A metabolomics investigation into the effects of
HIV protease inhibitors on HPV16 E6 expressing
cervical carcinoma cells†
Dong-Hyun Kim,‡§a J. William Allwood,§¶a Rowan E. Moore,b
Emma Marsden-Edwards,8b Warwick B. Dunn,¶a Yun Xu,a Lynne Hampson,c
Ian N. Hampsonc and Royston Goodacre*ad
Recently, it has been reported that anti-viral drugs, such as indinavir and lopinavir (originally targeted for
HIV), also inhibit E6-mediated proteasomal degradation of mutant p53 in E6-transfected C33A cells. In
order to understand more about the mode-of-action(s) of these drugs the metabolome of HPV16 E6
expressing cervical carcinoma cell lines was investigated using mass spectrometry (MS)-based metabolic
profiling. The metabolite profiling of C33A parent and E6-transfected cells exposed to these two anti￾viral drugs was performed by ultra performance liquid chromatography (UPLC)-MS and gas
chromatography (GC)-time of flight (TOF)-MS. Using a combination of univariate and multivariate
analyses, these metabolic profiles were investigated for analytical and biological reproducibility and to
discover key metabolite differences elicited during anti-viral drug challenge. This approach revealed
both distinct and common effects of these two drugs on the metabolome of two different cell lines.
Finally, intracellular drug levels were quantified, which suggested in the case of lopinavir that increased
activity of membrane transporters may contribute to the drug sensitivity of HPV infected cells.
Introduction
Cervical cancer is the major gynecological cancer among women
diagnosed in the UK. Each year in the UK, over 2800 women are
diagnosed and approximately 1000 deaths are caused by this cancer
(UK Cervical Cancer Statistics, Cancer Research UK. www.cancerre
searchuk.org). The global figures are even more astounding, with an
estimated 473 000 women affected by cervical cancer and 253 500
deaths each year (National Cervical Cancer Coalition. www.nccc￾online.org). Indeed in many low resource countries it is the greatest
cause of women’s cancer-related mortality.
Human papilloma virus (HPV) is the major cause of cervical
cancer1 and there are over 100 different types of HPV associated
with a variety of clinical lesions with approximately 20 of these
being associated with anogenital tract lesions.2 Of these HPVs,
the high-risk types (e.g., HPV16 and 18) are more often found in
association with pre-malignant cervical lesions and invasive
cancers.1,3 HPV16 and 18 are the most widespread high-risk
types associated with cervical cancer, accounting for over 60%
of cases, although there are 11 other high risk types reported.4,5
Although anti-HPV vaccines have been implemented these
generally only cover the high risk types (16 and 18) which mean
that there is still a significant proportion of other high risk
HPV-related cervical disease that will not be protected by this
strategy. In addition, since there are many women persistently
infected with high risk types of HPV and cervical cancer can take
from 10–20 years to develop, alternative therapies are required for
preventing HPV infection. Whilst surgery has been employed
widely for the treatment of HPV related pre-cancerous cervical
intraepithelial neoplasia (CIN),4,6 most surgical procedures of this
type carry an increased risk of infertility, which leads to a need
for simple, preferably self-administered non-surgical therapy
providing several advantages such as better preservation of
a School of Chemistry, Manchester Institute of Biotechnology, The University of
Manchester, 131 Princess Street, Manchester, M1 7DN, UK.
E-mail: [email protected]; Fax: +44 (0)161 3064519;
Tel: +44 (0)161 3064480 b Waters Corporation, Atlas Park, Simonsway, Manchester, M22 5PP, UK
The University of Manchester, Gynaecological Oncology Laboratories,
Human Development, St Mary’s Hospital, Manchester, M13 OJH, UK
d Manchester Centre for Integrative Systems Biology (MCISB), Manchester Institute
of Biotechnology, The University of Manchester, 131 Princess Street, Manchester,
M1 7DN, UK
† Electronic supplementary information (ESI) available. See DOI: 10.1039/
c3mb70423h
‡ Current address: Centre for Analytical Bioscience, School of Pharmacy,
University of Nottingham, University Park, Nottingham, NG7 2RD, UK.
§ These authors contributed equally to this work.
¶ Current address: School of Biosciences, University of Birmingham, Edgbaston,
Birmingham, B15 2TT, UK.
8 Current address: Thermo Fisher Scientific, Stafford House, Boundary Way,
Hemel Hempstead, HP2 7GE, UK.
Received 23rd September 2013,
Accepted 2nd January 2014
DOI: 10.1039/c3mb70423h
www.rsc.org/molecularbiosystems
Molecular
BioSystems
PAPER
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online View Journal | View Issue
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 399
obstetric function which would enable this treatment to be
offered for low grade disease.7
Expression of high risk types of the E6 and E7 viral onco￾proteins is largely responsible for the oncogenic properties of
HPV.8 One of the most intensively studied properties of the E6
protein is its ability to compromise the function of the p53
tumour suppressor protein.9 In association with the cellular E3
ubiquitin ligase E6-associated protein (AP), E6 binds to the p53
protein. E6 mediated activation of E6AP then catalyses the
ubiquitination and subsequent proteasomal degradation of
p53.10 Indeed this strategy of inappropriate activation of the
proteasome is used by many other viruses to subvert the function
of a variety of cellular proteins that would prove detrimental to
viral persistence.1,11,12 This implies that selective inhibition of
proteasomal function could prove to be an effective strategy for
the treatment of HPV infections.
Although FT-IR and Raman spectroscopies are reagentless
and non-invasive tools for global, sensitive and highly reproducible
metabolome analyses with minimal sample preparation, there is
the limitation that specific chemical structures cannot be identi￾fied. UPLC- and GC-MS are powerful and highly sensitive analytical
techniques not only for the quantification of metabolites but also
for the identification of known and unknown compounds in
biological samples. Whilst GC-MS can detect a range of volatile
compounds, non-polar fatty acids, or primary metabolites (when
derivatised), UPLC-MS is more suited to the analysis of secondary
metabolites and lipids when C18 reversed-phase LC is applied.
Since both MS techniques are fully automated they are particularly
amenable to high-throughput metabolomics analysis. Thus, UPLC￾MS and GC-MS have been used widely to investigate metabolic
changes in biological processes, to discover new biomarkers and
drugs, and diagnose diseases.13–21
We have recently reported that the anti-viral drugs indinavir
and lopinavir, which are currently used for the treatment of a
human immunodeficiency virus (HIV) infection, also inhibit the
ability of HPV16 E6 to degrade p53 and selectively kill E6-dependant
cervical carcinoma cells in vitro22 and that the exposure of these
drugs elicits phenotypic changes (i.e., metabolic alteration) of these
carcinoma cells as revealed with FT-IR spectroscopy.23 However,
whilst recent studies show that indinavir is targeted to the nucleus,24
the mode of action of these drugs against HPV is largely unknown.
Therefore, in this study, in order to contribute to an understanding
of the mechanism of these drugs against HPV on human cervical
cell lines, we investigate the level and compositional changes in
intracellular components of control and HPV16 E6 expressing
cervical carcinoma cells upon exposure to a series of physiological
relevant indinavir and lopinavir concentrations and quantify the
intra-cellular concentrations of these drugs.
Materials and methods
Cell line and culture medium
HPV-negative human C33A cervical carcinoma cells (termed
‘‘C33AP’’) were maintained in RPMI 1640 medium (Invitrogen,
Paisley, UK) supplemented with 10% fetal bovine serum (FBS)
from the same batch and 2 mM L-glutamine (Sigma-Aldrich
Company Ltd, Dorset, UK) (complete medium) at 37 1C, 5% CO2.
C33A cells stably transfected with HPV16 E6 (termed ‘‘C33AE6’’)
and pcDNA3.1 were derived and cultured as previously described.25
Protease inhibitor
Indinavir was obtained through the NIH AIDS Research and
Reference Reagent Program (Division of AIDS, NIAID, NIH) as
indinavir sulphate (8145). Lopinavir was provided as a generous
gift from Abbott Laboratories, Park Road, Abbott Park, IL 60064-
6187, USA. Indinavir and lopinavir were dissolved in sterile
distilled water and DMSO (Sigma-Aldrich Company Ltd, UK),
respectively, at working stock concentrations of 20 mM.
