Magnetic Resonance Spectroscopy Detectable
Metabolomic Fingerprint of Response to Antineoplastic
Alessia Lodi, Sabrina M. Ronen*
Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, California, United States of America
Targeted therapeutic approaches are increasingly being implemented in the clinic, but early detection of response
frequently presents a challenge as many new therapies lead to inhibition of tumor growth rather than tumor shrinkage.
Development of novel non-invasive methods to monitor response to treatment is therefore needed. Magnetic resonance
spectroscopy (MRS) and magnetic resonance spectroscopic imaging are non-invasive imaging methods that can be
employed to monitor metabolism, and previous studies indicate that these methods can be useful for monitoring the
metabolic consequences of treatment that are associated with early drug target modulation. However, single-metabolite
biomarkers are often not specific to a particular therapy. Here we used an unbiased 1H MRS-based metabolomics approach
to investigate the overall metabolic consequences of treatment with the phosphoinositide 3-kinase inhibitor LY294002 and
the heat shock protein 90 inhibitor 17AAG in prostate and breast cancer cell lines. LY294002 treatment resulted in
decreased intracellular lactate, alanine fumarate, phosphocholine and glutathione. Following 17AAG treatment, decreased
intracellular lactate, alanine, fumarate and glutamine were also observed but phosphocholine accumulated in every case.
Furthermore, citrate, which is typically observed in normal prostate tissue but not in tumors, increased following 17AAG
treatment in prostate cells. This approach is likely to provide further information about the complex interactions between
signaling and metabolic pathways. It also highlights the potential of MRS-based metabolomics to identify metabolic
signatures that can specifically inform on molecular drug action.
Citation: Lodi A, Ronen SM (2011) Magnetic Resonance Spectroscopy Detectable Metabolomic Fingerprint of Response to Antineoplastic Treatment. PLoS
ONE 6(10): e26155. doi:10.1371/journal.pone.0026155
Editor: Daniel Monleon, Instituto de Investigacio ´n Sanitaria INCLIVA, Spain
Received June 30, 2011; Accepted September 21, 2011; Published October 12, 2011
Copyright: ? 2011 Lodi, Ronen. 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: This work was supported by National Institutes of Health (NIH; www.nih.gov) grant RO1 CA130819. The funder had no role in study design, data
collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: firstname.lastname@example.org
The Warburg effect, wherein cancer cells have an abnormally
elevated rate of glucose consumption and aerobic glycolysis, was
first discovered in the 1920s . Research over the past decade is
increasingly demonstrating that several other aspects of metabo-
lism are also profoundly different in cancer cells [2,3]. Many of
these changes appear to result from the acquisition of mutations
that develop during oncogenesis and provide a growth advantage
to the cancerous cells in the tumor microenvironment. Knowledge
of these metabolic changes is now being used as the basis for
development of more specific molecular imaging methods. For
instance, the higher glycolytic demand characterizing cancer cells
has been exploited in [18F] 2-fluoro-2-deoxy-D-glucose positron
emission tomography (FDG-PET) imaging of tumors . In recent
years, owing to advances in dynamic nuclear polarization (DNP),
the elevated glycolytic rates in tumors have also been imaged using
13C magnetic resonance spectroscopy (MRS) by probing the
conversion of hyperpolarized pyruvate into lactate [5–7]. Con-
versely, several reports have been published reporting normaliza-
tion of glucose metabolism as an indication of response to targeted
treatment [8,9]. A decrease in pyruvate to lactate conversion in
response to treatment with phosphoinositide 3-kinase (PI3K) or
receptor tyrosine kinase (RTK) inhibitors was shown in different
tumor types by
clinically, total choline (tCho, comprised of choline, PC and
glycerophosphocholine) was also identified in several MR studies
as an important biomarker that is generally elevated in cancer cells
and associated with more aggressive and invasive phenotypes [12–
20]. Inhibition of cell proliferation following treatment with
targeted therapies, including inhibitors of Ras, PI3K, mitogen-
activated protein kinases (MAPK) and hypoxia inducible factor
(HIF) led in most cases to a drop in PC and tCho [21–31]. In
particular, inhibition of HIF-1a with PX-478 in HT-29 colorectal
cancer xenografts induced a drop in intracellular PC levels as well
as tCho . Pharmacological intervention with the PI3K
inhibitor PI-103 induced a drop in PC in PC3 prostate cancer
cells and HCT116 colorectal cancer cells . Similarly in MDA-
MB-231, MCF-7 and Hs578T breast cancer cells treated with
either LY294002 or Wortmannin PC dropped . Finally
orthotopic glioblastoma tumors treated with the PI3K inhibitor
PX-886 also lead to a drop in tCho . Interestingly, treatment
with the heat shock protein (HSP) 90 inhibitor 17-(Allylamino)-17-
demethoxygeldanamycin (17AAG), which likely has a more
complex effect on cellular signaling as it targets a number of
protein kinases (including Akt, MEK and c-Raf) as well as
hormone receptors , was reported to cause an increase in PC
in several cancer models including breast and colorectal [24,25].
