early Changes in Protein Expression Detected by Mass
Spectrometry Predict Tumor Response to Molecular
Therapeutics
Michelle L. Reyzer, Robert L. Caldwell, Teresa C. Dugger, et al.
Cancer Res 2004;64:9093-9100.
Updated version
Access the most recent version of this article at:
http://cancerres.aacrjournals.org/content/64/24/9093
Cited Articles
This article cites by 29 articles, 14 of which you can access for free at:
http://cancerres.aacrjournals.org/content/64/24/9093.full.html#ref-list-1
Citing articles
This article has been cited by 10 HighWire-hosted articles. Access the articles at:
http://cancerres.aacrjournals.org/content/64/24/9093.full.html#related-urls
E-mail alerts
Reprints and
Subscriptions
Permissions
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at pubs@aacr.org.
To request permission to re-use all or part of this article, contact the AACR Publications
Department at permissions@aacr.org.
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
[CANCER RESEARCH 64, 9093–9100, December 15, 2004]
Early Changes in Protein Expression Detected by Mass Spectrometry Predict
Tumor Response to Molecular Therapeutics
Michelle L. Reyzer,1,2 Robert L. Caldwell,1,2 Teresa C. Dugger,3 James T. Forbes,3 Christoph A. Ritter,6
Marta Guix,3 Carlos L. Arteaga,3,4,5 and Richard M. Caprioli1,2,5
1
Mass Spectrometry Research Center, Departments of 2Biochemistry, 3Medicine, and 4Cancer Biology, 5Vanderbilt-Ingram Comprehensive Cancer Center, Vanderbilt University
School of Medicine, Nashville, Tennessee; and 6Department of Pharmacology and Institute of Pharmacy, Peter Holtz Research Center of Pharmacology and Experimental
Therapeutics, Ernst-Moritz-Arndt-Universitat Greifswald, Greifswald, Germany
ABSTRACT
Biomarkers that predict therapeutic response are essential for the
development of anticancer therapies. We have used matrix-assisted laser
desorption/ionization mass spectrometry (MALDI-MS) to directly analyze
protein profiles in mouse mammary tumor virus/HER2 transgenic mouse
frozen tumor sections after treatment with the erbB receptor inhibitors
OSI-774 and Herceptin. Inhibition of tumor cell proliferation and induction of apoptosis and tumor reduction were predicted by a >80% reduction in thymosin ␤4 and ubiquitin levels that were detectable after 16
hours of a single drug dose before any evidence of in situ cellular activity.
These effects were time- and dose-dependent, and their spatial distribution in the tumor correlated with that of the small-molecule inhibitor
OSI-774. In addition, they predicted for therapeutic synergy of OSI-774
and Herceptin as well as for drug resistance. These results suggest that
drug-induced early proteomic changes as measured by MALDI-MS can
be used to predict the therapeutic response to established and novel
therapies.
INTRODUCTION
There is an increasing number of molecule-targeted anticancer
therapies, both established and in clinical development. In most cases,
the tumors that are dependent on the molecular targets of these
therapies and, therefore, potentially sensitive to these drugs, are not
easily recognizable. Therefore, it becomes imperative to identify a
molecular signature or predictive marker before or during preclinical
anticancer drug development to guide this process. Such information
would allow the selection of appropriate patients into efficacy studies
as well as the exclusion of patients in whom these drugs are unlikely
to exhibit any clinical benefit. In some cases, this predictive molecular
signature is known. For example, the presence of estrogen and progesterone receptors in breast cancers predicts for response to antiestrogens (1). Clinical responses to the anti-HER2 (erbB2) humanized
IgG1 trastuzumab (Herceptin) are limited to breast tumors with HER2
gene amplification (2). Likewise, gastrointestinal stromal tumors with
mutations in c-Kit respond to the small molecule c-Kit tyrosine kinase
inhibitor imatinib (Gleevec; ref. 3). On the other hand, recent trials
with epidermal growth factor receptor (EGFR) inhibitors in unselected patients with epithelial tumors that are known to express the
receptor RNA and protein have shown limited to no clinical activity
(4, 5). Interestingly, two groups recently reported EGFR gene mutations in lung cancers that exhibited robust clinical responses to the
EGFR tyrosine kinase inhibitor gefitinib (Iressa; refs. 6, 7), additionally underscoring the presence of molecular signatures or surrogate
markers than can predict drug action.
Proteomics approaches, including two-dimensional gel electrophoresis and matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS), have been shown to be powerful tools to
discover biomarkers that distinguish normal from diseased tissues.
Chaurand et al. (8) used MALDI-MS to directly analyze mouse colon
tissue and showed that protein patterns were significantly different
between tumor tissue and normal colon tissue. Several tumor-specific
protein markers were identified in this work (8). Yanagisawa et al. (9)
recently showed that tumor subsets of non–small-cell lung cancer give
rise to unique protein patterns. They reported that MALDI-MS–
derived protein profiles from human lung tissue could accurately
distinguish normal lung, adenocarcinoma, squamous-cell carcinoma,
and large-cell carcinoma. They also identified markers that were
indicative of patient prognosis (9). In this article, we show that protein
changes initiated by treatment with a small molecule inhibitor of the
EGFR tyrosine kinase and/or a HER2 blocking antibody predict for
cellular and clinical activity in oncogene-expressing mammary tumors. These changes can be observed as early as after 16 hours of
treatment and, thus, support the use of proteomic-based approaches to
potentially guide individualized anticancer therapies.