Sample preparation
For UPLC-MS metabolite profiling, after incubation, the com￾plete medium was removed and 800 mL of trypsin was added in
order to detach the adherent cells from the flask. Cells were
then incubated for 3 min at 37 1C, 5% CO2. After this period,
cells were resuspended in 10 mL of the complete medium and
were counted. Each 1 106 of C33AP and C33AE6 cells were
then seeded in large scale T75 culture flasks (to provide five
drug conditions and five biological replicates for each drug)
and allowed to adhere and reach 80–90% confluence at 37 1C,
5% CO2. Indinavir (0, 0.05, 0.15, 0.5 and 1.0 mM) and lopinavir
(0, 7.5, 15, 22.5 and 30 mM), or water and DMSO (as respective
controls) were added to the relevant flasks and cells were incubated
for 24 h at 37 1C, 5% CO2; these concentrations were investigated
previously and found to be physiologically relevant.22,23 Three flasks
were employed per drug concentration in order to obtain enough
biomass to be detected by UPLC-MS. After the incubation period,
the growth media were poured off and the cells were washed with
PBS warmed at 37 1C (matching the incubation temperature). 3 mL
of pure methanol (MeOH, 48 1C) was added for quenching
metabolism and cells were scraped from the surface of
the culture flask whilst being kept on ice. After combining
three flasks, cells were freeze-thaw extracted and lyophilised
(i.e., flash frozen in liquid N2 for 1 min and thawed at 4 1C then
vortexed for 30 s 4) and then centrifuged at 3000 g for
10 min. Next, the supernatants (cell extracts) were evaporated at
room temperature and each dried extract was weighed for
normalisation, and subsequently stored at 80 1C prior to
analysis. Due to the large numbers of cultures required for
the experiment, weekly batches were prepared for each cell line
(i.e., week 1 – parent + indinavir; week 2 – parent + lopinavir;
week 3 – E6 + indinavir; week 4 – E6 + lopinavir, each culture
batch included 0 drug dose as a control). Although direct
comparisons of the different cell lines and drug treatments
may be at risk of identifying significant metabolic differences
between the weekly culture batches, each drug treatment and
cell line is still comparable within each culture batch, therefore
allowing the selection of metabolic changes associated with
drug treatment prior to comparing the drug responses between
different cell lines and indinavir and lopinavir challenge. Before
analysis, the samples were reconstituted in 20 mL mg1 cell dry
weight (DW) of 20% aqueous MeOH (HPLC grade). In total four
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
400 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
groups (i.e. 1. parent + indinavir, 2. parent + lopinavir, 3.
E6 + indinavir, 4. E6 + lopinavir) of samples were prepared
and these were analysed as one batch by UPLC-MS using an
Acquity UPLCt (Waters Ltd, Manchester, UK) and SYNAPT
HDMSt (Waters Ltd, UK).
In order to assess instrument performance as detailed in
ref. 26–28 three pooled quality control (QC) samples were prepared:
an indinavir QC to represent equal mix of all indinavir exposed
samples; a lopinavir QC to represent equal mix of all lopinavir
exposed samples; and a mixed QC to represent equal mix of all
the samples.
The initial sample preparation for GC-MS followed the same
methods as applied for UPLC-MS sample generation, with the
addition of derivatisation prior to analysis. However, in order to
reduce the total number of cultures, the study focused upon a
smaller number of levels of anti-viral challenge. 1 106 C33AE6
cells were seeded to each of 90 flasks (to provide three concen￾tration levels for each drug and five biological replicates) and
allowed to adhere and reach 80–90% confluence at 37 1C, 5% CO2.
Treatments included 0, 0.2 and 1 mM of indinavir and 0, 15 and
30 mM of lopinavir, with additional controls of water and DMSO
respectively, the cultures were incubated for 24 h at 37 1C, and
5% CO2. Three flasks were pooled for each drug concentration
in order to obtain enough biomass to be detected by GC-MS.
Harvesting of the cells, metabolic quenching and extraction
of metabolites was performed as described for UPLC-MS.
To each GC-MS extract, 100 mL of internal standard solution
(0.25 mg ml1 succinic-d4 acid, malonic-d2 acid and glycine-d5
in HPLC-grade water) was added and the extract lyophilised at
room temperature by speed vacuum concentration and stored
at 80 1C before analysis.
UPLC-MS and MS/MS methodologies
For metabolite profiling, UPLC was performed on an ACQUITY
UPLCt system (Waters Corporation, Milford, MA, USA). Chromato￾graphic separations were performed on a 1 mm 100 mm, 1.7 mm
ACQUITY HSS T3 C18 column. The column was maintained at
45 1C and samples were eluted with a linear water (0.1% v/v formic
acid)/acetonitrile (0.1% v/v formic acid) gradient over 10 min at a
flow rate of 0.25 ml min1 as follows: 100% A (0 min) to 100% B at
7 min to 100% B at 8 min to 100% A at 8.1 min and held to 10 min.
A 5 mL injection of each sample was used for UPLC-MS(/MS).
A hybrid Quadrupole/Travelling Wave IMS-oa TOF device
SYNAPT HDMS (Waters, Manchester, UK) was operated in positive
ion electrospray (ES +ve) mode. Data were acquired in V mode
with a FWHM of 10 000 with mass accuracy typically within 3 ppm
root mean square (RMS). Data were acquired from 50–1000 Da,
using a source temperature of 120 1C, desolvation temperature of
350 1C and cone voltage of 30 V. Low-energy data were acquired in
function (1) using a collision energy (CE) of 6 eV on the Trap
collision cell and 4 eV on the transfer collision cell. High-energy
data were acquired in function (2) using a ramped CE on the Trap
collision cell of 15–25 eV and a fixed CE on the transfer collision
cell of 25 eV. Sample measurements were performed in triplicate,
to account for any analytical variability. For UPLC-MS, initially
the UPLC-MS profiling data were baseline corrected, aligned,
and exact mass retention time (EMRT) pairs extracted within
Waters MarkerLynxt XS thus producing an X and Y data matrix
of samples aligned for each EMRT. A text file of all the excipient
masses was generated and employed in the cell profiling
MarkerLynxt XS processing method, thus producing an output
containing only EMRT pairs that were associated with endogenous
metabolites. The final output was Pareto scaled29,30 prior to
statistical analysis.
UPLC-MS/MS analysis was performed under the same UPLC
conditions as applied to the UPLC-MS analysis, only linked to a
XEVOt QTOF MS system (Waters, Manchester, UK). Most of the
instrument parameters were made consistent with those applied to
the SYNAPT HDMS with the exception of the desolvation tempera￾ture which was set at 400 1C. The target mass ion for MS/MS was
selected by the quadrupole and was then subjected to CID using a
CE ramp typically in the range 15–35 eV. Sample measurements
were compared to those of pure analytical standards.
UPLC-MS and MS/MS metabolite identification
The process of marker identification was made up of four stages.
 The first stage employed the exact mass and accurate
isotope ratio data for each EMRT pair to calculate proposed
elemental composition. This was automatically calculated using
the Elemental Composition Calculator within MarkerLynx XS.
The mass accuracy (in ppm or mDa) and isotope ratio accuracy
(i-FIT value) of each proposed formula were provided allowing
the quality of the elemental composition results to be assessed.
 In the second stage, each EMRT pair and its proposed
elemental compositions were searched against predefined online
databases (Chemspider.com) using either the exact mass or the
elemental composition as the search criterion. The results from
the database search were compared with the proposed elemental
compositions and where a match was found (elemental com￾position and database result) the structure of the proposed
metabolite was obtained from the database.
 The third stage of marker identification was to use Mass￾Fragment software (Waters Ltd, UK) to assign the high energy
MSE fragments according to the structure of the proposed
metabolite. The method of marker identification is a simple,
logical process which employs all of the UPLC-MS data qualities
to build confidence in assignments and therefore reduces the
incidence of false positives.
 For unambiguous confirmation a fourth step is required,
where comparative UPLC-MS/MS analysis was performed on both
the sample EMRT pairs and an authenticated reference standard
for the predicted metabolite structures. Matching of both RT and
MS/MS spectra between the reference standards and sample
EMRT pairs is required for unambiguous confirmation.31
 In addition, automated workflows (PutMetID) have
been employed for the rapid and high-throughput annotation,
and putative metabolite identification of UPLC-MS-based meta￾bolomic data sets as described in ref. 32 and 33.
The UPLC-MS metabolite identification routine and reporting
adhered to standards set out by the Metabolomics Standards
Initiative.31 An example of this metabolite annotation approach
is shown in ESI,† Fig. S1.