13C MRS [10,11]. Phosphocholine (PC) or,
PLoS ONE | www.plosone.org1October 2011 | Volume 6 | Issue 10 | e26155
However, treatment with 17AAG in prostate cancer xenografts in
mice (hormone sensitive CWR22 and hormone resistant CWR22r)
induced a drop in the tCho pool . While demonstrating the
value of metabolic changes as biomarkers of response to targeted
therapies, these studies also highlight the fact that some metabolic
changes are common to different therapeutic agents. An
alternative approach is the use of completely untargeted and
global metabolic profiling methods coupled with a robust
chemometric analysis based on multivariate statistical methods
of analysis. These approaches have been proposed for the
detection and identification of global metabolic changes in human
biofluids as a diagnostic tool [33–35]. Similar approaches have
also been employed in cell model systems to investigate the
metabolic effects of drug treatments or different genetic pheno-
types [36–38]. Recently this approach was also used in studies of
tumor biopsy samples providing a method to distinguish between
normal and malignant tissue in different cancer types [39–43].
Due to the completely untargeted nature of these studies and the
size and complexity of the metabolic signature datasets, the
application of appropriate multivariate statistical methods of
analysis is key to identifying the most prominent changes in the
metabolic signature. In cell model systems principal component
analysis (PCA) is usually appropriate to efficiently identify and
discriminate the underlying metabolic variation in the datasets.
Moreover, PCA is a completely unsupervised method and does not
require any a priori information about the data allowing for a
completely unbiased analysis of the datasets .
Several PI3K inhibitors are currently in clinical trials for cancer
treatment . Similarly, the HSP90 inhibitor 17AAG has been
tested in the clinic for treatment of solid tumors [46,47]. We were
therefore interested in assessing the individual signatures of
response that could potentially be translated into the clinic. In
this study, we investigated two prostate cancer cell lines, PC3 and
LNCaP, and used an untargeted and unbiased1H MRS-based
metabolomics approach to investigate the metabolic consequences
of pharmacological inhibition of the PI3K signaling pathway and
the HSP90 protein chaperone using LY294002 and 17AAG,
respectively. Moreover, to confirm the generality of our findings,
we investigated the metabolic changes induced by LY294002 and
17AAG in MCF-7 breast cancer cells. Based on the analysis of the
comprehensive changes in the metabolome of these cell lines we
identified a pattern of metabolic changes that was different for
each of the two drugs but identical for the three different cell lines.
This approach could provide drug-specific metabolic readouts of
molecular drug action. Furthermore, this method could be used to
identify previously unrecognized metabolic changes associated
with modulation of specific signaling pathways.
Treatment doses and target inhibition in prostate cancer
Two prostate cancer cell lines, PC3 and LNCaP, were treated
for 48 hours with LY294002, a PI3K inhibitor, and 17AAG, a
HSP90 inhibitor. For each cell line, the treatment doses were
determined such that the cell viability remained approximately
constant during treatment (simulating tumor stasis during
treatment, as opposed to the control cells which proliferate
normally). These doses were 25 mM and 10 mM LY294002, and
1 mM and 0.25 mM 17AAG for PC3 and LNCaP prostate cancer
cells, respectively. Inhibition of the target proteins at the
determined doses was confirmed by Western blotting. Following
48 hours of treatment with LY294002, inhibition of the PI3K
signaling pathway was confirmed in both PC3 and LNCaP cells by
probing for p-4E-BP1 protein levels, which decreased following
treatment (Fig. 1). The effectiveness of the 48 hours 17AAG
treatment was verified by probing the levels of the HSP90-client
protein c-Raf, which decreased in both cell lines (Fig. 1).
Moreover, following 17AAG treatment, p-4E-BP1 decreased to
intermediate levels between the control and LY294002 treated
samples (Fig. 1).
Proton MRS-based metabolomics analysis in prostate
1H MR spectra were recorded on the polar fraction of the cell
extracts (8 replicates per treatment per cell line) of PC3 and
LNCaP prostate cancer cells. To investigate the effect of
treatment on the metabolome of the two prostate cancer cell
lines we performed multivariate statistical analysis on the
complete MRS datasets either including both cell lines and all
the treatment conditions or considering each individual cell line.
The scores plots obtained from the analysis (Fig. 2) clearly
highlight that, as expected, the two cell lines have extremely
different phenotypes (more than 90% of the total variability is
explained by the first principal component in Fig. 2A, segregat-
ing the two cell lines). Moreover, the scores plots obtained from
the PCAs performed on the individual cell lines (Fig. 2B–C)
demonstrate excellent clustering of samples within the same
Figure 1. Inhibition of target signaling pathways in cancer cells
following drug treatment. Schematic of signaling pathways
targeted by LY294002 and 17AAG and Western blots showing
modulation of p-4E-BP1 and c-Raf (b-actin as loading control) levels
following administration of DMSO (solvent control; C), LY294002 (L) or
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org2October 2011 | Volume 6 | Issue 10 | e26155
treatment group and clear separation between the 3 different
treatment conditions (dimethyl sulfoxide (DMSO) as solvent
control, LY294002 and 17AAG), for both PC3 and LNCaP cell
lines. In light of the excellent separation indicated by the above
analyses and with the objective of gaining a better understanding
of the metabolic changes underlying the observed differences, we
repeated the PCAs, performing comparisons of the spectra
acquired on untreated (solvent control) samples versus those
obtained from either LY294002 or 17AAG treated samples for
each cell line. As expected, the scores plots obtained from these
PCAs (Fig. 3) indicate the complete separation of solvent control
and each of the treatments along the first principal component
(PC1). The advantage of this analysis resides in the fact that the
PC1 loadings plots capture the metabolic changes specifically
induced by the considered drug treatment (compared to control).