MATERIALS AND METHODS
Transplantation and Treatment of Transgenic Tumors. FVB female
transgenic mice expressing the human HER2 cDNA under the control of the
mouse mammary tumor virus (MMTV) promoter develop mammary tumors
with a latency of 8 months (10). Snap-frozen MMTV/HER2 mammary tumors
were provided by Sharon Erickson (Genentech, Inc., South San Francisco,
CA). Through a small incision in the dorsal flank of syngeneic nontransgenic
female mice, an ⬃1-mm3 tumor pellet was inserted with a sterile probe in the
s.c. space near the #1 mammary fat pad just above the shoulder. The incision
site was closed with 1 to 2 sterile wound clips under a protocol approved by
the Vanderbilt Institutional Animal Care and Use Committee. Within 1 to 2
weeks, tumors measuring ⬎3 mm in diameter were palpable. Tumor size was
followed with calipers, and tumor volume was calculated by the formula
(volume ⫽ width2 ⫻ length/2). Herceptin was purchased from the Vanderbilt
University Hospital Pharmacy and administered in sterile isotonic saline at
Received 6/23/04; revised 8/13/04; accepted 9/1/04.
either 10 mg/kg or 30 mg/kg i.p. twice a week. OSI-774 was provided by Mark
Grant support: M. Reyzer is the recipient of a Philip Morris Post-Doctoral Research
Fellowship. Grants from the NIH including R01 CA80195 (to C. Arteaga), R01 CA86243
Sliwkowski (Genentech, Inc.) and prepared every week in captisol (from OSI
(to R. Caprioli), and R01 GM58008 (to R. Caprioli), the Breast Cancer Specialized
Pharmaceuticals, Boulder, CO) as a 20 mg/mL stock solution and administered
Program of Research Excellence (SPORE) Grant P50 CA98131, and the Vanderbiltdaily via orogastric gavage (p.o.) at a dose of 10 to 100 mg/kg in a volume of
Ingram Cancer Center Support Grant CA68485.
100 ␮L. Where indicated, lungs were examined for surface lung metastases as
The costs of publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked advertisement in accordance with
described previously (11).
18 U.S.C. Section 1734 solely to indicate this fact.
Immunoprecipitation, Western Blotting, and Immunohistochemistry.
Requests for reprints: Carlos L. Arteaga, Division of Oncology, Vanderbilt UniverTumors were harvested and either fixed in 10% formalin (VWR Scientific,
sity School of Medicine, 2220 Pierce Avenue, 777 Preston Research Building, Nashville,
West Chester, PA) or homogenized as described previously (12). Tumor
TN 37232-6307. Phone: (615) 936-3524; Fax: (615) 936-1790; E-mail: carlos.arteaga@
vanderbilt.edu or Richard M. Caprioli, Mass Spectrometry Research Center, 465 21st
lysates were precipitated with a COOH-terminus erbB2 (HER2) polyclonal
Avenue South, Medical Research Building III, Room 9160, Vanderbilt University School
antibody (NeoMarkers, Freemont, CA) and protein A-Sepharose (Sigma, St.
of Medicine, Nashville, TN 37232. Phone: (615) 322-4336; Fax: (615) 343-8372; E-mail:
Louis, MO) followed by Western blot procedures with HER2 (NeoMarkers)
r.caprioli@vanderbilt.edu.
and phosphotyrosine (Upstate Biotechnology, Lake Placid, NY) antibodies
©2004 American Association for Cancer Research.
9093
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
(13). Tumor cell proliferation was measured as described previously (14). In
brief, paraffin-embedded tumors were sectioned (5-␮m) and then subjected to
immunohistochemistry with a proliferation cell nuclear antigen (PCNA) antibody (NeoMarkers). Tumor cell apoptosis was assessed by terminal deoxynucleotidyl transferase (Tdt)-mediated nick end labeling (TUNEL) analysis
with the Apoptag Detection Kit (Serologicals Corp., Norcross, GA) according
to the manufacturer’s instructions.
Tissue Preparation and MALDI-MS Analysis. Tissues from animals
were snap-frozen in liquid N2 and stored at ⫺80°C until additional analysis.
Frozen tissues were cut at ⫺15°C into 12-␮m thick sections on a cryostat
(Leica Microsystems Inc., Bannockburn, IL); sections were transferred and
thawed onto gold-coated stainless-steel MALDI target plates. The sections
were desiccated for at least 1 hour before matrix application. For regional
analysis, 0.25 ␮L of the matrix solution (sinapinic acid, 20 mg/mL in 50:50
acetonitrile:0.2% trifluoroacetic acid) was deposited on the tissue, dried, and
an additional 0.25 ␮L was deposited directly on the first spot. For imaging
applications, sinapinic acid matrix was coated over the tissue until there was a
homogeneous layer of matrix crystals over the surface (15).
MALDI time-of-flight mass spectra were acquired on a Voyager DE-STR
mass spectrometer (Applied Biosystems, Foster City, CA) equipped with a
nitrogen laser (337 nm) run at 2 Hz. Linear mode was used in conjunction with
delayed extraction. Data were obtained with an accelerating voltage of 25 kV,
95% grid voltage, 0.05% ion guide wire voltage, a delay time of 150 nanoseconds, and a flight path length of 2.0 m. A QStar Pulsar i (Applied
Biosystems) hybrid QqTOF mass spectrometer equipped with a MALDI
source and a nitrogen laser (337 nm) was used to acquire MALDI-tandem mass
spectra (MS/MS) of the drug OSI-774. The laser was run at 20 Hz. OSI-774
was effectively analyzed by selected reaction monitoring as described previously (16). Briefly, the protonated molecular ion at m/z 394.2 was dissociated,
and the predominant fragment ion at m/z 278.1 was analyzed. Sensitivity was
increased by allowing a wide mass range into the collision quadrupole (⬃5 Da)
and by detecting only a small mass range surrounding the fragment of interest
(m/z 240 to 320) in the time-of-flight region.
Imaging. The Voyager DE-STR MALDI time-of-flight instrument is
equipped with imaging software described previously (17). The software
consists of a graphical user interface that allows the user to specify the area to
be imaged, the distance between laser shots (spatial resolution), the instrument
acquisition method to be used, the number of laser shots to be averaged in a
spectrum, and the mass ranges of interest to monitor. As the image is acquired,
the sample stage is moved from spot to spot, creating a raster of desorbed areas
over the tissue surface. A spectrum is acquired at each spot, and the intensity
of each signal from all of the mass ranges specified is plotted as a function of
the location on the tissue surface. Thus, ion density maps are obtained showing
the localization of compounds of interest within the tissue. The QStar Pulsar i
instrument is equipped with similar software (16).