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 401
Targeted quantification of lopinavir and indinavir by UPLC-MS
For drug quantification, UPLC-MS analysis for the quantifica￾tion of lopinavir within the samples was performed on a Waters
ACQUITY UPLC system, whereas for indinavir were performed
upon a Thermo Accela UPLC system, both UPLC systems were
operated using the same column and under the same condi￾tions as applied for metabolite profiling, but we were only
interested in the peak areas for the two drugs. A number of
serially diluted standards for each of the two drug compounds
were analysed six times each. The sample extracts initially
profiled by UPLC-MS were diluted three fold prior to analysis
on the Q-TOF micro, and 24 fold prior to analysis on the
LTQ-Orbitrap XL, thus preventing saturation of the MS detectors.
The indinavir and lopinavir extracted peak areas, for both drug
standards and samples, were transformed to their natural log
values, and calibration curves were built for the two drug
standards. Quantification of the indinavir and lopinavir mass
peaks within the cell line samples were predicted against
the calibration curves constructed using the serially diluted
drug standards. All stages of the quantification process were
performed within Matlab R2009.
GC-MS methodology
For GC-MS metabolite profiling, samples were derivatised and
analysed by GC-TOF/MS as described in ref. 34 following the
optimised method for analysis of yeast cells. All experiments
were run on a GC-TOF/MS instrument (Agilent 6890N gas
chromatograph and LECO Pegasus III TOF mass spectrometer)
using the manufacturer’s software (ChromaTOF version 2.12).
A DB-50 GC column (Supelco, Gillingham, UK; 30 m 0.25 mm
0.25 mm film thickness) was used. In the ChromaTOF software,
the S/N threshold was set to 10, baseline offset to 1.0, data
points for averaging to 3, and peak width to 3. Data were
deconvolved within ChromaTOF and the metabolites identified
by RI (10) and MS (80% forward and reverse match) library
matching with high confidence against an in-house library,32
the NIST 2002 library and the Golm Metabolome Database
(GMD).35,36 The GC-MS metabolite identification routine and
reporting adhered to standards set out by the Metabolomics
Standards Initiative.31 The deconvoluted profiles for each
sample were next aligned thus forming an X and Y matrix for
statistical analyses.
Data analysis
Data analysis followed the Metabolomics Standards Initiative
guidelines.37 The resultant data matrix of EMRTs was sub￾mitted to multivariate statistical analysis in Matlab version 7
(The Mathworks, Inc., Matick, MA, USA). Principal components
analysis (PCA) was performed on the pre-processed UPLC-MS
data set. Briefly, PCA is one of the oldest and most widely used
multivariate techniques, it is employed to reduce the dimension￾ality of metabolomics data whilst maintaining the majority of
its variance and is used to visualise general trends and outliers
among the observations. Therefore, it is often utilised as
an initial step prior to cluster or discriminant analysis.38–40
PCA was performed according to the NIPALS (nonlinear iterative
partial least squares) algorithm.41 In addition to the inspection
of loadings matrices from PCA for the discovery of which
metabolites were more discriminatory, as a final data mining
step univariate data analyses were applied. Here, N-way ANOVA
was performed within Matlab 7 software (The Mathworks, Inc.,
Matick, MA, USA) and was used to assess levels of significant
difference for each EMRT pair between control and drug doses
within each cell line individually.42
Finally, Matlab was used to perform multi-block consensus
PCA (cPCA) on the GC-MS data sets.42 PCA is typically applied to
all of the metabolite variables and reduces them to a small
number of new variables (PCs) which explain the greatest
sources of variance, however conventional PCA does not always
detect common trends between different sample classes. Recently,
cPCA has been introduced in which each sample class or anti-viral
drug that the cells have been challenged with can be divided into
several blocks, the cPCA then looks to fit the data within the
different blocks to discover common trends between them which
may aid greatly in the production of more interpretable models.
This multi-block method has been used in cases where the
number of variables is large and additional information is
available for blocking the variables into conceptually meaningful
blocks.43,44 Since multi-block PCA can offer the potential to
extend the scope of conventional PCA and to identify common
trends between different blocks (e.g., E6 + indinavir vs. E6 +
lopinavir), cPCA was applied to the GC-MS data sets, after
samples were rearranged into two separated blocks, each one
for each type of drug and their treated concentrations.
Results and discussion
UPLC-MS data quality
A total of 332 injections were performed totalling 59 h of analysis.
Amongst these were 5 biological replicates for each cell line/drug
treatment permutation, 3 technical replicates of each concentration
level of drug treatment in each cell line, 6 replicate injections of
each drug QC, 8 replicate injections of the mixed QC and drug
standards. All samples were analysed in a randomised manner.
Throughout the course of the analysis retention time drift was
minimal and 3 ppm RMS mass accuracy was maintained.
Multivariate analysis of UPLC-MS profiling data from C33A
parent and E6-transfected cells
Extracts from the C33AP and C33AE6 cells grown in both
the absence and presence of indinavir and lopinavir, at physio￾logically relevant levels, were analysed by UPLC-MS. The 3D
UPLC-MS data obtained were converted into a 2D matrix by
MarkerLynx XS. Each data point in the 2D matrix represents an
exact mass retention time (EMRT) pair (i.e., a potential marker
along with its intensity), and the matrix was converted to a tab
delimited text file suitable for import into Matlab software. PCA
was then carried out initially to determine basic biological
differences between the two different cell types and different
drug treatments, and to investigate trends determined by the
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
402 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
drug dosing. The resultant PCA scores plots are shown in Fig. 1.
As can be seen in Fig. 1A, trends between the cells treated with
two different drugs were clearly observed, indicating a strong
feasibility to investigate relationships between drug concentra￾tions and mass spectra in terms of anti-viral effects on their
biochemistry. In addition, the scores of the cells exposed to low
concentrations of both drugs are observed in the bottom left of
the PCA space (PC1 vs. PC2), as the concentration of the anti￾viral drugs increases, the clusters spread from the bottom to
the top left (indinavir exposed cells) or top right (lopinavir
exposed cells) of the scores plot (Fig. 1A).
Next, separate PCAs of indinavir and lopinavir challenged
samples were performed to investigate how differently each drug
affects intracellular metabolites on C33AP and E6 cells (Fig. 1B and
C). As can be seen in Fig. 1B and C, no clear separation between
parent and E6 cells exposed to indinavir is observed, whilst scores
of C33AP cells treated with lopinavir is separated clearly from
those of C33AE6 cells treated with lopinavir. However, better
discrimination between indinavir challenged C33AP and E6 cells
in the different PCA space (PC1 vs. PC3, data not shown) is
observed, indicating inherent differences between the parent
and E6 cells in terms of their biochemistry due to the presence
of the E6 oncogene in C33AE6 cells. We have recently confirmed
the phenotypic differences between host C33A cells and those
expressing E6 using the vibration technique of FT-IR spectro￾scopy.23 Interestingly on closer inspection of each PCA the scores
from C33AP and E6 cells exposed to indinavir show similar trends
whereas those from C33AP and E6 cells exposed to lopinavir reveal
markedly different trends, suggesting that the cells are interacting
with lopinavir in a different way when compared to indinavir.
Therefore, these results from PCA clearly reveal that metabolite
profiling using UPLC-MS is sensitive enough to detect the meta￾bolic changes elicited by each anti-viral drug.
UPLC-MS feature selection
To investigate which mass ions contributed to the separations and
trends from each cell line and drug treatment 2-way orthogonal
comparisons were made between 0 dose and mid-dose,
Fig. 1 PCA scores plots from C33A parent and E6-transfected cells exposed to indinavir and lopinavir. (A) Mixed, indinavir and lopinavir quality controls
(QC) are included as an indication of data quality, (B) indinavir exposed parent (PI) and E6 cells (EI) with indinavir QC, and (C) and lopinavir exposed parent
(PL) and E6 cells (EL) with lopinavir QC showing the anti-viral drug effect. (A) PC1 (54.08%) versus PC2 (21.28%); blue: indinavir treated parent and E6 cells;
green: lopinavir treated parent and E6 cells; (B) PC1 (51.37%) versus PC2 (17.88%); blue: indinavir challenged E6 cells; red: indinavir challenged parents
cells; (C) PC1 (64.28%) versus PC2 (14.61%); blue: lopinavir challenged E6 cells; red: lopinavir challenged parents cells. In all of these plots the first two
letters indicate the type of cells and the name of the drug, the third number indicates the number of biological replicates, and the last number represents
the drug concentrations (mM for indinavir, mM for lopinavir). Mix QC is an equal mix of all the samples; indinavir QC comprises an equal mix of all indinavir
challenged samples; lopinavir QC is an equal mix of all lopinavir challenged samples. Arrows are drawn as a visual guide indicating the relationship
between the drug concentrations and the mass spectra.