In fact, PC1 explains the large majority of the total variability
(between 63% and 77% depending on the cell line/treatment
considered), while PC2 explains less than 12% (between 9% and
12%) of the total variability.
The analysis of the loadings plots (Fig. 4) obtained from each
of the PCAs revealed discriminatory metabolites for the different
cell lines/treatments. Specifically, the loadings plots obtained
from the comparison of solvent control and LY294002 treatments
indicated the decrease of intracellular alanine, lactate, fumarate,
glutathione and phosphocholine concentrations following treat-
ment in both PC3 and LNCaP prostate cancer cells. The
administration of LY294002 also induced the accumulation of
branched amino acids (valine, leucine, isoleucine) and glutamine
in both prostate cancer cells. Some disparate changes among the
two prostate cancer cells were also observed, including the
decrease of taurine, myo-inositol and uridine diphosphate (UDP)-
glucose in PC3 cells and asparagine and glycine in LNCaP cells,
and the accumulation of glucose, glycine, phenylalanine, tyrosine
and histidine in PC3 cells and UDP-N-acetyl-glucosamine, UDP-
N-acetyl-galactosamine, creatine, phosphocreatine, choline, gly-
cerophosphocholine, taurine, myo-inositol and citrate in LNCaP
In the case of 17AAG treatment, a decrease in lactate, alanine,
fumarate and glutamine, and the intracellular accumulation of
valine, leucine, isoleucine, phosphocholine, myo-inositol, taurine
and citrate were observed in both PC3 and LNCaP cells. It is
worth noting that citrate represents the largest peak in the loadings
plot, Fig. 4B, for PC3 cells treated with 17AAG. This metabolite
was virtually undetected in untreated and LY294002-treated PC3
cells, but accumulated in substantial amounts following 17AAG
treatment, as depicted in Fig. 5.
Other metabolic changes induced by treatment with 17AAG
included the intracellular accumulation of glucose, glycine,
phenylalanine, tyrosine and hystidine in PC3 and accumulation
of creatine, phosphocreatine, choline and glycerophosphocholine
in LNCaP cells. Concurrently glutathione and UDP-glucose
decreased in PC3 and asparagine, glycine, UDP-N-acetyl-
LNCaP cells. Interestingly, our results indicated that several
metabolic changes in response to treatment with LY294002 and
17AAG were common to both prostate cancer cells.
Targeted analysis of proton MR spectra of breast cancer
MCF-7 cells were treated with 25 mM LY294002 or 3 mM
17AAG. Similarly to the prostate samples, these doses were
previously determined such that they induced inhibition of target
proteins and inhibition of cell growth [24,48].
To confirm the generality of our findings in prostate cells we
next conducted a targeted analysis of the metabolic consequences of
LY294002 and 17AAG treatment in the MCF-7 cells. We focused
on the common metabolic changes detected in the prostate
samples when using the untargeted approach and therefore
quantified the modulations in the intracellular levels of lactate,
alanine, fumarate, phosphocholine, glutamine and glutathione. In
complete agreement with the results reported above for the
prostate cells, treatment with LY294002 induced the decrease of
intracellular alanine, lactate, fumarate, glutathione and phospho-
choline in MCF-7 breast cancer cells. Similarly, the decrease of
lactate, alanine, fumarate and glutamine, and the accumulation of
phosphocholine induced by the 17AAG treatment in PC3 and
LNCaP cells was also confirmed in MCF-7 breast cancer cells.
Quantification of the main metabolic changes is summarized in
In this study we investigated the global metabolic effects
induced in PC3 and LNCaP prostate cancer cell lines through
the pharmacological intervention with LY294002, a PI3K
inhibitor, and 17AAG, an HSP90 chaperone-function inhibi-
tor. We characterized the metabolic fingerprints of the prostate
cancer cells with and without treatment using
then determined the metabolic changes associated with
response to each treatment using a completely untargeted and
unbiased multivariate statistical approach (PCA). Our aim was
to identify the commonalities in the two cell lines of the
1H MRS and
Figure 2. Multivariate statistical analysis of the MR spectra. Scores plots (PC1 vs PC2) obtained by performing PCA on the MR spectra
acquired on polar extracts (8 replicates per treatment condition) of (A) both PC3 and LNCaP cells, and individual (B) PC3 and (C) LNCaP prostate
cancer cells following a 48-hrs treatment with DMSO (solvent control, black), LY294002 (green) and 17AAG (red).
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org3October 2011 | Volume 6 | Issue 10 | e26155
metabolic signatures associated with response to treatment for
each of the inhibitors. A common metabolic profile was
observed in our studies that indicates that several metabolites
are simultaneously modulated following treatment with each of
the inhibitors. Moreover, we also confirmed the generality of
the two treatment-specific metabolic signatures in a breast
cancer cell line (MCF-7).
