Toxicology 313 (2013) 113–121

Contents lists available at ScienceDirect

Toxicology
journal homepage: www.elsevier.com/locate/toxicol

In vitro approach to predict post-translational phosphorylation response to

mixtures
Jonathan Boyd ∗ , Julie A. Vrana, Holly N. Williams
C Eugene Bennett Department of Chemistry, West Virginia University, Morgantown, WV 26506, United States

a r t i c l e i n f o a b s t r a c t

Article history: While exposure to chemical mixtures is an everyday reality, an understanding of their combined effects,
Received 18 April 2012 and any potential prediction thereof, is extremely limited. Realistic exposures potentially consist of
Received in revised form hundreds to thousands of chemicals per day, but even relatively simple binary mixture interactions
19 September 2012
can be inherently difficult to predict based upon the lack of temporal and spatial mechanisms for the
Accepted 1 October 2012
individual constituents. To this end, we explore the concept of capitalizing on downstream convergence
Available online 9 November 2012
of intracellular signal transduction to experimentally simplify the means of determining xenobiotics that,
when combined, could result in increased or unexpected toxicity. In a proof of principle study, we exposed
Keywords:
Mixtures
HepG2 cells to deguelin, a natural isoflavonoid, alone and in combination with KCN, and determined the
Signal transduction relative post-translational phosphorylation responses to several key proteins related to mitochondrial
Apoptosis outer membrane permeabilization. Dose-dependent phosphorylation activity provides a clear identifi-
cation of threshold response to low-level exposures, and crosstalk amongst selected proteins correctly
forecasts mixtures interactions that may lead to increased toxicity. We then used Bliss Independence
to determine if the experimentally measured mixture phosphorylation responses could be predicted
with individual responses. Independence accurately predicted mixture interactions for deguelin and
KCN (87.5%). To more fully exhaust independence as a model for determining potential pharmacody-
namic interactions, we exposed HepG2 cells to deguelin and staurosporine, a broad kinase inhibitor;
independence accurately predicted these mixture responses (77.5%). In this study, we demonstrate the
potential of a new in vitro approach for the prediction of toxic mixtures interactions that is fundamentally
driven by the interdependence of signal transduction and apoptosis.
© 2012 Elsevier Ireland Ltd. All rights reserved.

1. Introduction
Exposure to chemical mixtures is an everyday reality, how-
ever an understanding of their combined effects, and any potential
prediction thereof, is extremely limited. From an environmental
perspective, realistic exposures potentially consist of hundreds to
thousands of chemicals per day. From a pharmaceutical perspec-
tive, adverse drug–drug interactions are of toxicological concern,
but novel regimens involving mixtures have also shown therapeu-
tic potential (Ma and Waxman, 2008; Cottarel and Wierzbowski,
2007). Understanding and ultimately predicting chemical mixture
effects could not only elucidate potential risk factors associated
with exposures, but also offers the opportunity to develop a greater
understanding of cellular mechanisms for enhanced drug discovery
(Jonker et al., 2005; Fitzgerald et al., 2006).
∗ Corresponding author at: P.O. Box 6045, Morgantown, WV 26506, United States.
Tel.: +1 304 615 9627; fax: +304 293 4904.
E-mail address: jonathan.boyd@mail.wvu.edu (J. Boyd).
Mixtures research takes many forms, but the basic conceptual
framework is simple: the physiological response to a mixture of
xenobiotics may be significantly different than the contributions
of its individual constituents. These differences are the result of
biochemical–xenobiotic interactions which can be pharmacoki-
netic and/or pharmacodynamic in nature. Most current mixtures
research is focused on pharmacokinetics (absorption, distribution,
metabolism and excretion) because alterations in the means by
which the body processes a primary xenobiotic can increase or
decrease concentrations of a secondary xenobiotic in the blood
and tissues by an order of magnitude or more (U.S. DHHS, 2006;
for reviews see Zhang et al., 2009a, 2009b). On the other hand,
pharmacodynamic mixture effects can occur when the tissue- or
cellular-level response to the primary xenobiotic enhances sensi-
tivity to the secondary exposure; these responses are thought to
amplify or dampen the effects of secondary xenobiotics (Wu et al.,
2010; Natarajan et al., 2006).
The pharmacodynamic response of cells to xenobiotics
is primarily coordinated by signal transduction networks
which typically follow a fundamental motif, the phospho-
rylation/dephosphorylation cycle mediated by kinases and