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 403
and 0 dose and high-dose samples for each drug treatment
using PCA. To clarify the drug effect between drug concentra￾tions, mid- (0.15 mM for indinavir and 15 mM for lopinavir)
and high-dose (1.0 mM for indinavir and 30 mM for lopinavir)
were chosen (ESI,† Fig. S2). Furthermore, to determine the
chemical differences between each two sample groups, a scatter
plot (S-plot) based upon PC loadings from the 2-way PCA was
performed. Examples of each 2-way PCA ordination and S-plot
are provided in Fig. S2 (ESI†).
PCA ordination scores plots from all 2-way comparisons
showed clearly separated clusters (data not shown) between
non-drug exposed control cells, and mid- and high-dose
exposed cells which indicate metabolic differences in terms
of the level and compositional changes of intracellular meta￾bolites caused by the anti-viral drugs. S-plots were constructed
to determine significant variables contributing to the class
separation. EMRT pairs making significant positive or negative
contributions to the PC1 axis, which separates the classes, were
selected. As a result, 225 key mass ions out of 1860 EMRT pairs
were selected and cross checked with significant variables
selected by univariate N-way ANOVA (with the false discovery
rate Q r 0.05).45 p-Values and putative identification based on
1860 EMRT pairs are shown in Table S1 (ESI†). This resulted in
71 of the EMRT pairs being selected by both multivariate and
univariate analyses. Possible adducts species were then removed,
and finally 34 of the common key mass ions were selected for
unambiguous metabolite identification (Table 1). For identifi￾cation of key metabolites, the proposed elemental composition,
C5H9NO2 of m/z 116.0714 [M + H]+ was calculated automatically
and was searched against online databases. Following this,
MassFragment software was used to assign the high energy
MSE fragments according to the proposed metabolite, proline
(ESI,† Fig. S1A and B) and then UPLC-MS/MS analysis was
performed on both m/z 116.0714 (RT 0.3856) and the standard
compound of proline (Fig. S1C and D, ESI†). As can be seen in
Fig. S1B (ESI†), m/z 116.0714 was unambiguously confirmed as
proline by comparing UPLC-MS/MS fragmentation of sample
EMRT with that of the authenticated proline standard. Using the
same procedure of UPLC-MS/MS analysis as above, m/z 120.0806
(RT 2.1775) was determined as 2,3-dihydro-1H-indole (indoline).
Further putative metabolite identifications of significant variables
were provided by applying the PutMetID workflow.39,40
Drug quantification in C33A parent and E6-transfected cells
exposed to indinavir and lopinavir using UPLC-MS
To determine the intra-cellular concentration of anti-viral drugs,
indinavir and lopinavir in drug challenged C33AP and E6 cells,
sets of serially diluted drug standards of known concentration
(indinavir, 0.5, 0.75, 1, 2.5, 5, 7.5, 10, 25, 50, 75, 100, 250, 500 mM;
lopinavir, 5, 10, 15, 20, 25 and 30 mM) and extracts from C33AP and
E6 cells exposed to indinavir and lopinavir (profiling samples) were
initially analysed by UPLC-MS as described above. Calibration
curves containing a minimum of six data points were then created
using data sets from the standards of known concentration
(analysed six times and averaged for each concentration). The
linearity of the calibration curves of indinavir and lopinavir (R2
)
by UPLC-MS were 0.9979 and 0.9301, respectively, and so they
proved to be suitable for quantification. Following this, the
extract concentrations of the two drugs for each cell line were
determined against the calibration curves by comparison of the
mass spectral drug intensity (extracted peak area for lopinavir
or indinavir parent mass ion converted to the natural log value)
recorded in the cellular extracts. The drug levels within the
cellular extracts are reported in Table 2. The quantification can
only be viewed as relative and the reported values as arbitrary.
This relates back to issues in relating the predicted extract
molar concentration back to the cell weight in each extract. Due
to limited volumes of material and the requirement to perform
sample processing and extraction rapidly, neither determina￾tion of sample FW or a cell count were performed, instead
sample normalisation was performed according to the DW of
the extract. However, as it is known that treatment with the
anti-viral compounds does not reduce cell growth;23,24 in total
three flasks were pooled (each containing B5 106 cells) to
provide enough material (B15 106 cells) for metabolite
extraction. Thus the values presented in Table 2 can only be
related back to the estimated cell number of B15 106 cells
and not to a true cell weight, as one would desire, and thus the
values can only be considered as arbitrary but are still suitable
for the relative comparison of anti-viral levels between the
different C33A cultures.
As can be seen in Table 2, intra-cellular levels of lopinavir
were approximately two-fold lower in the C33AE6 than C33AP
cells, suggesting that C33AE6 is actively excreting the lopinavir
anti-viral drug from the cell. It is known that multidrug
resistance is a major function of cancer cells, which develop
resistance to toxic or chemotherapy drugs.46,47 This multidrug
resistance has been highly correlated to the function of mole￾cular ‘efflux pumps’, which actively excrete chemotherapy drugs
from the cell.46 Interestingly, it has been reported that p53
mutations and/or a loss of p53 function strengthens multidrug
resistance in neuroblastoma cell lines.48 Thus, it could be
hypothesised that since expression of viral E6 proteins in
C33AE6 cells inappropriately activates the 26S proteasome to
degrade p53,22 very low levels of p53 proteins in the cells could
cause high-level multidrug resistance, and thus lopinavir could
be excreted from the cells via membrane efflux pumps. On the
contrary, intra-cellular levels of indinavir were greater in the
C33AE6 than C33AP cells. Perhaps one would expect that
indinavir and lopinavir would act in the same way upon the
C33AE6 and C33AP cells, however the concentrations of the
two anti-viral compounds used for treatment differ massively
(7.5–30 mM lopinavir, 0.05–1 mM indinavir), even the structures
of the two anti-viral compounds differ significantly, and it is
possible that a culmination of structural and dosing differences
between the two anti-viral compounds may explain these effects
within challenged C33AP and E6 cultures. Previous work
revealed that the concentration of indinavir was approximately
eight-fold greater in the nucleus than in the cytoplasm of C33A
E6 cells, which demonstrated that indinavir undergoes enhanced
nuclear translocation in E6 expressing cells only and this
suggests that the nucleus is the most likely site of action for
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
404 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
Table 1 The significant metabolite features selected by PCA and N-way ANOVAa
RT m/z
True
mass
Elemental
composition Adduct Putative IDs Confidence
0.5956 80.0505 79.0432 C5H5N Pyridine 3
0.6129 86.992 85.9847 Unidentified 4
0.5292 87.0266 86.0193 C4H6S 2,3-Dihydrothiophene; 2,5-dihydrothiophene;
but-3-yne-1-thiol
0.3856 116.0714 115.0641 C5H9NO2 Proline 1
2.3639 118.0662 117.0589 Unidentified 4
0.7755 119.0498 118.0425 C9H8O3 HCOOH Coumarate; hydroxy-nonene-diynoic acid; benzoyl acetate;
caffeic aldehyde; hydroxycinnamate; coumarinate;