When considering the two inhibitors, both LY294002 and
17AAG have the ability to affect signaling via the PI3K pathway,
as indicated by decreased p-4E-BP1 levels downstream of
mammalian target of rapamycin (mTOR). The results of the
metabolic profiling indicate that regardless of the inherently
different metabolic fingerprint and genetic background of the two
prostate and one breast cancer cell lines, changes in the metabolic
signatures induced by the drug treatments demonstrate common-
alities which are probably associated with the PI3K/Akt pathway
and downstream modulation of HIF-1a. Both drug treatments
induce the intracellular depletion of lactate and alanine in all 3 cell
lines. Lactate dropped between 30% and 65% after LY294002
treatment and between 17% and 35% after 17AAG treatment,
depending on the cell line. Alanine dropped between 20% and
50% with LY294002 and between 27% and 58% with 17AAG
depending on the cell line. These results points to hindered
glycolysis upon administration of the drug treatments and is
consistent with the known activation of glucose uptake and
glycolysis by PI3K/Akt signaling . Intracellular fumarate also
decreased (between 43% and 72% with LY294002 and between
36% and 88% with 17AAG treatment) in all prostate and breast
cancer cells after the administration of either drug indicating that
it might be a key mediator of this response. In fact, it has been
previously reported that the inhibition of fumarate hydratase (FH)
and the associated accumulation of intracellular fumarate coincide
with HIF upregulation .
In contrast, other metabolites showed a ‘‘drug-specific’’
behavior. Phosphocholine dropped (13 to 50% depending on the
cell line) following PI3K inhibition but increased (between 10 and
83%) following HSP90 inhibition, in line with previous studies.
Phosphocholine was previously reported to decrease following
treatment with LY294002 and wortmannin (another PI3K
inhibitor; ) and increase following 17AAG treatment in
human breast and colon cancer cell lines [24,25].
The modulation of glutamine was also drug specific: it
increased (approximately 25% in the 2 prostate cancer cell lines)
or stayed constant following PI3K inhibition and dropped (20–
Figure 3. Multivariate statistical analysis of the MR spectra. Scores plots (PC1 vs PC2) obtained by performing PCA on the MR spectra
acquired on polar extracts (8 replicates per treatment condition) of PC3 and LNCaP prostate cancer cells. Control samples were compared to samples
treated for 48 hours with either LY294002 ((A) for PC3 and (C) for LNCaP cells) or 17AAG ((B) for PC3 and (D) for LNCaP cells).
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org4 October 2011 | Volume 6 | Issue 10 | e26155
30%) following HSP90 inhibition. To the best of our knowledge,
these findings have not been previously reported. Whereas
further studies are needed to fully understand the underlying
mechanism of these changes, these observations are in line with
the potentially fundamental role of glutaminolysis in cancer cell
growth and the molecular/metabolic links reported between Myc
and both glutaminase expression and glutamine-uptake regula-
The observation that the intracellular concentration of citrate
is increased following 17AAG treatment is highly significant in
the context of prostate cancer as it might indicate a specific shift
towards a more physiologic metabolism of prostate cancer cells.
Citrate is known to be physiologically present in large amounts
in the healthy prostate. However, citrate levels dramatically
drop upon development of prostate cancer. Although 17AAG
has not shown significant efficacy in a phase II clinical trial in
prostate cancer patients, it is possible that the use of this drug
combined with other agents would contribute to improving
outcome by mediating a normalization of metabolism in
prostate cancer cells.
In conclusion, this research highlights the potential of MRS-
based untargeted metabolomics. Using this approach we identified
a possible global metabolic signature associated with PI3K and
HSP90 inhibition. This study not only confirmed previous work
Figure 4. Multivariate statistical analysis of the MR spectra. Loadings plots (on PC1) obtained by performing the PCA comparisons on the MR
spectra of control and one treatment per analysis (as shown in Fig. 3) acquired on polar extracts of PC3 (black line) and LNCaP (red line) prostate
cancer cells following 48 hours of treatment with (A) LY294002 or (B) 17AAG. Enlarged sections of the loadings plots represent the region of 1.9–4.1
ppm. Ala: alanine; Asn: asparagine; Cho: choline; Cit: citrate; Cre: creatine; Fum: fumarate; Glc: glucose; Gln: glutamine; Gly: glycine; GPcho:
glycerophosphocholine; GSH: glutathione; His: histidine; Ile: isoleucine; Lac: lactate; Leu: leucine; m-Ino: myo-inositol; Pcho: phosphocholine; Pcre:
phosphocreatine; Phe: phenylalanine; Tau: taurine; Tyr: tyrosine; Val: valine, UDPS: UDP sugars.
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org5 October 2011 | Volume 6 | Issue 10 | e26155
but also identified previously unknown putative links between
signaling and metabolic pathways. Once individual metabolic
changes are validated through detailed mechanistic studies, a
combination of metabolic alterations could be envisaged which
would provide a more specific signature of response than single
metabolite biomarkers. Additional studies are under way in our
lab in animal xenograft models of breast and prostate cancer to
verify and validate in vivo the response signature to treatment with
PI3K and HSP90 inhibitors using high resolution magic angle
spinning (HR-MAS) MRS and MR spectroscopic imaging. The
potential value of HR-MAS MR analysis of tumor biopsies to
predict long-term survival and evaluate response to treatment has
been previously reported . In the long term, knowledge of the
expected metabolic signature of response to targeted therapies
could lead to the development of automated decision-support tools
based on in vivo noninvasive patient MRS data similar to the
approach developed in the context of the INTERPRET project
(http://gabrmn.uab.es/INTERPRET) wherein the MRS signa-
ture is proposed as a tool to assist in the diagnosis and grading of
brain tumors and other abnormal brain masses [55,56].
Ultimately, this could lead to specific non-invasive methods for
monitoring response in the in vivo clinical setting.