Protein Identification and Tandem MS. Tumor aliquots of ⬃300-mg
were suspended in homogenization buffer [0.25 mol/L sucrose, 10 mmol/L
Tris, protease inhibitors (pH 7.6)] and homogenized. The homogenate was
centrifuged at 680 ⫻ g for 10 minutes at 4°C, and the pellet was labeled “crude
nuclear fraction.” The supernatant was transferred to an Eppendorf tube and
centrifuged at 10,000 ⫻ g for 10 minutes at 4°C. The soluble portion was
transferred to a new Eppendorf tube and labeled “cytosolic fraction.” To
solubilize the crude nuclear fraction, the pellet was thawed and resuspended in
Tris HCl buffer (0.1 mol/L), agitated with a motor-driven pestle for 1 minute,
and centrifuged at 16,000 ⫻ g for 10 minutes at 4°C. The resulting soluble
material was transferred to a new Eppendorf tube and designated “extract 1.”
The pellet was additionally extracted with 2 mol/L NaCl, 5 mol/L urea, 0.01
mol/L Tris (pH 7.54), and centrifuged (designated extract 2). The resulting
pellet was resuspended in water and treated as described above (extract 3) and
pooled with extract 2. The final pellet was resuspended in 8/2/0.01 (v/v/v)
acetonitrile/water/trifluoroacetic acid (TFA; extract 4). All of the fractions
were frozen in liquid N2 and stored at ⫺70°C until further analysis. To
determine which tumor fraction contained the proteins of interest, dilutions of
the cytosolic and nuclear extracts were analyzed via MALDI-MS. If necessary,
proteins were additionally isolated from others in the extracts by reverse-phase
high-performance liquid chromatography (Agilent 1000, Waldbronn, Germany). Fractions were collected, lyophilized, resuspended in 15 ␮L 50/50/0.1
(v/v/v) water/acetonitrile/TFA, and analyzed as described above. An aliquot of
each protein subfraction (50 to 100 pmol) was reduced in ammonium bicar-
bonate buffer (0.4 mol/L, 25 ␮L) and DTE (5 ␮L, 45 mmol/L) for 15 minutes
at 60°C and alkylated with iodoacetimide (5 ␮L, 100 mmol/L) for 15 minutes
at room temperature in the dark. Protein digestion was done by adding trypsin
(0.05 ␮g) and incubating for 24 hours at 37°C. Trypsin activity was inactivated
by freezing the fractions at ⫺80°C.
Digested samples were thawed on ice, and peptides were purified and
concentrated with ZipTip reversed-phase chromatography per the manufacturer’s instructions (Millipore, Billerica, MA). A sample aliquot (500 nl) was
mixed on the target plate with ␣-cyano-4-hydroxycinnamic acid matrix [500
nl; prepared at 10 mg/mL in 50/50/0.1 (v/v/v) water/acetonitrile/TFA]. The
samples were allowed to dry and were analyzed with a 4700 Proteomics
Analyzer MALDI TOF-TOF mass spectrometer (Applied Biosystems). Peptide
mass fingerprinting was done in MS mode, and peptide masses were searched
against the National Center for Biotechnology Information and SwissProt
databases. Results were confirmed by sequencing the tryptic peptides in
MS/MS mode and searching for the amino acid m/z values in the National
Center for Biotechnology Information and SwissProt databases using a commercially available database searching algorithm (MASCOT).
RESULTS
Drug-Mediated Inhibition of HER2 Induces Rapid Changes in
the Proteome. We first determined the efficacy of signaling inhibitors against tumors arising in MMTV/HER2 transgenic mice (10).
These mice develop metastatic mammary tumors with a median
latency of 8 months. These transgenic tumors can be serially transplanted in FVB wild-type mice where they retain HER2 overexpression. Unless otherwise specified, all of the experiments were done
with a tumor from an F2 mouse (F2–1282). To inhibit tumor growth
we used the following: (1) OSI-774 (erlotinib, Tarceva), an ATPcompetitive small molecule (MW 393) that inhibits the EGFR and the
HER2 tyrosine kinases with an in vitro IC50 of 0.02 and 0.35 ␮mol/L,
respectively (18, 19); and (2) the humanized IgG1 trastuzumab (Herceptin), which binds the ectodomain of the HER2 receptor (20).
Treatment with either OSI-774 or Herceptin eliminated established
MMTV/HER2 tumors of ⱖ100 mm3 in volume (Fig. 1A). Within 48
hours, OSI-774 –treated tumors exhibited loss of HER2 phosphorylation (Fig. 1B). We next compared global protein expression patterns
between treated and untreated tumor transplants via MALDI mass
spectrometry.
MMTV/HER2 tumors from untreated mice were first analyzed to
obtain a baseline protein profile. An example of such a profile,
representing an average of three mass spectra, is shown in Fig. 1C.
There are ⬎200 signals present over the mass range of 2,000 to
30,000 Da. These signals are thought to arise from abundant, soluble
proteins present in the tissue section. Several abundant proteins are
labeled, including thymosin ␤4 (T␤4), ubiquitin, acyl-CoA binding
protein (ACBP), histone H4, and the ␣ and ␤ chains of hemoglobin
(␣⫹ and ␤⫹, singly charged, and ␣2⫹, doubly charged).
Therapy-Induced Changes in the Tumor Proteome Are Doseand Time-Dependent. To examine if the effects of OSI-774 are
dose-dependent, mice bearing two established (ⱖ250 mm3) contralateral tumor transplants each were treated with different daily drug
concentrations. Twenty hours later, the right transplant was harvested.
To control for an antitumor effect, treatment was continued for 9 days,
and volume of the contralateral transplant was measured serially. Only
tumors treated with the higher dose (100 mg/kg) of OSI-774 exhibited
evidence of apoptosis at 20 hours as well as growth arrest after 9 days
of treatment (Figs. 2, A and B).
Mass spectra were acquired and averaged from a total of 3 tumors
per dose harvested 20 hours after the first dose of OSI-774. Each
average spectrum was compared with a control mass spectrum acquired from 9 spectra collected from 3 animals. Selected mass ranges
of the four averaged traces are shown in Fig. 2C. The top spectrum
9094
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
Fig. 1. Drug-mediated inhibition of HER2 induces rapid changes in the proteome of MMTV/
HER2 tumors. A, FVB nontransgenic mice bearing
MMTV/HER2 tumors measuring ⬎100 mm3 were
treated with Herceptin 30 mg/kg i.p. twice a week
or OSI-774 100 mg/kg via orogastric gavage daily.