phenylpyruvate; cyclopenta[b]pyran
2.1775 120.0806 119.0733 C8H9N Indoline 1
0.6267 123.0537 122.0464 Unidentified 4
0.5355 150.0596 149.0523 C5H11NO2S Methionine 2
0.7716 165.0551 164.0478 C9H8O3 Coumarate; hydroxy-nonene-diynoic acid; benzoyl acetate;
caffeic aldehyde; hydroxycinnamate; phenylpyruvate;
trans-hydroxycinnamate
2.2151 166.0870 165.0797 C9H11NO2; C9H8O H; NH3 Phenylalanine; pyridyl-butanoate; aminohydrocinnamic
acid; tyrosinal; hydroxy-indanone; isochromanone;
hydroxycinnamyl aldehyde; coumaraldehyde;
dihydrocoumarin; trans-cinnamate
2.3923 173.0826 172.0753 C8H12O4 2-Octenedioic acid 3
2.3676 179.0499 178.0426 Unmatched adduct 4
2.4619 187.0987 186.0914 C8H12O4 2-Octenedioic acid adduct 2
2.4501 227.0861 226.0788 C5H12N6O3 Na Dimethylenetriurea 3
2.4034 232.1559 231.1486 C11H21NO4 (iso)Butyrylcarnitine 2
2.4998 245.0991 244.0918 C15H13FO2;
C10H22O2S;
C7H14N4
H; K; Na_HCOONa Flurbiprofen; MPC-7869; octyl-hydroxyethyl sulfoxide;
guancydine
2.4427 246.1701 245.1628 C12H23NO4;
C13H19N5; C12H20O4
H; NH3 Methylbutyroylcarnitine; isovalerylcarnitine; pinacidil;
dodecenedioic acid; dioxo-dodecanoic acid; traumatic
acid
2.458 267.0804 266.0731 C10H14N4O2;
C10H14O4; C8H16O4
Na_Na; HCOONa;
Na_HCOONa
IBMX; morinamide; (dimethoxylphenyl)ethane-diol;
dihydroxymint lactone; dihydroxy-dihydro-cumate;
guaifenesin; dihydroxy-octanoic acid; cladinose
2.4194 281.1052 280.0979 C15H18N2O;
C14H14ClNO2;
C9H11F2N3O4;
C9H16N4S
K; NH3; NH3;
HCOONa
Octahydroindolo[2,3-a]quinolizin-(6h)-one; huperzine A;
selagine; clopirac; gemcitabine; burimamide
2.3863 305.0824 304.0751 C13H18C12N2O2;
C15H14N4S;
C17H14N2;
C14H18N2O
H; Na; NaCl; KCl Melphalan; methylphenylsulfanyl-quinazolinediamine;
ellipticine; olivacine; camoensine; ibudilast;
isopropylantipyrine
0.5752 308.0923 307.085 C10H17N3O6S;
C18H13NO4;
C13H19NO4S;
C16H15N3
H; H; Na; NaCl Glutathione; stealthin C; letrozole; probenecid; epinastine 3
2.3854 328.1427 327.1354 C18H23NO;
C11H23N7
NaCl; KCl Bifemelane; orphenadrine; meladrazine 3
0.5226 348.0719 347.0646 C10H14N5O7P Deoxyguanosine-monophosphate; AMP; azido￾deoxythymidine-monophosphate; ribosyladenine￾phosphate; adenosine-phosphate; dGMP;
formycin-monophosphate; vidarabine phosphate
2.3935 367.1521 366.1448 C17H20CIN3O3;
C19H22N4O;
C19H22O3; C16H26O4
NH3; Na_Na;
HCOONa; HCOOK
Azasetron; aminacrine; doisynoestrol; glepidotin C;
ostruthin; hydroxy juvenile hormone III; oxo-hydroxy￾hexadecadienoate
3.073 374.2459 373.2386 C26H31NO 131-I-TM-601; androsta-dienoquinolinol 3
2.4834 392.1374 391.1301 C18H28NO4P;
C22H23NO2;
C19H27NO3;
C19H21N3O;
C16H24FNO
K; NaCl; KCl;
HCOOK;
NaCl_HCOONa
Dimemorfan phosphate; (alphaS, betaS)-alpha-ethyl￾alpha-(4-methoxyphenyl)-beta-phenyl-2-pyridineethanol;
enpromate; glycopyrrolate; nateglinide; protoemetine;
tetrabenazine; alcaftadine; talastine; zolpidem; UH-301
5.4651 426.3594 425.3521 Unidentified 4
2.3931 434.1229 433.1151 C22H21NO7;
C18H20FN5O4;
C18H23NO6;
C18H24ClN3O2;
C17H18FN5O2;
C19H27NO3;
C19H21N3O
Na; Na_Na;
HCOOK; HCOOK;
Na_HCOONa;
NaClx2;
NaCl_HCOONa
Cetocycline; gemifloxacin; riddelline; pancopride;
8-benzo[1,3]dioxol-,5-ylmethyl-9-butyl-2-fluoro-9h-purin-6-
ylamine; glycopyrrolate; nateglinide; protoemetine;
tetrabenazine; alcaftadine; talastine; zolpidem
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 405
this compound against HPV.24 The current data perhaps reflects
that due to the nuclear localisation of indinavir within C33AE6
cells, the final concentration of the drug is maintained at a
greater level than in the parent cells where no such nuclear
localisation has been observed in previous studies. It must also
be taken into account that the reported concentrations of
indinavir within the various cell lines can be quite variable; this
variability was even observed within the same cell line at the
same dosing concentration. It is likely that this is an indication
of the problems caused due to indinavir crystallising within the
challenged cell cultures, which relates to the high concen￾tration of indinavir that the cells were subjected to. Indinavir
crystallisation has even been observed within HIV patients
being treated with high concentrations of indinavir.
Trend plots were employed to determine which mass ions
are altered significantly by drug concentrations and these are
shown in Fig. 2. In these plots the peak intensities for the
selected analytes from univariate and multivariate analyses are
compared across the entire concentration series, visualising
those metabolites which are reduced or elevated. As can be seen in
Fig. 2, there are several patterns explaining the drug associated level
changes of metabolites. In indinavir challenged C33AP and E6 cells,
the intensities of m/z 80.0505 at RT 0.5956 (pyridine; putative
assignment), m/z 116.0714 at RT 0.3856 (proline; confirmed
assignment), m/z 173.0826 at RT 2.3923 (2-octenedioic acid;
putative assignment), m/z 232.1559 at RT 2.4034 ((iso)butyryl￾carnitine; putative assignment) and m/z 348.0719 at RT 0.5226
(deoxyguanosine-monophosphate; putative assignment) decrease
as the drug concentration increases.
The intensities of m/z 87.0266 at RT 0.5292 (2,3-dihydro￾thiophene, 2,5-dihydrothiophene or but-3-yne-1-thiol; putative
assignment), m/z 120.0806 at RT 2.1775 (indoline; confirmed assign￾ment), m/z 150.0596 at RT 0.5355 (methionine; putative assignment)
and m/z 165.0551 at RT 0.7716 (coumarate; putative assign￾ment) and m/z 166.0870 at RT 2.2151 (phenylalanine; putative
assignment) increase as the drug concentration increases.
In lopinavir challenged C33AP and E6 cells, like indinavir
challenged cells, the intensities of m/z 80.0505, m/z 173.0826,
m/z 232.1559 and m/z 348.0719 decrease as the drug dose
increases whilst the intensity of m/z 166.0870 increases as the
drug dose increases. However, the interesting features are
observed in C33AP and E6 cells exposed to lopinavir in terms
of the drug response against the two different cell lines. With
respect to m/z 87.0266, m/z 116.0714, m/z 120.0806, m/z 150.0596
and m/z 165.0551, their intensities decrease in C33AE6 cells but
increase in C33AP cells as the drug concentration increases. This
trend analysis also confirms that indinavir and lopinavir interact
differently with cells.
Overall, the concentration of phenylalanine increases in both
C33AP and E6 cells as the doses of indinavir and lopinavir increase
and the concentrations of 2-octenedioic acid, (iso)butyrylcarnitine
and deoxyguanosine-monophosphate decrease in both cells as
the doses of both drugs increase. The fact that the levels of
Table 1 (continued)
RT m/z
True
mass
Elemental
composition Adduct Putative IDs Confidence
1.6819 466.1115 465.1042 C15H21N7O7S;
C19H25NO10;
C20H21N5O6;
C20H25NO2S2
Na; K; K;
Na_HCOONa
-O-(N-(L-Prolyl)-sulfamoyl)adenosine; vicianin;
lucumin; pemetrexed; tiagabine
0.3735 635.1423 634.135 C20H32N6O12S2 Na Glutathione disulfide; oxidized glutathione; oxiglutatione 3
2.3512 685.1863 684.179 C34H32N4O9;
C31H36O12
Na_Na; HCOOK 2-Octenedioic acid/(iso)butyrylcarnitine adduct; nicomol 3
2.3686 768.1238 767.1165 C21H36N7O16P3S 2-Octenedioic acid/(iso)butyrylcarnitine adduct; coenzyme
5.3537 949.6295 948.6222 C59H90O7; C53H90O6 K; NaCl_HCOONa Thermocryptoxanthin-13; TG(50 : 6) 3
a MSI metabolite identification confidence levels,31 level 1 – RT and MS/MS match to reference standard; level 2 – MS/MS match but no standard;
level 3 – MS match to MMDB, GMD, or NIST 05 library; level 4 – unidentified.