Figure 5. Accumulation of citrate following treatment with
17AAG in PC3 prostate cancer cells. Enlarged section (2.49 – 2.72
ppm) of the MR spectra acquired on polar extracts of PC3 cells
following 48 hours of treatment with DMSO (solvent control, black),
LY294002 (green) and 17AAG (red). Spectra were normalized according
to the probabilistic quotient normalization method.
Figure 6. Common metabolic changes in prostate and breast cancer cells following drug treatment. Quantification of selected
metabolites (shown as percent of control, mean 6 standard deviation) from MR spectra acquired on prostate (PC3 and LNCaP, N=8) and breast
(MCF-7, N=3) cancer cell lines following 48 hours of treatment with (A) LY294002 or (B) 17AAG. *: p,0.05; **: p,0.005; ***: p,0.0005. Pcholine:
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org6 October 2011 | Volume 6 | Issue 10 | e26155
Materials and Methods
Cell culture and treatments
PC3, LNCaP (prostate) and MCF-7 (breast) cancer cell lines
were obtained from American Type Culture Collection via
University of California San Francisco (UCSF) Cell Culture
Facility (San Francisco, CA, USA) and were maintained in
exponential proliferation in Dulbecco’s Modified Eagle Medium
(DMEM) supplemented with 10% heat-inactivated fetal bovine
serum, 2 mM L-glutamine, 100 units ml21of penicillin and
100 mg ml21of streptomycin. The cells were cultured in a
humidified chamber at 37uC and with 5% CO2.
For all the experiments cells were incubated with drug for
48 hours as follows. PC3 cells with 25 mM LY294002 (PI3K
inhibitor) and 1 mM 17AAG (HSP90 inhibitor), LNCaP cells with
10 mM LY294002 and 0.25 mM 17AAG and MCF-7 cells with
25 mM LY294002 and 3 mM 17AAG. The treatment doses were
determined such that they decreased cell viability to approxi-
mately 50% of solvent control after 48 hours of treatment by
using the cell proliferation assay detailed below. All treatments
were performed with matching DMSO solvent control (1:1000
final concentration in culture medium) and were replenished after
Cell viability assay
The effect of different drug treatment doses on cell viability was
determined using the WST-1 reagent assay (Roche). Cells were
seeded in 96-well plates and treated for 4 to 48 hours with 4
different treatment doses for each drug. After treatment, WST-1
reagent was incubated in wells for approximately 1 hour and cell
viability was determined by spectrophotometric (Tecan) quantifi-
cation of absorbance at 440 nm.
The effect of treatment with LY294002 and 17AAG on the
levels of target proteins was analyzed by Western blotting. Whole
cell lysates from treated and untreated cells were separated on 4%
to 20% SDS-PAGE gels (Bio-Rad). Proteins were then transferred
onto nitrocellulose membranes, blocked and incubated with
primary and secondary (anti-IgG horseradish peroxidase-linked,
Cell Signaling) antibodies. Primary antibodies against p-4E-BP1,
c-Raf and b-actin (as loading control) (Cell Signaling) were used.
Immunocomplexes were visualized using enhanced chemilumi-
nescence (ECL) Western Blotting Substrate (Pierce).
MR sample preparation
In all cases, cells were extracted using a dual phased extraction,
as described previously . The extraction of MCF-7 breast
cancer cells (3 replicates) was previously described in detail
[24,48]. In the case of PC3 or LNCaP prostate cancer cells,
following 48 hour treatments, approximately 56106cells were
washed twice with phosphate buffered saline (PBS) in the tissue
culture flask and then fixed using ice-cold methanol. Cells were
then scraped off and transferred with the methanol to a glass
centrifuge tube. Chloroform and water were then added to the
methanol in equal volumes (final solution 1:1:1 methanol:chlor-
oform:water). The solution was vortexed and centrifuged to
separate the aqueous and lipid phases. The two phases were then
collected separately and dried. The dried polar extracts (8
replicates per treatment per cell line) were then redissolved in
600 ml of 100 mM phosphate buffer (pH 7.0) prepared in 90%
H2O - 10% D2O and containing 0.5 mM sodium 3-(trimethylsi-
lyl)propionate-2,2,3,3-d4 (TMSP, Cambridge Isotope Laborato-
ries) as internal reference.
MR data acquisition and processing
One dimensional (1D)
performed on the aqueous fraction of the PC3 and LNCaP cell
extracts using a 600 MHz spectrometer equipped with a
cryogenically cooled probe. 90u pulse and 4 s relaxation delay
were used and the water resonance was suppressed using
excitation sculpting . The acquisition of1H MR data from
MCF-7 cells extracts was previously described [24,48].
All the MRS datasets were processed using NMRLab  in
the MATLAB programming environment (The MathWorks,
Inc.). Following standard processing steps, spectra were aligned,
selected signals arising from residual solvents (water, methanol
and chloroform) and from TMSP were excluded. Spectra were
normalized according to the probabilistic quotient method .