0300-483X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tox.2012.10.010
114 J. Boyd et al. / Toxicology 313 (2013) 113–121
phosphatases (Kholodenko, 2006; Sauro and Kholodenko, 2004).
This simple cycle is embedded in early response networks that use
positive and negative feedback to generate extremely diverse func-
tions like signal amplification and perfect adaptation, reversible
and irreversible switches, homeostasis, and oscillations (Tyson
et al., 2003). One key irreversible switch is the mitochondrial
pathway of apoptosis, and both the signaling network and infra-
structure surrounding the decision to flip the switch is complex.
Multiple signals, generated by either external or internal stresses,
converge upon the mitochondria and result in mitochondrial
outer membrane permeabilization (MOMP), which is currently
considered the point of no return for apoptosis (i.e., irreversible
switch) (Chipuk and Green, 2008; Galluzzi et al., 2012).
In the study presented here, we propose that by capitalizing
on downstream convergence of signal transduction toward MOMP,
one may experimentally simplify the means for determining xeno-
biotics that, when combined, could result in increased apoptosis.
Rather than assess all possible interactions between xenobiotics
and multiple cellular targets, this paper suggests that it may be
possible to screen this pharmacodynamic response of cells to indi-
vidual compounds and map their crosstalk for the prediction of
significant interactions. This approach is not reliant on the entire
xenobiotic mechanism, but instead, it monitors the cellular inter-
pretation of the exposure; thus, it has the potential to capture both
on- and off-target activity. A benefit of this type of approach is that
it can identify doses of xenobiotics that may lead to aberrant mix-
ture effects, rather than focusing on the mode(s) of action of the
individual xenobiotics.
In this initial proof of principle study, we exposed human hepa-
tocellular carcinoma-derived cells (HepG2) to two mitochondrial
inhibitors, deguelin and KCN, to test our hypothesis that spatiotem-
porally matched post-translational phosphorylation of proteins
related to MOMP could be predictive of mixtures interactions that
lead to enhanced toxicity. Deguelin is a natural isoflavonoid that
inhibits NADH ubiquinone oxidoreductase (complex I), AKT (pro-
tein kinase B) and HSP90; further, it has shown promise as an
effective agent for a variety of cancer types, including hepatocel-
lular carcinoma (Lee et al., 2005; Peng et al., 2007; Oh et al., 2007,
2008). KCN is an inhibitor of cytochrome c oxidase (complex IV) of
the mitochondrial electron transport chain, and is often released
in the body following ingestion of cyanogenic glycosides which
are found in over 2000 plants, including sorghum and cassava
(Haque and Bradbury, 2002). To further evaluate the boundaries of
this approach, we co-exposed HepG2 cells to deguelin and stau-
rosporine. Staurosporine is a microbial alkaloid natural product
and broad kinase inhibitor that has affinity for the ATP bind-
ing site of numerous kinases (Omura et al., 1995; Fabian et al.,
2005).
2. Materials and methods
2.1. Materials
Deguelin, potassium cyanide, and staurosporine were obtained from Sigma
Aldrich (St. Louis, MO). RPMI-1640 containing phenol red, RPMI-1640 without
phenol red, sodium pyruvate, HEPES, l-glutamine, fetal bovine serum, and penicillin-
streptomycin were obtained from Invitrogen (Carlsbad, CA). Cell lines and MTT
assay kits were obtained from American Type Culture Collection (Manassas, VA).
MitoXpress oxygen probe was obtained from Luxcel Corporation (Cork, Ireland).
2.2. Cell culture
Human hepatocellular carcinoma-derived HepG2 cells were cultured in RPMI-
1640, supplemented with 1 mM sodium pyruvate, 5 mM HEPES, 2 mM l-glutamine,
10% fetal bovine serum, 100 U/ml penicillin, and 100 ␮g/ml streptomycin. Cells were
maintained in a humidified atmosphere at 37 ◦ C, 5% CO2 and passaged at 80% con-
fluence.
2.3. Dosing
Cells were seeded into clear-bottom, black-sided 96-well plates at a concen-
tration of 4 × 104 cells per well in RPMI-1640 without phenol red and allowed to
grow for 24 h before dosing. Media was then aspirated from wells and cells were
challenged with varying concentrations of single and mixed compounds (1 ␮L injec-
tions) in fresh media (100 ␮L). Compounds were prepared in either DMSO (deguelin,
staurosporine) or water (KCN) to microplate well concentrations of 0.01 ␮M to
100 ␮M. Media and injection volumes were held constant for all assays and DMSO