Tumor volumes in mm3 were calculated as indicated in Materials and Methods. Each data point
represents three mice; bars, ⫾SD. B. Untreated and
treated tumor transplants were harvested 48 hours
after initiation of therapy and homogenized. Lysates were precipitated with a COOH-terminus
HER2 antibody and immune complexes separated
in equal parts by SDS-PAGE followed by immunoblot procedures for HER2 and phosphotyrosine.
C. A representative mass spectrum from one section of an untreated MMTV/HER2 tumor is shown.
Mass spectra acquired from the three matrix spots
shown on the tumor section were averaged and
baseline subtracted. Numerous soluble proteins are
observed in the mass range from 2,000 to 30,000
Da, including T␤4, ACBP, histone H4, and the ␣
and ␤ chains of hemoglobin (␣⫹ and ␤⫹, singly
charged, and ␣2⫹, doubly charged).
shows the protein signals in the range of 4,700 to 5,100 Da. There is
a signal at m/z 4,913 that is of similar intensity in all four groups. The
signal at m/z 4,965, identified as T␤4(21), is of similar intensity in the
control group and the groups treated with 10 mg/kg and 30 mg/kg, but
is significantly decreased (⬃80%) in the 100 mg/kg group. T␤4 is an
actin-sequestering protein and has recently been implicated as having
a role in tumor metastasis by promoting cell migration and angiogenesis (22). In contrast, there was no signal at m/z 4,794 in the untreated
or low dose groups, which was prominent in the tumors treated with
100 mg/kg (Fig. 2C). This signal was sequenced by tandem MS and
identified as a fragment of an E-cadherin binding protein (SwissProt
accession number Q8C7W5). Electrospray ionization-tandem MS
(LTQ, ThermoFinnigan) on the triply charged ion at m/z 1599.95
returned the following sequence via Sequest searching: KAPQGDEGFD YNEEQRYDCK GGELFGNQRR FPGHLFWDFK, with a
predicted molecular weight of Mr 4,783. This sequence corresponds to
residues 53-92 of the protein. E-cadherin is a transmembrane protein
that mediates cell-cell adhesion. Inactivating mutations or inhibition
of E-cadherin function have been associated with tumor cell motility
and progression (23–25). Recently, a novel E-cadherin binding protein called Hakai was described (26), which acts as an ubiquitin-ligase
(E3) involved in ubiquitination of E-cadherin after tyrosine phosphorylation. The sequence of the protein fragment observed from the
treated tumors differs from the analogous sequence of Hakai by the
loss of a single glutamic acid residue after Glu59. A possible role of
this protein on OSI-774 –induced tumor growth inhibition remains to
be studied.
The bottom mass spectrum of Fig. 2C highlights the range of 8,300
to 8,800 Da, where the same dose dependence observed for T␤4 is
observed for ubiquitin (m/z 8,565). Many proteins are targeted for
proteolysis by ubiquitination (27). Therefore, a marked decrease in
free ubiquitin suggests the possibility that more ubiquitin is covalently
bound to proteins targeted for degradation or that there is a decrease
in the production of ubiquitin. Whereas the actual fate of ubiquitin in
these tumors remains unknown, the reduction of ubiquitin correlates
with tumor reduction.
Protein changes described above between the tumors treated with
100 mg/kg and controls were observed 20 hours after a single administration of OSI-774. To determine how early these drug-induced
protein changes can be detected, other mice received 100 mg/kg
OSI-774; tumors were removed at 2, 4, 7.25, and 16 hours after a
single dose and subjected to MALDI-MS. The averaged spectra from
each group were compared with a control spectrum (Fig. 2D). T␤4
and ubiquitin were substantially decreased at 16 hours but not at the
earlier time points. Interestingly, the signal for the E-cadherin binding
protein fragment at m/z 4,794 that had increased at ⬃20 hours was not
observed at any of the earlier time points.
Drug Localization in Tumors Correlates with Proteomic Markers of Cellular Response. To study the relationship between the drug
distribution in tumors and the resulting protein alterations, mass
spectral images were obtained on serial sections of a tumor removed
16 hours after a therapeutic dose of OSI-774 (21, 28). The presence of
OSI-774 in sections of tumor tissue was determined via MALDI
QqTOF mass spectrometry with selected reaction monitoring of the
transition m/z 394.23278.1 (ref. 16; Fig. 3A). An optical image of a
section of treated mouse tumor is shown with four sinapinic acid
matrix spots. Each spot was analyzed by monitoring the intensity of
the primary fragment (m/z 278.1) of protonated OSI-774 (m/z 394.2).
A representative spectrum of detected OSI-774 is shown for spot #3.
Mass spectra were acquired from 2,000 to 30,000 Da every 140 ␮m
across the tissue sections. Each spectrum is an average of 20 laser
shots. The optical image of the tissue sections along with selected ion
images is presented in Fig. 3B. The distribution of a representative
protein that does not change as a result of OSI-774 treatment is
illustrated with histone H4 (dimethylated, diacetylated, m/z 11,344).
In contrast, whereas ubiquitin is homogeneously distributed in the
control tumor, it is markedly decreased throughout the treated tumors.
These ion images support the data obtained from profiling discrete
areas of the tissue surface. There were some signals from both
proteins in the necrotic areas of the tumors. These arose from cell
material and fluid still present in the necrotic area. However, the
9095
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
Fig. 2. Dose- and time-dependence of therapy-induced proteomic changes. A. Two (approximately) 1-mm3 pellets from an MMTV/HER2 mammary tumor were implanted via small
surgical incisions in the s.c. space of the right and left dorsum of wild-type female FVB mice, respectively. Once they reached a volume ⬎250 mm3, treatment with OSI-774 at the
indicated doses p.o. was started. Twenty hours after the first dose, all right-sided tumors were harvested, and mice continued on daily therapy for the next 9 days. Left-sided tumor
volumes are shown; bars, 3 ⫾ SD. B. TUNEL analysis of MMTV/HER2 sections from tumors harvested 20 hours after a variable single dose of OSI-774. Values represent the average;
⫾SD TUNEL⫹ nuclei/10 400⫻ fields per mouse (n ⫽ 3 mice per condition). Only the 100 mg/kg dose produced a significant increase in tumor cell apoptosis compared with controls
(Students t test, P ⬍ 0.05). C. Mass spectral analysis of MMTV/HER2 sections from tumors harvested 20 hours after a variable single dose of OSI-774 reveals several proteomic changes
induced from the 100 mg/kg dose, including down-regulation of T␤4 (m/z 4,965) and ubiquitin (m/z 8,565) and up-regulation of m/z 4,794 (E-cadherin binding protein fragment). D.