Spectra acquired on PC3 and LNCaP cell samples were then
binned at approximately 0.0017 ppm and the generalized-log
transformation was applied prior to conducting the multivariate
statistical analysis . Principal component analysis (PCA) of the
complete PC3 and LNCaP extract MRS datasets was carried out
using MATLAB. For all datasets, MRS resonances of metabolites
were assigned by comparison with spectra of standard com-
pounds (www.bml-nmr.org) and the peak integrals of selected
metabolites were calculated using ACD/Spec Manager version
9.15 software (Advanced Chemistry Development) for relative
quantification. Data are reported as mean values 6 standard
deviation. It should be noted that the MRS datasets were
acquired under slightly different conditions for the breast and the
prostate cancer samples. However, for each cell line the spectra of
treated samples were always acquired under the exact same
conditions as the matching control samples. The relative
quantifications (as percent change when treatment is compared
to control) are thereby unaffected by the differing acquisition
conditions. For the prostate cancer samples (N=8 per treatment
condition and per cell line) statistical significance was determined
using a Mann-Whitney U test with p,0.05 considered signifi-
cant. For the breast cancer samples (N=3) which were used to
confirm the trends observed in the prostate samples, statistical
significance was assessed using a one-sided Mann-Whitney U test
with p,0.05 considered significant.
1H MR spectra acquisition was
We wish to thank Judy Su and Alissa Brandes for providing MRS data.
Conceived and designed the experiments: AL SMR. Performed the
experiments: AL. Analyzed the data: AL. Wrote the paper: AL SMR.
1. Warburg O (1956) On the origin of cancer cells. Science 123: 309–314.
2. Hsu PP, Sabatini DM (2008) Cancer cell metabolism: Warburg and beyond.
Cell 134: 703–707.
3. Kaelin WG, Thompson CB (2010) Q&A: Cancer: Clues from cell metabolism.
Nature 465: 562–564.
4. Jones RG, Thompson CB (2009) Tumor suppressors and cell metabolism: A
recipe for cancer growth. Genes Dev 23: 537–548.
5. Albers MJ, Bok R, Chen AP, Cunningham CH, Zierhut ML, et al. (2008)
Hyperpolarized 13C lactate, pyruvate, and alanine: Noninvasive biomark-
ers for prostate cancer detection and grading. Cancer Res 68: 8607–
6. Day SE, Kettunen MI, Gallagher FA, Hu D-E, Lerche M, et al. (2007) Detecting
tumor response to treatment using hyperpolarized 13C magnetic resonance
imaging and spectroscopy. Nat Med 13: 1382–1387.
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org7 October 2011 | Volume 6 | Issue 10 | e26155
7. Seth P, Grant A, Tang J, Vinogradov E, Wang X, et al. (2001) On-target
inhibition of tumor fermentative glycolysis as visualized by hyperpolarized
pyruvate. Neoplasia 13: 60–71.
8. Stroobants S, Goeminne J, Seegers M, Dimitrijevic S, Dupont P, et al. (2003)
18FDG-Positron emission tomography for the early prediction of response in
advanced soft tissue sarcoma treated with imatinib mesylate (Glivec).
Eur J Cancer 39: 2012–2020.
9. Van den Abbeele AD, Ertuk M (2008) FDG-PET to measure response to
targeted therapy: The example of gastrointestinal stromal tumor and imatinib
mesylate (Gleevec). PET Clinics 3: 77–87.
10. Ward CS, Venkatesh HS, Chaumeil MM, Brandes AH, VanCriekinge M, et al.
(2010) Noninvasive detection of target modulation following phosphatidylinositol
3-kinase inhibition using hyperpolarized 13C magnetic resonance spectroscopy.
Cancer Res 70: 1296–1305.
11. Dafni H, Larson PEZ, Hu S, Yoshihara HAI, Ward CS, et al. (2010)
Hyperpolarized 13C spectroscopic imaging informs on hypoxia-inducible factor-
1 and Myc activity downstream of platelet-derived growth factor receptor.
Cancer Res 70: 7400–7410.
12. Glunde K, Ackerstaff E, Mori N, Jacobs MA, Bhujwalla ZM (2006) Choline
phospholipid metabolism in cancer: Consequences for molecular pharmaceutical
interventions. Mol Pharm 3: 496–506.
13. Jacobs MA, Barker PB, Bottomley PA, Bhujwalla Z, Bluemke DA (2004) Proton
magnetic resonance spectroscopic imaging of human breast cancer: A
preliminary study. J Magn Reson Imaging 19: 68–75.
14. Jagannathan NR, Singh M, Govindaraju V, Raghunathan P, Coshic O, et al.
(1998) Volume localized in vivo proton MR spectroscopy of breast carcinoma:
Variation of water-fat ratio in patients receiving chemotherapy. NMR Biomed
15. Kurhanewicz J, Vigneron DB, Hricak H, Narayan P, Carroll P, et al. (1996)
Three-dimensional H-1 MR spectroscopic imaging of the in situ human prostate
with high (0.24-0.1-cm(3)) spatial resolution. Radiology 198: 795–805.
16. Mori N, Delsite R, Natarajan K, Kulawiec M, Bhujwalla ZM, et al. (2004) Loss
of p53 function in colon cancer cells results in increased phosphocholine and
total choline. Mol Imaging 3: 319–323.
17. Nakagami K, Uchida T, Ohwada S, Koibuchi Y, Suda Y, et al. (1999) Increased
choline kinase activity and elevated phosphocholine levels in human colon
cancer. Japanese Journal Cancer Research 90: 419–424.
18. Roebuck JR, Cecil KM, Schnall MD, Lenkinski RE (1998) Human breast
lesions: Characterization with proton MR spectroscopy. Radiology 209:
19. Ross B, Michaelis T (1994) Clinical applications of magnetic resonance
spectroscopy. Magn Reson Q 10: 191–247.