well content was <1%.
2.4. MTT assay
After 24 h of exposure to single compounds or mixtures, cell viability was deter-
mined using the MTT assay, according to the manufacturer’s protocol. The assay is
based on the reduction of tetrazolium MTT to formazan by metabolically active
cells, in part by the action of dehydrogenase enzymes, to generate reducing equiv-
alents such as NADH and NADPH. Briefly, MTT reagent was added to the wells of
the microplate, and after two hours of incubation at 37 ◦ C, intracellular formazan
crystals were solubilized with the provided detergent solution. Absorbance values
were obtained using the Safire2 microplate reader (Tecan US, Raleigh, NC) with a
measurement wavelength of 570 nm and a reference wavelength of 700 nm, read
from the bottom. Experiments were each performed at least in triplicate on separate
days.
2.5. Oxygen consumption assay
Immediately after dosing single compounds, cellular oxygen consumption was
assessed using the MitoXpress probe, according to manufacturer’s protocol. Briefly,
oxygen-sensitive probe was diluted to a stock concentration of 1 ␮M, and stock
probe was diluted 1:15 in each well of a 96-well plate containing cells; 100 ␮L of pre-
warmed mineral oil was also added to each well to block ambient oxygen from the
cells. After pre-warming the plates at 37 ◦ C for 1 h, cells were challenged with varying
concentrations of single compounds. Immediately following addition of compound,
oxygen consumption was determined by measuring fluorescence. Fluorescent signal
was obtained using the Safire2 microplate reader (Tecan US, Raleigh, NC) with exci-
tation wavelength of 380 nm and emission wavelength of 650 nm, reading from the
bottom every 2 min for 24 h after dosing. Experiments were performed in triplicate
on separate days.
2.6. Bio-plex multiplex immunoassay
After 400 min exposure to increasing doses of deguelin (0.01, 0.1, 1.0, 10,
100 ␮M) in 1% DMSO alone and in combination with the EC50 concentrations of stau-
rosporine (10 ␮M) or KCN (0.01–100 ␮M) obtained from Sigma Aldrich (St. Louis,
MO), cells were lysed using lysis buffer (BioRad, Hercules, CA) with 500 ␮M phenyl-
methanesulfonylfluoride (PMSF) (Sigma, St. Louis, MO) and phosphatase inhibitors
(BioRad, Hercules, CA). Total protein concentration was determined using the DC
Protein Assay (BioRad, Hercules, CA) according to the manufacturer’s instructions.
Protein phosphorylation was detected using multiplex assays analyzed with the
flow based BioPlex suspension array system (Bio-Rad, Hercules, CA). Beads con-
taining antibodies against phosphorylated ERK1/2 (Thr202/Tyr204, Thr185/Tyr187),
AKT (Ser473), HSP27 (Ser78), I␬B␣ (Ser32/Ser36), JNK (Thr183/Tyr185), p38MAPK
(Thr180/Tyr182), p53 (Ser15), and p90RSK (Thr359/Ser363) were obtained from Bio-
Rad (Hercules, CA). All experiments were performed on separate days in triplicate
and error bars reflect standard error of the mean.
2.7. Statistical analysis
Dose–response curves for MTT assays were generated by best-fit Hill-plot
regression of scatter plot data using Prism V5 (Graphpad Software, San Diego, CA).
Oxygen consumption curves were generated by choosing the best-fit polynomial
regression of scatter plot data using Prism V5 (Graphpad Software, San Diego, CA).
The time point of interest (400 min) was selected using SAS JMP V8 (Cary, NC) where
the change in the slope of the curve reached a minimum for most exposures. Sta-
tistical significance was assessed by using a two-way analysis of variance (ANOVA)
with Bonferroni post-test. A difference at P < 0.05 level was considered statistically
significant. For all graphs, error bars reflect standard error of the mean.
3. Results and discussion
3.1. KCN enhances deguelin toxicity
For this initial study, deguelin and KCN were selected due to
their intracellular activity: we hypothesize that deguelin, a com-
plex I, AKT and HSP90 inhibitor, and KCN, a complex IV inhibitor,
will have both convergent and divergent post-translational phos-
phorylation responses that may be used to determine mixtures
 J. Boyd et al. / Toxicology 313 (2013) 113–121
Fig. 1. KCN enhances the toxicity of deguelin in HepG2 cells. HepG2 viability dose–response curves for deguelin and KCN, alone and in combination. Viability was measured
by an MTT assay and is shown as percent viability. Percent viability was calculated relative to control cells which received dosing vehicle (<1% DMSO), but no deguelin or
KCN. Mixture responses found to be significantly different (p < 0.05) from individual deguelin or KCN are marked with * and # , respectively. Vertical dashed line indicates the