Mass spectral analysis of MMTV/HER2 tumor sections from tumors harvested 2 hours, 7.25 hours, or 16 hours after a single 100 mg/kg dose of OSI-774. Down-regulation of T␤4
and ubiquitin was observed at 16 hours but not at the earlier time points. The spectrum from each time point is the average of 4 spectra from one animal, whereas the control spectrum
is the average of 54 spectra from 9 animals.
presence of histone H4 in these areas seems to be reduced relative to
the rest of the tumor.
The spatial localization of OSI-774 was analyzed on a serial tumor
section via MALDI QqTOF tandem mass spectrometry as described
previously (16). Fig. 3B shows an optical image of the tumor section
mounted on the MALDI plate before matrix application. The ion
density map obtained for OSI-774 via selected reaction monitoring of
the transition m/z 394.23278.1 is also shown in Fig. 3B. Spectra were
obtained every 200 ␮m in the x-direction and 400 ␮m in the ydirection. Forty laser shots were averaged at each spot. As shown,
OSI-774 is distributed throughout the tumor section 16 hours after
drug administration predominantly in the more vascularized peripheral sections of the tumor.
Drug-Induced Changes in the Proteome Correlate with Therapeutic Synergy. The EGFR tyrosine kinase inhibitor gefitinib has
been shown to synergize with Herceptin against HER2-overexpressing breast cancer xenografts (29). Therefore, we next tested if the
combination of OSI-774 and Herceptin would be synergistic against
MMTV/HER2 tumors and whether this synergy can be predicted by
early proteomic changes in situ. Tumors measuring ⱖ500 mm3 were
randomized to no treatment, OSI-774, a less than therapeutic dose of
Herceptin, or the combination (n ⫽ 13/condition). At 72 hours, 3
tumors per condition were harvested for subsequent analysis, and the
remaining 10 tumors in each group were treated for 3 weeks. Six of
10 tumors treated with the combination underwent a complete response (Table 1), which was maintained off therapy for the next 9
months; no complete responses were observed in the other groups.
Whereas single treatment with the small molecule or the antibody
induced transient or more sustained growth delay, only the combination induced sustained tumor regressions (Fig. 4A). Tumors receiving
OSI-774 alone or in combination but not those treated with the
subtherapeutic dose of Herceptin showed a ⬎50% inhibition of proliferation as well as a ⬎10-fold increase in apoptosis measured by
TUNEL (Fig. 4B). Lung surface metastases were only observed in
controls but not in both groups receiving OSI-774 (Table 1).
Three tumors for each condition were analyzed by MALDI-MS,
with 5 to 10 spectra acquired per tumor. The resulting spectra were
baseline-subtracted and normalized with a computer program written
in-house, and spectra from each treatment group were averaged together (spectra in which signals for hemoglobin were dominant were
not included). Fig. 4C shows mass spectra in the ranges of m/z 4,700
to 5,100 and m/z 8,300 to 8,800. In tumors treated with OSI-774 or the
9096
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
Fig. 3. Localization of OSI-774 in tumors correlates with mass spectrometric evidence of cellular response. A. Wild-type FVB mice bearing MMTV/HER2 tumors measuring ⬃200
mm3 were treated with 100 mg/kg OSI-774 p.o. Sixteen hours later tumors were harvested. A mass spectrum is shown for the analysis of OSI-774 in spot #3 of one tumor section.
The drug is analyzed via selected reaction monitoring of the transition m/z 394.2 3 278.1. B. One section of untreated tumor and two serial sections of treated tumor were analyzed
via imaging MS. Two selected protein images are shown, one for histone H4 (dimethylated, diacetylated, m/z 11,344) in yellow and one for ubiquitin (m/z 8,565) in green. Whereas
histone is fairly homogenously distributed in both untreated and treated tumor sections, ubiquitin has been markedly down-regulated in treated compared with the untreated tumor. A
mass spectral image of the localization of the drug itself, OSI-774, was done on a serial section of the treated tumor. OSI-774 is distributed throughout the tumor section but is less
evident in the necrotic center.
combination, a decrease in T␤4 and ubiquitin was observed. The
signal at m/z 4,913 is the same among all four groups, as previously
observed (see Fig. 2C). Interestingly, the largest changes observed for
T␤4 and for the E-cadherin binding protein fragment are from the
group treated with both drugs.
The changes observed for the OSI-774-treated group are of a
smaller magnitude than the changes observed in Fig. 2, C and D.
There are several possible reasons for this difference. First, the tumors
used in the drug synergy study were much larger (⬃3 to 4⫻) than the
tumors examined in Fig. 2. This results in slower-growing tumors and
potentially slower turnover of proteins. This would potentially result
in lesser alterations in protein content in response to any stimulus.
Consistent with this change, however, the response of the larger
tumors to OSI-774 alone was not as large as the response observed in
smaller tumors. Thus, the lesser magnitude in protein alteration may
reflect the lower level of tumor reduction.
Table 1
Complete responses
Deaths
Lung surface metastases
Lung weight (mg)
Control
Herceptin
OSI-774
Herceptin
⫹ OSI-774
0/10
4/10
4.5 ⫾ 1.3
179.3 ⫾ 10.5
0/10
2/10
3.1 ⫾ 0.6
168.8 ⫾ 12.5
0/10
1/10
0.0 ⫾ 0.0
148.6 ⫾ 12.8
6/10
2/10
0.0 ⫾ 0.0
138.0 ⫾ 3.0
Two additional observations can be made from these spectra.