Table 2 Quantification of indinavir and lopinavir in C33A samplesa
Sample type
Lopinavir/indinavir
(arbitrary)
Standard
error
C33AE6 0 mM lopinavir challenged NDb ND
C33AE6 7.5 mM lopinavir challenged 8.8 1.182793
C33AE6 15 mM lopinavir challenged 11.54 1.282809
C33AE6 22.5 mM lopinavir challenged 15.32 1.119553
C33AE6 30 mM lopinavir challenged 20.54 1.216799
C33AP 0 mM lopinavir challenged ND ND
C33AP 7.5 mM lopinavir challenged 15.36 1.087474
C33AP 15 mM lopinavir challenged 26.92 3.163131
C33AP 22.5 mM lopinavir challenged 33.18 2.55918
C33AP 30 mM lopinavir challenged 37.52 2.847174
C33AE6 0 mM indinavir challenged ND ND
C33AE6 0.05 mM indinavir challenged 52.96 13.91149
C33AE6 0.15 mM indinavir challenged 277.96 19.0113
C33AE6 0.5 mM indinavir challenged 343.44 25.98102
C33AE6 1 mM indinavir challenged 583.24 67.82022
C33AP 0 mM indinavir challenged ND ND
C33AP 0.05 mM indinavir challenged 121.88 29.02114
C33AP 0.15 mM indinavir challenged 162.08 30.96893
C33AP 0.5 mM indinavir challenged 102.44 19.96671
C33AP 1 mM indinavir challenged 195.28 61.89755
a Indinavir and lopinavir were quantified in C33AP and C33AE6 cells
that had been previously challenged with various concentrations of the
two anti-viral compounds. The profiling extracts were diluted so as to
not saturate the mass spectrometers detector. Analytical standards for
the two anti-viral compounds were serially diluted, analysed six times
and the data averaged. The extracted peak areas for the anti-viral
compounds parent masses within both the analytical standards and
the sample extracts were transformed to their natural log values, calibra￾tion curves were constructed for the two anti-viral compounds, the
concentration of the anti-viral compounds within the cellular extracts
were predicted against the calibration curves. b ND: Not detected.
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
406 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
Fig. 2 Trend plots of the 10 most significant metabolites. Blue line: C33A E6-transfected cell, green line: C33A parent cell, (A, K) 2-methylenebut-3-
enenitrile (tentative assignment); (B, L) 2,3-dihydrothiophene, 2,5-dihydrothiophene or but-3-yne-1-thiol (tentative assignment); (C, M) proline
(unambiguously confirmed assignment); (D, N) indoline (unambiguously confirmed assignment); (E, O) methionine (tentative assignment); (F, P) 6-
acetyl-2-hydroxy-cyclohepta-2,4,6-trien-1-one or 9-hydroxynon-7-en-3,5-diynoic acid (tentative assignment); (G, Q) phenylalanine (tentative assign￾ment); (H, R) 2-octenedioic acid (tentative assignment); (I, S) (iso)butyrylcarnitine (tentative assignment); (J, T) 20
-deoxy-guanosine 50
-phosphate
(tentative assignment).
Fig. 3 The plots of the super scores (A) and each block score (B, indinavir; C, lopinavir) from cPCA. Arrows are drawn as a visual guide indicating the
relationship between the spectra and the concentrations of indinavir and lopinavir.
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 407
Fig. 4 Box plots of identified metabolites (A, glutathione; B, aspartic acid; C, cysteine; D, sugar phosphate; E, malic acid) deemed as significant common
trends by cPCA resulting from exposure of C33A E6 cell cultures to indinavir (left) and lopinavir (right).
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
408 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
these four compounds were changed in C33AP cells (which do
not produce E6 oncoproteins) as well as C33AE6 could assume
that these cellular components represent general drug effects
on the two different cells. By contrast, only differences between
C33AP and E6 cells exposed to lopinavir in terms of the level
changes of metabolites is that concentrations of proline, indo￾line and methionine decrease in C33AE6 but increase in C33AP
cells as lopinavir concentration increases. Generally, since
methionine is known as the major methyl group donor of
various intermediates in vivo such as methyl groups of DNA and
RNA intermediates, the decrease in the level of this compound in
lopinavir challenged C33AE6 cells could be connected to the drop
in total cellular nucleotides and carbohydrates, as observed in our
previous work.23 A decrease in carbohydrates for instance could
also be potentially significant, a previous investigation reported
that insulin resistance in vitro can be induced by lopinavir,
which also inhibits glucose and 2-deoxyglucose uptake into
primary rat adipocytes in vitro.49,50 Thus a decrease in the
levels of carbohydrates and carbohydrate-based nucleotides
could be related to the documented effect of lopinavir on
carbohydrate metabolism.
Multivariate analysis of GC-MS profiling data from C33A parent
and E6-transfected cells
The C33A E6-transfected cells were grown in the presence of
indinavir at concentrations of 0, 0.2 and 1 mM, and lopinavir at
concentrations of 0, 15, and 30 mM for 24 h at 37 1C, 5% CO2,
giving a total of each three conditions and five biological
replicates including the controls. The cells were quenched
and extracted using 100% MeOH (48 1C) and then measured
using GC-MS. All the peak areas were normalised to that of
the succinic-d4 acid internal standard effectively producing a
semi-quantitative output. The deconvoluted and library
matched GC-MS profiling data set from each condition was
exported as a Microsoft excel sheet and analysed using Matlab.
Multi-block cPCA was then carried out to investigate the common
effect of the two anti-viral drugs on E6 cells.
The plots of super scores and each block scores from cPCA
are shown in Fig. 3. The common trend of the two blocks is
represented by the super scores space, along with block weights
which represent the contribution of each block to the super
scores. As can be seen in Fig. 3A, the super scores of C33AE6
cells exposed to low concentrations of indinavir and lopinavir
are observed in the bottom right hand corner, as the concen￾tration of the anti-viral drug increases, the cluster spreads from
right to left. The two drugs could not be detected by GC-MS due
to their large molecular weights, thus the clustering of sample
groups is dependent on cellular metabolism and not drug￾associated peaks. In successive analyses, each block scores were
plotted with the variables within the corresponding block.
Again, each block scores from indinavir or lopinavir treated
C33AE6 cells show similar patterns to the super score plot,
indicating that these two anti-viral drugs have a similar effect
on the cells. It is therefore valuable in the sense that these trends
are directly associated with the concentrations of indinavir and
lopinavir. This result clearly reveals that the metabolic profiling
data from C33A HPV16 E6 expressing cervical carcinoma cells
using GC-MS contain valuable information for studying the
phenotypic effect of the anti-viral drug in terms of level changes
of intracellular metabolites.
To investigate which specific metabolites are associated with
the anti-viral drug effects, the loadings vectors were calculated
for each individual block and plotted (ESI,† Fig. S3). As can be
seen in this figure, several significant variables marked by red
were identified and confirmed as significant by the Friedman
test (non-parametric 2-way ANOVA). From these loadings plots,
24 significant metabolite variables were determined and only
five metabolites were unambiguously identified through
in-house GC-MS libraries as reduced glutathione, aspartic acid,
malic acid, cysteine and sugar phosphate, respectively, they
show significant quantitative differences that correlate to the
different drug doses. As can be seen in Fig. 4, the levels of these
metabolites are reduced as the concentration of the anti-viral
drugs increases. In addition, other than significant variables
Fig. 5 Box plots of metabolites (A, octadecenoic acid; B, lactose) against the concentrations of indinavir (left hand side) and lopinavir (right hand side).
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 409
from the loadings plots, which show the common effect between
two anti-viral drugs, some of the metabolites such as octadecenoic
acid and lactose reveal the significant concentration change
associated with increasing indinavir doses whereas no clear
level changes of these metabolites are shown against increasing
lopinavir concentrations (Fig. 5).
Conclusion
Intracellular metabolite changes caused by indinavir and lopinavir
in the presence and absence of HPV16 E6 against an isogenic
background (C33A cervical cancer cells) were investigated using
MS-based metabolomics. We believe that this could help
towards providing a means of defining the mode of action of
these compounds against HPV, which is not the designed target
for these HIV protease inhibitors. Thus the objective of the
study was to identify key metabolites involved in the anti-viral
response and to provide information related to the pathway
relationships between these components.