deguelin EC50 with gray highlight representing standard error of the mean (SEM).
interactions. To establish if these disparate xenobiotics lead to
increased toxicity when combined as a mixture, we first measured
viability of HepG2 cells in response to 24 h exposures of deguelin
and KCN, alone and in combination (Fig. 1). Individually, these
compounds showed greatly different dose–response curves with
calculated effective concentrations resulting in 50% viability (EC50 )
of 40.2 ± 1.2 ␮M for deguelin and >100 ␮M (i.e., non-toxic at the
doses measured) for KCN. However, when combined, the mixture
dose–response curves dramatically change. Even the lowest con-
centration of KCN tested (0.01 ␮M) enhanced deguelin toxicity, as
shown in Fig. 1. For all mixture combinations with KCN, the average
deguelin EC50 was 18.1 ± 3.0 ␮M.
3.2. Observed phosphorylation responses
We next explored the signal transduction response, by means
of phosphorylation activity, of HepG2 cells to each xenobiotic.
Unfortunately to date, there is not an assay that can monitor
post-translational phosphorylation in real-time without poten-
tially interfering with the response of our protein targets of interest.
Therefore, a method for the temporal estimation of critical signaling
events is required. The ideal approach to determine key time points
in signal transduction response following exposure to deguelin and
KCN would involve an extracellular method, so as to decrease the
risk of additional intracellular interference. Both of these com-
pounds inhibit mitochondrial electron transport chain activity,
which directly impacts the availability of ATP derived from oxida-
tive phosphorylation. Relative rates of oxidative phosphorylation
can be readily estimated extracellularly by the measurement of
oxygen consumption. Measuring oxygen consumption in response
to inhibitors is a proven technique that has been used to identify
key temporal events relevant to both signal transduction (Jannsen-
Heininger et al., 2008) and apoptosis progression (Sung et al., 2010;
Tuttle et al., 2007). Thus, we measured the oxygen consumption
of HepG2 cells in response to the individual compounds, deguelin
and KCN (Appendix 1). From this assay, doses of deguelin and
115
KCN initially inhibit oxygen consumption (as compared to controls
which have been normalized to a value of 1). However, with the
exception of 10 and 100 ␮M doses of deguelin, the cells appear to
recover from both deguelin and KCN induced oxygen consumption
inhibition and are shown to consume oxygen at rates equal to, or
greater than, controls around 400 min post-exposure, implicating
this time-point as a key signaling event.
With this critical signaling time-point in hand, we simulta-
neously measured the phosphorylation of ERK1/2 (Thr202/Tyr204,
Thr185/Tyr187), AKT (Ser473), HSP27 (Ser78), I␬B␣ (Ser32/Ser36),
JNK (Thr183/Tyr185), p38MAPK (Thr180/Tyr182), p53 (Ser15),
and p90RSK (Thr359/Ser363) at 400 min post-exposure to both
deguelin and KCN, alone and in combination. Our aim with
the selection of these protein targets was to cast a wide net
with respect to signal transduction proteins relevant to MOMP,
and each of these proteins has demonstrated regulatory activ-
ity over BCL-2 proteins that ultimately govern this process (Cory
et al., 2003; Green and Kroemer, 2004). Fig. 2A shows the rel-
ative phosphorylation response of HepG2 cells to exposures of
increasing doses of deguelin and KCN, individually (normalized
to controls). Two-way ANOVA with Bonferroni post-test deter-
mined significant differences (p < 0.05) between deguelin and
control for ERK1/2, p38MAPK, and p90RSK at 1 ␮M; ERK1/2,
p38MAPK and HSP27 at 10 ␮M; p38MAPK, I␬B␣ and p53 at 100 ␮M
(Fig. 2A). Thus, a sub-lethal threshold response exists for HepG2
cells exposed to deguelin at doses between 0.1 and 1 ␮M. KCN
was not found to be significantly different from control for any
protein–dose combination, which is in agreement with viability
data that demonstrated all doses of KCN that were tested are
non-toxic.
To identify signal transduction proteins that may be related to
the toxicity of the mixture of deguelin and KCN, we compared the
observed mixture protein phosphorylation responses to those of
its individual contributors. Results are shown in Fig. 2B but briefly,
the phosphoprotein response of the mixture was found to be sig-
nificantly different (p < 0.05) from deguelin for 17 protein-dose
116 J. Boyd et al. / Toxicology 313 (2013) 113–121