First, there is no change in the ubiquitin signal between the control
and Herceptin-treated groups. In addition, there is no difference
between the ubiquitin signal in the OSI-774 group and the combination group, but the ubiquitin signal from both OSI-774 –treated
groups is notably reduced compared with the signal in control and
Herceptin-treated groups. Thus, it seems that the reduction of
ubiquitin originates solely from administration of the small molecule inhibitor OSI-774. The second observation concerns the signal
at m/z 8,719, subsequently identified by peptide mapping as a
calmodulin fragment, which is considerably increased only in the
combination group. In fact, many signals over the entire mass
range (2,000 to 40,000 Da) showed larger changes in the group
treated with the combination with intermediate levels in the single
drug groups, possibly reflecting the enhanced therapeutic effect of
the combined therapy (Fig. 4D; ANOVA, P ⬍ 0.05). In addition to
T␤4 and ubiquitin, these included several histones (H3 and both
dimethylated, diacetylated H4 at m/z 11,344 and dimethylated,
monoacetylated H4 at m/z 11,307), acyl-CoA binding protein, and
calgizzarin, as well as a number of as yet unidentified signals.
Finally, the E-cadherin binding protein fragment and a signal at
m/z 6,081 were also up-regulated in the tumors treated with the
combination.
9097
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
Fig. 4. Drug-induced changes in tumor proteome predict for therapeutic synergy. A. FVB mice bearing large tumors were randomized to no therapy, Herceptin 10 mg/kg i.p. twice
a week, OSI-774 100 mg/kg/day p.o., or the combination of both drugs. Three mice per group were sacrificed 72 hours after treatment initiation, and their tumors were harvested for
the studies below. Mean tumor volumes in mm3⫾ SD were followed in the remaining 10 mice per group as indicated in Materials and Methods. In six of 10 mice receiving both drugs,
tumors were completely eliminated and did not recur off therapy up to 14 months later. B. Immunohistochemical detection of PCNA⫹ and TUNEL⫹ tumor cells as indicated in
Materials and Methods in sections from MMTV/HER2 tumors 3 days after initiation of treatment. Values represent the average; ⫾SD PCNA⫹ and TUNEL⫹ nuclei/1,000 nuclei per
mouse (n ⫽ 3). C. Mass spectral analysis of MMTV/HER2 sections from tumors harvested 72 hours after treatment reveals several proteomic changes induced from the treatment,
including down-regulation of thymosin ␤4 (m/z 4,965) and ubiquitin (m/z 8,565) and up-regulation of m/z 4,796 (E-cadherin binding protein fragment) and m/z 8,719 (calmodulin
peptide). In all cases, the largest changes are observed from the group treated with the combination. D. Response of various protein signals to treatment, plotted as the intensity difference
of the signals (⫾SEM) in treated compared with untreated tumors. The down-regulation of five signals (m/z 6,252, 17,837, 17,877, 15,334, and 11,307, corresponding to an unidentified
protein, peptidyl-prolyl cis-trans isomerase A, acetylated peptidyl-prolyl cis-trans isomerase A, histone H3, and dimethylated, monoacetylated histone H4, respectively) was shown to
be statistically significant (*, ANOVA, P ⬍ 0.05) for the combination therapy compared with controls. Peptidyl-prolyl cis-trans isomerase A was detected in unmodified (a) and
acetylated (b) forms. Histone H4 was detected in dimethylated, diacetylated (c) and dimethylated, monoacetylated (d) forms.
Therapy-Induced Changes in the Proteome Predict for Drug
Resistance. We finally studied if MALDI-MS can be predictive for
drug resistance by comparing the Herceptin-induced protein profile in
F2–1282 tumors versus a second tumor from the same transgenic
founder line (Fo5), which acquired spontaneous resistance to Herceptin. Fo5 tumor cells bind Herceptin and exhibit equal levels of HER2
protein as the F2–1282 tumors.7 Within 10 days of continuous administration of a therapeutic dose of Herceptin (30 mg/kg), all F2–
1282 tumors were eliminated, whereas Fo5 tumors continued to grow
(Fig. 5A). Forty-eight hours after the second dose of antibody, F2–
1282 tumors exhibited a ⬎10-fold increase in TUNEL⫹ cells compared with controls. Consistent with the lack of response, this was not
seen in Fo5 tumors (Fig. 5B).
Fig. 5C shows mass spectra in the mass range of 8,880 to 9,400 Da
obtained for both sensitive (1282) and resistant (Fo5) tumor lines. An
increase in m/z 9,212 is observed in Herceptin-treated 1282 tumors. In
contrast, the same mass range for the resistant tumors shows no
difference in the signal at m/z 9,212 between the untreated and treated
7
R. Schwall, personal communication.
groups, suggesting that m/z 9,212 is a potential biomarker of tumor
response to the HER2 antibody. Several other signals seemed different
between the sensitive and resistant tumors after Herceptin treatment,
and they are presented in Fig. 5D. The signals are plotted as the
absolute intensity difference from untreated tumors (either sensitive or
resistant). As shown, most signals display an increase in intensity on
Herceptin treatment in the sensitive tumors but show no significant
change or a decrease on treatment of resistant tumors. Signals at m/z
9,739, 10,164 (calgranulin A), 4,795 (E-cadherin binding protein
fragment), and 9,212 were determined to be significantly different (t
test, P ⬍ 0.05) in Herceptin-treated sensitive tumors compared with
untreated controls (* in Fig. 5D). Additional studies will be required
to address whether these molecular changes in the sensitive tumors are
causally associated with drug resistance in the Fo5 tumors.