UPLC-MS-based metabolic profiling in combination with a
variety of univariate and multivariate analyses such as N-way
ANOVA and PCA is a very useful method for the determination
of specific and common metabolic effects of indinavir and
lopinavir on these cells. Along with identifying all peaks that
are seen in UPLC-MS we also selected significantly altered
metabolites from univariate and multivariate statistical methods
to be subsequently identified employing UPLC-MS/MS. This
process is valuable since unequivocal identification of all low
molecular weight metabolites is considered to be a challenging
step in the application of metabolomics. In addition, we
also confirm that the change in the levels of phenylalanine,
2-octenedioic acid, (iso)butyrylcarnitine and deoxyguanosine￾monophosphate in drug challenged C33AP and E6 cells could
be thought of as general drug effects on both cell lines whereas
a reduction of the levels of proline, indoline and methionine in
only lopinavir exposed C33AE6 cells represents E6 oncogene
specified effects of the drug. These compounds can potentially
be used for the specific biomarkers in order to understand the
mechanism of the anti-viral drug effect against HPV, and their
role within the mode of action of these protease inhibitors will
be an area of future work. Furthermore, we also report that the
level of the lopinavir anti-viral drug detected in C33AE6 cells is
significantly lower than in C33AP cells treated with the same
dosing concentration. By contrast, the level of the indinavir
anti-viral detected in C33AE6 cells is significantly higher than
in C33AP cells treated with the same concentration. Although
this is currently difficult to explain fully, several hypotheses
have been generated for future research based upon the leads
presented by this metabolomics investigation and our previous
observation that indinavir is found at concentrations eight-fold
higher in the nucleus compared to the cytoplasm.24 The multi￾functional effects of HIV protease inhibitors are well known51
and the nuclear translocation of indinavir is potentially inter￾esting. High risk HPV16 E6 also localizes predominantly in the
nucleus52 and it is known that it participates not only in the
proteosomal destruction of TP53 but also acts to prevent the
binding of the TP53/p300 transcriptional complex to its nuclear
transactivation target genes.29 Thus it is possible that nuclear
indinavir may participate in suppressing the ability of E6 to
block the transactivation function of TP53. However, as we have
stated, the off target activities of HIV PIs are very diverse and
will clearly form the basis of future studies.
To supplement this UPLC-MS analysis GC-MS was used on a
subset of the samples. Unfortunately, as with the UPLC-MS
metabolic profiling, the number of key statistically significant
metabolites that can be identified is limited, indicating that
current GC-MS libraries also need to be improved substantially
for this cervical cell culture-based target matrix, however as
with UPLC-MS this is greatly limited by the availability of high
purity reference standards allowing unambiguous identification.31
Despite of this limitation, GC-MS analyses have resulted in the
detection of several statistically significant and potentially
clinically interesting metabolites such as the detoxification related
compound glutathione (m/z 308.0923 at RT 0.5752) in a reduced
form. A reduction in levels of the sugar lactose and also a
concurrent reduction in sugar-phosphate and the unknown sugars
were seen as anti-viral dosing levels increased. This may be related
to a reduction in energy metabolism or the arrest more specifically
of glycolysis and potentially enhanced mitochondrial energy meta￾bolism. Malic acid was also lowered as anti-viral dosing increased
potentially indicating that the anti-virals were also having a direct
influence upon the TCA cycle. Reductions of the amino acid,
aspartic acid, also indicate that the anti-virals are impacting greatly
upon central metabolism. Unfortunately despite all amino acids
being present in our in house metabolite libraries further amino
acids and organic acid intermediates of the TCA cycle were not
detected and/or identified by GC-MS, although the amino acids
proline and methionine were shown to be reduced in C33AE6 cells
and phenylalanine increased in C33AE6 cells with increased anti￾viral dosing by UPLC-MS profiling. The lack of other amino and
organic acids detected by GC-MS profiling indicates the need for
further sample bulking in order to produce highly metabolite rich
extracts in future experimentation. Octadecanoic acid levels were
also seen to decrease with increasing levels of anti-viral exposure
which could potentially also be related to the cellular stresses
induced by the anti-viral modes of action.
In conclusion we have demonstrated that a combination of
UPLC-MS based metabolic profiling with appropriate chemo￾metric analysis is a valuable approach for studying cellular
responses to anti-viral drugs. In addition, we have quantified
different intracellular drug levels in C33AP and E6 cells which
suggest, certainly in the case of lopinavir, that increased activity
of membrane transporters may contribute to the drug sensitivity of
HPV infected cells, no previous work has been carried out in this
area. In future studies, the application of several metabolomics
platforms (i.e., UPLC-MS, GC-MS, and potential UPLC-SPE-NMR
for aiding in identification),53–55 following the same regime as
presented for UPLC-MS here, and/or the use of radiolabeled anti￾viral compounds for flux analyses,56 could potentially uncover a
large area of effected metabolism leading to in-depth insights as
to the anti-viral modes of action.
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
410 | Mol. BioSyst., 2014, 10, 398–411 This journal is © The Royal Society of Chemistry 2014
Conflicts of interest
The authors have no conflict of interest.
Acknowledgements
This research was funded by an ORS award to D-H.K. In addition,
J.W.A., Y.X. and R.G. wish to thank Cancer Research UK (including
Experimental Cancer Medicine Centre award) and the Wolfson
Foundation. E.C. and R.G. also acknowledge the EU Commonsense
(www.fp7projectcommonsense.eu/) project (Grant 261809) financed
by the European Commission under the seventh Framework
Programme for Research and Technological Development. R.G.
and W.B.D. are also grateful to both the UK BBSRC and EPSRC
for financial support of the MCISB. The authors also wish to
acknowledge the support of the Humane Research Trust, the
Caring Cancer Research Trust and Cancer Research UK.
References
1 H. zur Hausen, Biochim. Biophys. Acta, Rev. Cancer, 1996,
1288, 55–78.
2 E. M. de Villiers, J. Virol., 1989, 63, 4898.
3 M. Scheffner, B. A. Werness, J. M. Huibregtse, A. J. Levine
and P. M. Howley, Cell, 1990, 63, 1129–1136.
4 D. M. Harper, E. L. Franco, C. Wheeler, D. G. Ferris,
D. Jenkins, A. Schuind, T. Zahaf, B. Innis, P. Naud,
N. S. De Carvalho, C. M. Roteli-Martins, J. Teixeira,
M. M. Blatter, A. P. Korn, W. Quint and G. Dubin, Lancet,
2004, 346, 1757–1765.
5 N. Munoz, F. X. Bosch, S. de Sanjose, R. Herrero,
X. Castellsague, K. V. Shah, P. J. F. Snijders and
C. J. L. M. Meijer, N. Engl. J. Med., 2003, 348, 518–527.
6 L. L. Villa, R. L. R. Costa, C. A. Petta, R. P. Andrade, K. A. Ault,
A. R. Giuliano, C. M. Wheeler, L. A. Koutsky, C. Malm,
M. Lehtinen, F. E. Skjeldestad, S. E. Olsson, M. Steinwall,
D. R. Brown, R. J. Kurman, R. M. Ronnett, M. H. Stoler,
A. Ferenczy, D. M. Harper, G. M. Tamms, J. Yu, L. Lupinacci,
R. Railkar, F. J. Taddeo, K. U. Jansen, M. T. Essser, H. L. Sings,
A. J. Saah and E. Barr, Lancet Oncol., 2005, 6, 271–278.
7 M. Kyrgiou, G. Koliopoulos, P. Martin-Hirsch, M. Arbyn,
W. Prendiville and E. Paraskevaidis, Lancet, 2006, 367,
489–498.
8 H. zur Hausen, J. Natl. Cancer Inst., 2000, 92, 690–698.
9 J. M. Huibregtse, M. Scheffner and P. M. Howley, EMBO J.,
1991, 10, 4129–4135.
10 J. M. Huibregtse, M. Scheffner and P. M. Howley, Mol. Cell.
Biol., 1993, 13, 775–784.
11 L. Banks, D. Pim and M. Thomas, Trends Biochem. Sci., 2003,
28, 452–459.
12 M. Thomas, D. Pim and L. Banks, Oncogene, 1999, 18,
7690–7700.