Fig. 2. The relative phosphorylation response of eight proteins from HepG2 cells exposed to deguelin, KCN and their mixture. (A) Phosphorylation response of protein targets
(from cell lysates) to 400 min exposures of deguelin and KCN, individually. Phosphorylation was measured in HepG2 cell lysates (following exposures) using a multiplex
bead-based ELISA assay. Phosphorylation is shown relative to control cells, which received dosing vehicle (<1% DMSO), but no deguelin or KCN. Phosphorylation responses
found to be significantly different (p < 0.05) from controls are marked by * . (B) Phosphorylation response of protein targets to 400 min exposures of mixtures of deguelin and
KCN. Phosphorylation responses found to be significantly different (p < 0.05) from deguelin or KCN are marked with * and # , respectively.

combinations that involved JNK (0.01 ␮M DEG + 100 ␮M KCN), I␬B␣
(0.01 ␮M DEG + 100 ␮M KCN; 0.10 ␮M DEG + 0.10–10 ␮M KCN),
HSP27 (10 ␮M DEG + 0.01 ␮M KCN), p90RSK (10 ␮M DEG + 100 ␮M
KCN), and p38MAPK (10,100 ␮M DEG + 0.01–100 ␮M KCN). Due to
the general lack of protein phosphorylation activity in response
to individual KCN, it was not surprising that the phosphoprotein
response of the mixture was found to be significantly different
(p < 0.05) from KCN for many more protein-dose combinations
(65 total, see Fig. 2B). I␬B␣, JNK, and p38MAPK phosphorylation
responses from mixtures were significantly different from both
deguelin and KCN, individually, and these may be contributing fac-
tors to the increased toxicity of the mixture.
 J. Boyd et al. / Toxicology 313 (2013) 113–121 117