20. Yeung DKW, Cheung HS, Tse GMK (2001) Human breast lesions:
Characterization with contrast-enhanced in vivo proton MR spectroscopy -
Initial results. Radiology 220: 40–46.
21. Al-Saffar NMS, Jackson LE, Raynaud FI, Clarke PA, de Molina AR, et al.
(2010) The phosphoinositide 3-kinase inhibitor PI-103 downregulates choline
kinase alpha leading to phosphocholine and total choline decrease detected by
magnetic resonance spectroscopy. Cancer Res 70: 5507–5517.
22. Beloueche-Babari M, Jackson LE, Al-Saffar NMS, Eccles SA, Raynaud FI, et al.
(2006) Identification of magnetic resonance detectable metabolic changes
associated with inhibition of phosphoinositide 3-kinase signaling in human
breast cancer cells. Mol Cancer Ther 5: 187–196.
23. Beloueche-Babari M, Peak JC, Jackson LE, Tiet MY, Leach MO, et al. (2009)
Changes in choline metabolism as potential biomarkers of phospholipase C
gamma 1 inhibition in human prostate cancer cells. Mol Cancer Ther 8:
24. Brandes A, Ward C, Ronen SM (2010) 17-allyamino-17-demethoxygeldanamy-
cin treatment results in a magnetic resonance spectroscopy-detectable elevation
in choline-containing metabolites associated with increased expression of choline
transporter SLC44A1 and phospholipase A2. Breast Cancer Res 12: R84.
25. Chung YL, Troy H, Banerji U, Jackson LE, Walton MI, et al. (2003) Magnetic
resonance spectroscopic pharmacodynamic markers of the heat shock protein 90
inhibitor 17-allylamino,17-demethoxygeldanamycin (17AAG) in human colon
cancer models. J Natl Cancer Inst 95: 1624–1633.
26. Jordan BF, Black K, Robey IF, Runquist M, Powis G, et al. (2005) Metabolite
changes in HT-29 xenograft tumors following HIF-1a inhibition with PX-478 as
studied by MR spectroscopy in vivo and ex vivo. NMR Biomed 18: 439.
27. Le HC, Lupu M, Kotedia K, Rosen N, Solit D, et al. (2009) Proton MRS detects
Metabolic Changes in Hormone Sensitive and Resistant Human Prostate
Cancer Model CWR22 and CWR22r. Magn Reson Med 62: 1112–1119.
28. Romanska HM, Tiziani S, Howe RC, Gunther UL, Guizar Z, et al. (2009)
Nuclear magnetic resonance detects phosphoinositide 3-Kinase/Akt-indepen-
dent traits common to pluripotent murine embryonic stem cells and their
malignant counterparts. Neoplasia 11: 1301–1308.
29. Ronen SM, Jackson LE, Beloueche M, Leach MO (2001) Magnetic resonance
detects changes in phosphocholine associated with Ras activation and inhibition
in NIH 3T3 cells. Br J Cancer 84: 691–696.
30. Sankaranarayanapillai M, Tong WP, Yuan Q, Bankson JA, Dafni H, et al.
(2008) Monitoring histone deacetylase inhibition in vivo: Noninvasive magnetic
resonance spectroscopy method. Mol Imaging 7: 92–100.
31. Koul D, Shen R, Kim YW, Kondo Y, Lu Y, et al. (2010) Cellular and in vivo
activity of a novel PI3K inhibitor, PX-866, against human glioblastoma. Neuro
Oncology 12: 559–569.
32. Goetz MP, Toft DO, Ames MM, Erlichman C (2003) The Hsp90 chaperone
complex as a novel target for cancer therapy. Annals of Oncology 14:
33. Clayton AT, Lindon JC, Cloarec O, Antti H, Charuel C, et al. (2006)
Pharmaco-metabonomic phenotyping and personalized drug treatment. Nature
34. Holmes E, Loo RL, Stamler J, Bictash M, Yap IKS, et al. (2008) Human
metabolic phenotype diversity and its association with diet and blood pressure.
Nature 453: 396–400.
35. Tiziani S, Lopes V, Gu ¨nther UL (2009) Early stage diagnosis of oral cancer
using 1H-NMR-based metabolomics. Neoplasia 11: 269–276.
36. Peet AC, McConville C, Wilson M, Levine BA, Reed M, et al. (2007) H-1 MRS
identifies specific metabolite profiles associated with MYCN-amplified and non-
amplified tumour subtypes of neuroblastoma cell lines. NMR Biomed 20:
37. Tiziani S, Lodi A, Khanim FL, Viant MR, Bunce CM, et al. (2009)
Metabolomic profiling of drug responses in acute myeloid leukaemia cell lines.
PLoS One 4: e4251.
38. Lodi A, Tiziani S, Khanim FL, Drayson MT, Gu ¨nther UL, et al. (2011)
Hypoxia triggers major metabolic changes in AML cells without altering
indomethacin-induced TCA cycle deregulation. ACS Chem Biol 6: 169–175.
39. Bathen TF, Jensen LR, Sitter B, Fjoesne HE, Halgunset J, et al. (2007) MR-
determined metabolic phenotype of breast cancer in prediction of lymphatic
spread, grade, and hormone status. Breast Cancer Res Treat 104: 181–189.
40. Bathen TF, Sitter B, Sjobakk TE, Tessem MB, Gribbestad IS (2010) Magnetic
resonance metabolomics of intact tissue: a biotechnological tool in cancer
diagnostics and treatment evaluation. Cancer Res 70: 6692–6696.