13 S. Wagner, K. Scholz, M. Donegan, L. Burton, J. Wingate and
W. Voelkel, Anal. Chem., 2006, 78, 1296–1305.
14 M. Carrascal, K. Schneider, R. E. Calaf, S. van Leeuwen,
D. Canosa, E. Gelpi and J. Abian, J. Pharm. Biomed. Anal.,
1998, 17, 1129–1138.
15 O. P. Bondar, D. R. Barnidge, E. W. Klee, B. J. Davis and
G. G. Klee, Clin. Chem., 2007, 53, 673–678.
16 T. Kind, V. Tolstikov, O. Fiehn and R. H. Weiss, Anal.
Biochem., 2007, 363, 185–195.
17 B. Winther and J. L. Reubsaet, J. Sep. Sci., 2007, 30, 234–240.
18 C. Chen, X. Ma, M. A. Malfatti, K. W. Krausz, S. Kimura,
J. S. Felton, J. R. Idle and F. J. Gonzalez, Chem. Res. Toxicol.,
2007, 20, 531–542.
19 C. Chen, F. J. Gonzalez and J. R. Idle, Drug Metab. Rev., 2007,
39, 581–597.
20 P. Begley, S. Francis-McIntyre, W. B. Dunn, D. I. Broadhurst,
A. Halsall, A. Tseng, J. Knowles, R. Goodacre and D. B. Kell,
Anal. Chem., 2009, 81, 7038–7046.
21 J. C. W. Rijk, A. A. C. M. Peijnenburg, M. H. Blokland,
A. Lommen, R. L. A. P. Hoogenboom and T. F. H. Bovee,
Chem. Res. Toxicol., 2012, 25, 1720–1731.
22 L. Hampson, H. C. Kitchener and I. N. Hampson, Antiviral
Ther., 2006, 11, 813–825.
23 D.-H. Kim, R. M. Jarvis, Y. Xu, A. W. Oliver, J. W. Allwood,
L. Hampson, I. N. Hampson and R. Goodacre, Analyst, 2010,
135, 1235–1244.
24 D.-H. Kim, R. M. Jarvis, J. W. Allwood, G. Batman, R. E. Moore,
E. Marsden-Edwards, L. Hampson, I. N. Hampson and
R. Goodacre, Anal. Bioanal. Chem., 2010, 398, 3051–3061.
25 L. Hampson, E. El Hady, J. V. Moore, H. Kitchener and
I. N. Hampson, FASEB J., 2001, 15, 1445–1447.
26 T. Sangster, H. Major, R. Plumb, A. J. Wilson and
I. D. Wilson, Analyst, 2006, 131, 1075–1078.
27 W. B. Dunn, I. D. Wilson, A. W. Nicholls and D. Broadhurst,
Bioanalysis, 2012, 4, 2249–2264.
28 W. B. Dunn, D. Broadhurst, P. Begley, E. Zelena, S. Francis￾McIntyre, N. Anderson, M. Brown, J. D. Knowles, A. Halsall,
J. N. Haselden, A. W. Nicholls, I. D. Wilson, D. B. Kell,
R. Goodacre and T. H. S. M. H. Consortium, Nat. Protoc.,
2011, 6, 1060–1083.
29 M. C. Thomas and C.-M. Chiang, Mol. cell, 2005, 17, 251–264.
30 R. van den Berg, H. Hoefsloot, J. Westerhuis, A. Smilde and
M. van der Werf, BMC Genomics, 2006, 7, 142.
31 L. W. Sumner, A. Amberg, D. Barrett, M. H. Beale, R. Beger,
C. A. Daykin, T. W. M. Fan, O. Fiehn, R. Goodacre,
J. L. Griffin, T. Hankemeier, N. Hardy, J. Harnly,
R. Higashi, J. Kopka, A. N. Lane, J. C. Lindon, P. Marriott,
A. W. Nicholls, M. D. Reily, J. J. Thaden and M. R. Viant,
Metabolomics, 2007, 3, 211–221.
32 M. Brown, W. B. Dunn, P. Dobson, Y. Patel, C. L. Winder,
S. Francis-McIntyre, P. Begley, K. Carroll, D. Broadhurst,
A. Tseng, N. Swainston, I. Spasic, R. Goodacre and
D. B. Kell, Analyst, 2009, 134, 1322–1332.
33 M. Brown, D. C. Wedge, R. Goodacre, D. B. Kell, P. N. Baker,
L. C. Kenny, M. A. Mamas, L. Neyses and W. B. Dunn,
Bioinformatics, 2011, 27, 1108–1112.
34 S. O’Hagan, W. B. Dunn, M. Brown, J. D. Knowles and
D. B. Kell, Anal. Chem., 2005, 77, 290–303.
Molecular BioSystems Paper
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online
This journal is © The Royal Society of Chemistry 2014 Mol. BioSyst., 2014, 10, 398–411 | 411
35 J. Kopka, N. Schauer, S. Krueger, C. Birkemeyer, B. Usadel,
E. Bergmuller, P. Dormann, W. Weckwerth, Y. Gibon,
M. Stitt, L. Willmitzer, A. R. Fernie and D. Steinhauser,
Bioinformatics, 2005, 21, 1635–1638.
36 J. Hummel, N. Strehmel, J. Selbig, D. Walther and J. Kopka,
Metabolomics, 2010, 6, 322–333.
37 R. Goodacre, D. Broadhurst, A. K. Smilde, B. S. Kristal,
J. D. Baker, R. Beger, C. Bessant, S. Connor, G. Capuani,
A. Craig, T. Ebbels, D. B. Kell, C. Manetti, J. Newton,
G. Paternostro, R. Somorjai, M. Sjostrom, J. Trygg and
F. Wulfert, Metabolomics, 2007, 3, 231–241.
38 I. T. Jolliffe, Principal Component Analysis, Springer-Verlag,
New York, USA, 1986.
39 L. W. Sumner, P. Mendes and R. A. Dixon, Phytochemistry,
2003, 62, 817–836.
40 R. Goodacre, E´. M. Timmins, R. Burton, N. Kaderbhai,
A. M. Woodward, D. B. Kell and P. J. Rooney, Microbiology,
1998, 144, 1157–1170.
41 H. Wold, Multivariate Analysis, Academic Press, New York,
USA, 1966.
42 Y. Xu and R. Goodacre, Metabolomics, 2012, 8, 37–51.
43 J. A. Westerhuis, T. Kourti and J. F. Macgregor, J. Chemom.,
1998, 12, 301–321.
44 A. K. Smilde, J. A. Westerhuis and S. de Jong, J. Chemom.,
2003, 17, 323–337.
45 Y. Benjamini and Y. Hochberg, J. R. Statist. Soc. B, 1995, 57,
289–300.
46 A. Persidis, Nat. Biotechnol., 1999, 17, 94–95.
47 S. L. McKenna and R. A. Padua, Br. J. Haematol., 1997, 96,
659–674.
48 N. Keshelava, J. J. Zuo, P. Chen, S. N. Waidyaratne,
M. C. Luna, C. J. Gomer, T. J. Triche and C. P. Reynolds,
Cancer Res., 2001, 61, 6185–6193.
49 M. A. Noor, O. P. Flint, J. F. Maa and R. A. Parker, AIDS,
2006, 20, 1813–1821.
50 Q. Yan and P. W. Hruz, J. Acquired Immune Defic. Syndr.,
2005, 40, 398–403.
51 G. Batman, L. Hampson and I. N. Hampson, Future Virol.,
2011, 6, 1021–1023.
52 T. Mesple`de, D. Gagnon, F. Bergeron-Labrecque, I. Azar,
H. Se´ne´chal, F. Coutle´e and J. Archambault, J. Virol., 2012,
86, 94–107.
53 S. Moco, R. J. Bino, R. C. H. De Vos and J. Vervoort, TrAC,
Trends Anal. Chem., 2007, 26, 855–866.
54 E. C. Tatsis, S. Boeren, V. Exarchou, A. N. Troganis, J. Vervoort
and I. P. Gerothanassis, Phytochemistry, 2007, 68, 383–393.
55 G. Glauser, E. Grata, L. Dubugnon, S. Rudaz, E. E. Farmer
and J.-L. Wolfender, J. Biol. Chem., 2008, 283, 16400–16407.
56 G. G. Harrigan, J. Colca, S. Szalma and L. G. Boros, Metabolomics,
2006, 2, 21–29.
Paper Molecular BioSystems
Published on 06 January 2014. Downloaded by University of Pittsburgh on 28/10/2014 07:51:39. View Article Online