Fig. 3. Observed and predicted relative phosphorylation responses of HepG2 cells exposed to mixtures of deguelin and KCN. Phosphoprotein responses are shown as rows,
and mixture combinations are shown in columns. Observed responses found to be significantly different (p < 0.05) from predicted responses are marked with * .

3.3. Using independence to determine signal transduction
mixture effects
In addition to experimental collection of signal transduction
responses to disparate chemicals (spatiotemporally matched
by protein target and time course), understanding and ulti-
mately predicting mixtures effects requires mathematical models.
Several mathematical techniques for predicting pharmacodynamic
interactions have been proposed (Jonker et al., 2005; Tallarida,
2006, 2007), but most focus on interactions at specific receptor
sites. Anderson and Dennison (2004) introduce the concept of
signal transduction cascades as a means for prediction of potential
pharmacodynamic interactions, but Jonker et al. (2005) extends
this to a mathematical expression based upon the location where
118 J. Boyd et al. / Toxicology 313 (2013) 113–121

Fig. 4. Viability and phosphorylation responses of HepG2 cells to mixtures of deguelin and staurosporine. (A) HepG2 viability in response to mixtures of deguelin and 10 ␮M
staurosporine. Viability, shown as % viability, was measured as above and was calculated relative to control cells which received vehicle (<1% DMSO), but no deguelin or
staurosporine. Mixture viability responses found to be significantly different (p < 0.05) from deguelin alone are marked with * . (B) Phosphorylation response of HepG2 cells
to exposures of 10 ␮M staurosporine (EC50 dose). Phosphorylation was measured in cell lysates (following exposures) as stated above. Phosphorylation is shown relative to
control cells which received vehicle (<1% DMSO). (C) Observed and predicted relative phosphorylation responses of HepG2 cells exposed to mixtures of deguelin plus 10 ␮M
staurosporine. Observed responses found to be significantly different (p < 0.05) from predicted responses are marked with * .
 J. Boyd et al. / Toxicology 313 (2013) 113–121
the interaction is occuring- sequential vs. dual (i.e. convergent)
pathways, which appears to be conceptually based upon Loewe
additivity (Loewe, 1928) and Bliss independence (Bliss, 1939). Bliss
Independence (also known as response addition) states that the
response from the combination of chemicals in a mixture is equal
to the conditional sum of component responses as defined for
the sum of independent event probabilities (Bliss, 1939; U.S. EPA,
2000). Independence represents a simple probabilistic model that
calculates the predicted effects f (xmix ) of a mixture with a known
composition using the expression
 responses for deguelin with staurosporine from their individ-
f (xmix ) = 1 − (1 − F(xi )) (1)
i the 40 protein–dose combinations (8 proteins and 5 different
where F(xi ) can be any measured quantal response of an indi-
vidual chemical. If one uses the phosphorylation of signal
transduction proteins for F(xi ), and the phosphoprotein tar-
gets are sufficiently downstream of any direct activity, then
independence may be able to predict potential interactions
between two or more xenobiotics given the appropriate temporal
scale.
To determine if the phosphoprotein response of a mixture of
deguelin and KCN follows independence, predicted values deter-
mined from Eq. (1) were tested against observed mixture responses.
For a full factorial dosing design (5 × 5) with 8 protein targets,
observed responses (200 total) were found to be significantly differ-
ent (p < 0.05) from predicted in only 25 protein–dose combinations
(see Fig. 3). This results in an overall 87.5% prediction rate. The
differences included 7 out of 8 proteins tested, but most discrepan-
cies involved either ERK1/2 (6) or p38MAPK (11). While changes in
JNK and p38MAPK phosphorylation that may be relevant to mix-
ture viability were not effectively predicted with independence,
I␬B␣ may be contributing to enhanced toxicity (see Fig. 1) and was
accurately predicted at doses that are relevant to our viability data.
Interestingly, AKT is the only protein tested that is known to be
directly inhibited by deguelin (Dell’Eva et al., 2007), and indepen-
dence correctly predicted phosphorylation activity of AKT across
all dosing combinations.
3.4. Signal transduction independence with a broad kinase
inhibitor
While independence was 87.5% accurate for predicting deguelin
and KCN mixture phosphorylation responses, we wanted to explore
the limits of this model by using a broad kinase inhibitor that has
affinity for a large number of kinases. Staurosporine is a microbial