DISCUSSION
The discovery of biomarkers that predict therapeutic response is of
increasing importance in the process of drug development. This need
has prompted investigators into presurgical studies in which drug-
9098
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
Fig. 5. Drug-induced changes in the proteome predict for therapeutic resistance. Mice bearing established ⬎300 mm3 Fo5 (Herceptin-resistant) and 1282 (Herceptin-sensitive) tumors
were treated with Herceptin 30 mg/kg i.p. twice a week. Each data point represents mean tumor volume ⫾ SD (Fo5, n ⫽ 3; 1282, n ⫽ 6). B. Tumor cell apoptosis measured in
representative tumor harvested 24 hours after the second dose of antibody. Data represents mean ⫾ SD TUNEL⫹ tumor cell nuclei in 10 random 400⫻ fields. A statistically significant
increase in apoptosis was observed in 1282 tumor sections (Students t test, P ⫽ 0.0004). C. Fo5 and 1282 tumors of equivalent size were harvested 24 and 48 hours after a single dose
of Herceptin 30 mg/kg i.p. and subjected to mass spectral proteomic analysis. An example of a statistically significant (two-tailed t-test with Welch correction, P ⫽ 0.008) change
observed after Herceptin-treatment in the 1,282 tumors that is not observed in the Fo5 tumors is shown. The sensitive tumor line traces consist of untreated tumors (average of 20 spectra
from 6 tumors) and Herceptin-treated tumors (average of 13 spectra from 4 tumors). The resistant tumor line traces consist of untreated tumors (average of 11 spectra from 3 tumors)
and Herceptin-treated tumors (average of 20 spectra from 4 tumors). D. Response of various protein signals to Herceptin treatment, plotted as the intensity difference of the signals
(⫾SEM) in treated compared with untreated tumors. The up-regulation of four signals (m/z 9,739, 10,164, 4,795, and 9,212, corresponding to an unidentified protein, calgranulin A,
E-cadherin binding protein fragment, and an unidentified protein, respectively) was shown to be statistically significant (*, unpaired t test, P ⬍ 0.05) for 1282 compared with Fo5 tumors.
induced “cellular activity” in situ is assessed after a short therapy
interval. For example, in patients with breast cancer, evidence of
inhibition of proliferation and/or apoptosis in tumor cells as measured
by Ki67 immunohistochemistry or TUNEL, respectively, in a second
biopsy done as early as 10 days after antiestrogens or chemotherapy
has been shown to predict for clinical response to therapy (30). In
these studies, antiestrogen-induced inhibition of proliferation was
limited to estrogen receptor-positive cancers, suggesting that this
approach could have potentially identified estrogen receptor expression as the molecular signature predictive of good odds of response to
antiestrogens. Chang et al. (31) treated locally advanced HER2overexpressing breast cancers with weekly Herceptin followed by the
addition of docetaxel after week 3. Significant tumor regressions at 3
weeks correlated with a 2-fold increase in tumor cell apoptosis, as
measured by cleaved caspase 3 in a biopsy obtained on day 8 after
start of therapy. Additionally, the IMPACT trial randomized women
with estrogen receptor-positive breast cancers to 12 weeks of the
aromatase inhibitor anastrozole, the selective estrogen receptor modulator tamoxifen, or the combination. In a core biopsy obtained after
2 weeks of therapy, a statistically superior inhibition in proliferation,
as measured by Ki67 immunohistochemistry, was observed in tumors
treated with anastrozole compared with the other two arms. This result
predicted for improved clinical outcome as measured by anastrozole
increasing the odds of breast-conserving surgery (32). Taken together,
these data suggest that drug-induced cellular activity in situ can
predict clinical outcome. However, the assays used in these studies are
time consuming, cumbersome, expensive, and require a few days of
treatment to allow sensitive tumor cells to exhibit evidence of drug
action.
Because drug-induced protein changes precede inhibition of proliferation or cell death, we proposed the use of MALDI-MS to detect
therapy-induced early changes in the tumor proteome that would
predict for clinical efficacy. Using MMTV/HER2 transgenic tumors,
we show herein that early changes in tumor protein profiles predict for
dose- and time-dependent effects of signaling inhibitors. Druginduced changes in the proteome also predicted for therapeutic synergy and drug resistance. Finally, in drug-sensitive tumors, the spatial
distribution of the erbB tyrosine kinase inhibitor OSI-774 mapped
with biomarkers of antitumor activity.
These changes were detected in frozen material that did not require
fixation or any additional preparation and occurred as early as 16
hours after a single drug dose. Some of the proteins that changed as
9099
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.
PREDICTION OF THERAPEUTIC RESPONSE WITH MALDI-MS
a function of response may well be causally related to this response.
T␤4 was decreased in tumors treated with the therapeutic dose of
OSI-774 as well as by the combination of OSI-774 and Herceptin.
Interestingly, T␤4 is an actin-sequestering protein involved in cell
migration, angiogenesis, ␤-catenin-mediated transcription, and metastatic progression (22). The association of this change with the inhibitory drug intervention suggests that MALDI-MS may be able to
identify molecular mechanisms associated with the antitumor response. Taken together, these data also illustrate the ability of this
technique to correlate changes in protein expression with drug biodistribution and, therefore, potentially evaluate the effect of molecularly targeted therapeutics on the microenvironment of the tumor.
These features in combination with clinical endpoints and more traditional assays (for apoptosis, cell proliferation, immunohistochemistry, and others) should allow for a more complete understanding of the
biochemical effects of novel therapeutics and thus allow for welldesigned clinical trials and optimal drug development.
In summary, imaging MS can be applied to detect the molecular
fingerprints predictive of the antitumor action of molecular therapies.
This methodology is fast with analysis times of ⬃5 minutes per frozen
unprocessed single tumor section. In addition, it is unbiased, reproducible, and amenable to high throughput screening. The amount of
information that MALDI-MS can provide from a single frozen tumor
section should improve the predictive value of current, more cumbersome antibody-based assays.
REFERENCES
1. Osborne CK, Zhao H, Fuqua SA. Selective estrogen receptor modulators: structure,
function, and clinical use. J Clin Oncol 2000;18:3172– 86.
2. Arteaga CL. Trastuzumab, an appropriate first-line single-agent therapy for HER2overexpressing metastatic breast cancer. Breast Cancer Res 2002;5:96 –100.
3. Heinrich MC, Corless CL, Demetri GD, et al. Kinase mutations and imatinib response
in patients with metastatic gastrointestinal stromal tumor. J Clin Oncol 2003;21:
4342–9.
4. Fukuoka M, Yano S, Giaccone G, et al. A multi-institutional randomized phase II trial
of gefitinib for previously treated patients with advanced non-small-cell lung cancer
(THE IDEAL 1 Trial). J Clin Oncol 2003;21.
5. Giaccone G, Herbst RS, Manegold C, et al. Gefitinib in combination with gemcitabine
and cisplatin in advanced non-small-cell lung cancer: a phase III trial–INTACT 1.
J Clin Oncol 2004;22:777– 84.