41. Chan ECY, Koh PK, Mal M, Cheah PY, Eu KW, et al. (2009) Metabolic
profiling of human colorectal cancer using high-resolution magic angle spinning
nuclear magnetic resonance (HR-MAS NMR) spectroscopy and gas chroma-
tography mass spectrometry (GC/MS). J Proteome Res 8: 352–361.
42. Cheng LL, Burns MA, Taylor JL, He WL, Halpern EF, et al. (2005) Metabolic
characterization of human prostate cancer with tissue magnetic resonance
spectroscopy. Cancer Res 65: 3030–3034.
43. Sitter B, Bathen T, Hagen B, Arentz C, Skjeldestad FE, et al. (2004) Cervical
cancer tissue characterized by high-resolution magic angle spinning MR
spectroscopy. MAGMA Magn Reson Mater Phys Biol Med 16: 174–181.
44. Schnackenberg LK, Beger RD, Dragan YP (2009) NMR-based metabonomics
in toxicology research In: Vaidyanathan S, Harrigan GG, Goodacre RE, eds.
Metabolome Analyses, Strategies for Systems Biology. Springer. pp 159–172.
45. Courtney KD, Corcoran RB, Engelman JA (2010) The PI3K Pathway As Drug
Target in Human Cancer. Journal of Clinical Oncology 28: 1075–1083.
46. Heath EI, Gaskins M, Pitot HC, Pili R, Tan W, et al. (2005) A Phase II Trial of
17-Allylamino-17-Demethoxygeldanamycin in Patients with Hormone-Refrac-
tory Metastatic Prostate Cancer. Clinical Prostate Cancer 4: 138–141.
47. Solit DB, Ivy SP, Kopil C, Sikorski R, Morris MJ, et al. (2007) Phase I Trial of
17-Allylamino-17-Demethoxygeldanamycin in Patients with Advanced Cancer.
Clinical Cancer Research 13: 1775–1782.
48. Su JS, Woods SM, Ronen SM (2011) Metabolic Consequences of PI3K/AKT
signaling in MCF-7 Breast Cancer Cells. NMR Biomed: doi: 10.1002/
49. Elstrom RL, Bauer DE, Buzzai M, Karnauskas R, Harris MH, et al. (2004) Akt
stimulates aerobic glycolysis in cancer cells. Cancer Res 64: 3892–3899.
50. Isaacs JS, Jung YJ, Mole DR, Lee S, Torres-Cabala C, et al. (2005) HIF
overexpression correlates with biallelic loss of fumarate hydratase in renal
cancer: Novel role of fumarate in regulation of HIF stability. Cancer cell 8:
51. Gao P, Tchernyshyov I, Chang T-C, Lee Y-S, Kita K, et al. (2009) c-Myc
suppression of miR-23a/b enhances mitochondrial glutaminase expression and
glutamine metabolism. Nature 458: 762–765.
52. Regan PL, Jacobs J, Wang G, Torres J, Edo R, et al. (2011) Hsp90 inhibition
increases p53 expression and destabilizes MYCN and MYC in neuroblastoma.
Int J Oncol 38: 105–112.
53. Wise DR, DeBerardinis RJ, Mancuso A, Sayed N, Zhang X-Y, et al. (2008) Myc
regulates a transcriptional program that stimulates mitochondrial glutaminolysis
and leads to glutamine addiction. Proc Natl Acad Sci USA 105: 18782–18787.
54. Cao MD, Sitter B, Bathen TF, Bofin A, Lønning PE, et al. (2011) Predicting
long-term survival and treatment response in breast cancer patients receiving
neoadjuvant chemotherapy by MR metabolic profiling. NMR Biomed: doi:
55. Perez-Ruiz A, Julia-Sape M, Mercadal G, Olier I, Majos C, et al. (2010) The
INTERPRET Decision-Support System version 3.0 for evaluation of Magnetic
Resonance Spectroscopy data from human brain tumours and other abnormal
brain masses. BMC Bioinformatics 11: 581.
56. Tate AR, Underwood J, Acosta DM, Julia `-Sape ´ M, Majo ´s C, et al. (2006)
Development of a decision support system for diagnosis and grading of brain
tumours using in vivo magnetic resonance single voxel spectra. NMR Biomed
57. Hwang TL, Shaka AJ (1995) Water suppression that works - Excitation sculpting
using arbitrary wave-forms and pulsed-field gradients. J Magn Reson 112:
58. Gu ¨nther UL, Ludwig C, Ruterjans H (2000) NMRLAB - Advanced NMR data
processing in MATLAB. J Magn Reson 145: 201–208.
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org8 October 2011 | Volume 6 | Issue 10 | e26155
59. Dieterle F, Ross A, Schlotterbeck G, Senn H (2006) Probabilistic quotient Download full-text
normalization as robust method to account for dilution of complex biological
mixtures. Application in H-1 NMR metabonomics. Anal Chem 78: 4281–4290.
60. Parsons HM, Ludwig C, Gu ¨nther UL, Viant MR (2007) Improved classification
accuracy in 1-and 2-dimensional NMR metabolomics data using the variance
stabilising generalised logarithm transformation. BMC Bioinformatics 8: 234.
MRS Metabolomic Fingerprint of Treatment Response
PLoS ONE | www.plosone.org9 October 2011 | Volume 6 | Issue 10 | e26155