6. Paez JG, Janne PA, Lee JC, et al. EGFR mutations in lung cancer: correlation with
clinical response to gefitinib therapy. Science (Wash DC) 2004;304:1497–500.
7. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth
factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib.
N Engl J Med 2004;350:2129 –39.
8. Chaurand P, DaGue BB, Pearsall RS, Threadgill DW, Caprioli RM. Profiling proteins
from azoxymethane-induced colon tumors at the molecular level by matrix-assisted
laser desorption/ionization mass spectrometry. Proteomics 2001;1:1320 – 6.
9. Yanagisawa K, Shyr Y, Xu BJ, et al. Proteomic patterns of tumour subsets in
non-small-cell lung cancer. Lancet 2003;362:433–9.
10. Finkle D, Quan ZR, Asghari V, et al. HER2-targeted therapy reduces incidence and
progression of midlife mammary tumors in female murine mammary tumor virus
huHER2-transgenic mice. Clin Cancer Res 2004;10:2499 –511.
11. Muraoka RS, Dumont N, Ritter CA, et al. Blockade of TGF-beta inhibits mammary
tumor cell viability, migration, and metastases. J Clin Investig 2002;109:1551–9.
12. Lenferink AE, Simpson JF, Shawver LK, et al. Blockade of the epidermal growth
factor receptor tyrosine kinase suppresses tumorigenesis in MMTV/Neu ⫹ MMTV/
TGF-alpha bigenic mice. Proc Natl Acad Sci USA 2000;97:9609 –14.
13. Lenferink AE, Busse D, Flanagan WM, Yakes FM, Arteaga CL. ErbB2/neu kinase
modulates cellular p27(Kip1) and cyclin D1 through multiple signaling pathways.
Cancer Res 2001;61:6583–91.
14. Muraoka RS, Koh Y, Roebuck LR, et al. Increased malignancy of Neu-induced
mammary tumors overexpressing active transforming growth factor beta1. Mol Cell
Biol 2003;23:8691–703.
15. Schwartz SA, Reyzer ML, Caprioli RM. Direct tissue analysis using matrix-assisted
laser desorption/ionization mass spectrometry: practical aspects of sample preparation. J Mass Spectrom 2003;38:699 –708.
16. Reyzer ML, Hsieh Y, Ng K, Korfmacher WA, Caprioli RM. Direct analysis of drug
candidates in tissue by matrix-assisted laser desorption/ionization mass spectrometry.
J Mass Spectrom 2003;38:1081–92.
17. Stoeckli M, Farmer TB, Caprioli RM. Automated mass spectrometry imaging with a
matrix-assisted laser desorption ionization time-of-flight instrument. J Am Soc Mass
Spectrom 1999;10:67–71.
18. Moyer JD, Barbacci EG, Iwata KK, et al. Induction of apoptosis and cell cycle arrest
by CP-358,774, an inhibitor of epidermal growth factor receptor tyrosine kinase.
Cancer Res 1997;57:4838 – 48.
19. Hidalgo M, Siu LL, Nemunaitis J, et al. Phase I and pharmacologic study of OSI-774,
an epidermal growth factor receptor tyrosine kinase inhibitor, in patients with advanced solid malignancies. J Clin Oncol 2001;19:3267–79.
20. Carter P, Presta L, Gorman CM, et al. Humanization of an anti-p185HER2 antibody
for human cancer therapy. Proc Natl Acad Sci USA 1992;89:4285–9.
21. Stoeckli M, Chaurand P, Hallahan DE, Caprioli RM. Imaging mass spectrometry: a
new technology for the analysis of protein expression in mammalian tissues. Nat Med
2001;7:493– 6.
22. Cha HJ, Jeong MJ, Kleinman HK. Role of thymosin b4 in tumor metastasis and
angiogenesis. J Natl Cancer Inst (Bethesda) 2003;95:1674 – 80.
23. Pedersen KB, Nesland JM, Fodstad O, Maelandsmo GM. Expression of S100A4,
E-cadherin, a- and b-catenin in breast cancer biopsies. Br J Cancer 2002;87:1281– 6.
24. Bukholm IK, Nesland JM, Karesen R, Jacobsen U, Borresen-Dale AL. E-cadherin
and a-, b-, and g-catenin protein expression in relation to metastasis in human breast
carcinoma. J Pathol 1998;185:262– 6.
25. Kremer M, Quintanilla-Martinez L, Fuchs M, et al. Influence of tumor-associated
E-cadherin mutations on tumorigenicity and metastasis. Carcinogenesis (Lond) 2003;
24:1879 – 86.
26. Fujita Y, Krause G, Scheffner M, et al. Hakai, a c-Cbl-like protein, ubiquitinates and
induces endocytosis of the E-cadherin complex. Nat Cell Biol 2002;4:222–30.
27. Hochstrasser M. Ubiquitin-dependent protein degradation. Annu Rev Genet 1996;30:
405–39.
28. Chaurand P, Fouchecourt S, DaGue BB, et al. Profiling and imaging proteins in the
mouse epididymis by imaging mass spectrometry. Proteomics 2003;11:2221–39.
29. Moulder SL, Yakes FM, Muthuswamy SK, et al. Epidermal growth factor receptor
(HER1) tyrosine kinase inhibitor (Iressa) inhibits HER2/neu (erbB2)-overexpressing
breast cancer cells in vitro and in vivo. Cancer Res 2001;61:8887–95.
30. Assersohn L, Salter J, Powles TJ, et al. Studies of the potential utility of Ki67 as a
predictive molecular marker of clinical response in primary breast cancer. Breast
Cancer Res Treat 2003;82:113–23.
31. Chang JC, Mohsin S, Weiss H, et al. Induction of apoptosis without change in cell
proliferation in primary breast cancers with neoadjuvant trastuzumab. Breast Cancer
Res Treat 2003;82:S13.
32. Dowsett M, Smith I. Greater Ki67 response after 2 weeks neoadjuvant treatment with
anastrozole (A) than with tamoxifen (T) or anastrozole plus tamoxifen (C) in the
IMPACT trial: a potential predictor of relapse-free survival. Breast Cancer Res Treat
2003;82:S6.
9100
Downloaded from cancerres.aacrjournals.org on October 7, 2014. © 2004 American Association for Cancer
Research.