alkaloid that has activity at many different kinases via interactions
with their ATP-binding site, and thus we hypothesize that there
will be a vast range of direct phosphorylation responses for the
8 proteins measured that are relevant to viability. We first deter-
mined viability of staurosporine alone at its manufacturer reported
EC50 (10 ␮M), and in combination with deguelin (Fig. 4A). For stau-
rosporine alone, HepG2 viability was found to be 61.58 ± 2.09%
when normalized to controls. Mixtures of deguelin (0.01–100 ␮M)
with staurosporine (10 ␮M) were statistically different (p < 0.05)
when compared to deguelin alone for the 0.01–10 ␮M doses and
decreased viability by 50% over this range.
We next determined the post-translational phosphorylation
response of the same 8 proteins (as described in Section 3.2)
to exposures of staurosporine alone and in combination with
deguelin. The post-translational phosphorylation responses of
staurosporine alone (10 ␮M), as shown in Fig. 4B, varied in mag-
nitude and all responses measured were statistically different
119
(p < 0.05) from controls, which received dosing vehicle (1% DMSO)
only. Only one phosphoprotein response (p53) was signifi-
cantly increased, whereas all other responses were significantly
decreased from control. As a mixture, deguelin (0.01–100 ␮M)
with staurosporine (10 ␮M) also showed disparate phosphopro-
tein responses of various magnitudes (Fig. 4C). As shown in Fig. 4C,
most phosphoprotein responses were lower than controls; how-
ever, p53 phosphoprotein responses were significantly increased
over controls for deguelin (0.01 and 0.1 ␮M) with staurosporine.
We determined the predicted phosphoprotein mixture
ual responses using independence, as calculated from Eq. (1). For
mixture doses), independence correctly predicted 31 phospho-
protein responses (Fig. 4C), when compared to observed mixture
responses. By using deguelin in combination with staurosporine,
a broad kinase inhibitor exhibiting activity at many kinases,
we found that independence was a reasonable model for the
prediction of post-translational phosphorylation activity with an
accuracy of 77.5%.
In summary, monitoring a limited number of convergent path-
ways, such as MOMP/apoptosis, representative of the cellular
interpretation of most on- or off-target activity, allows one to
screen disparate xenobiotics for potential mixtures interactions.
We believe that this approach may be particularly suited for the
identification of potentially toxic xenobiotic combinations based
upon dose, rather than mode(s) of action. By determining pharma-
codynamic mixture responses of xenobiotics that exhibit activity
at similar sites (deguelin with KCN; electron transport chain) or
xenobiotics with disparate activity at various magnitudes (deguelin
with staurosporine), independence was thoroughly examined as
a model to predict mixture phosphoprotein interactions. Predic-
tions based upon Bliss Independence can give insight as to potential
post-translational phosphorylation responses that could lead to
increased or unexpected mixture effects, which may be a signif-
icant first step toward understanding, controlling, and ultimately
predicting mixtures effects.
Conflict of interest statement
The authors declare that there are no conflicts of interest.
Acknowledgements
The author would like to thank the Defense Advanced Research
Projects Agency, the Johns Hopkins University Applied Physics
Laboratory, and West Virginia University whose funding sup-
ported portions of this research. Further, the author wishes to
thank K.K., K.S., and N.B. for their thoughtful discussions and
reviews of the manuscript, as well as countless others who dis-
cussed the theory of using signal transduction to predict mixtures
interactions.
Appendix A.
Oxygen consumption response of HepG2 cells to deguelin
and KCN (individually) measured over 24 h. Oxygen consump-
tion was measured using MitoXpress extracellular fluorescent
probes (Luxcel Corp, Cork, Ireland) per manufacturer’s proto-
cols. Oxygen consumption is shown relative to control cells
which received vehicle (<1% DMSO), but no deguelin or KCN
(Fig. A1).
120 J. Boyd et al. / Toxicology 313 (2013) 113–121

Fig. A1. Oxygen consumption response of HepG2 cells to deguelin and KCN (individually) measured over 24 h. Oxygen consumption was measured using MitoXpress
extracellular fluorescent probes (Luxcel Corp, Cork, Ireland) per manufacturer’s protocols. Oxygen consumption is shown relative to control cells which received vehicle (less
than 1% DMSO), but no deguelin or KCN. Solid curves indicate oxygen consumption with respective error (dotted curves representing ± S.E.M.). Vertical dashed line indicates
the selected 400 min time-point.

References
Anderson, M.E., Dennison, J.F., 2004. Mechanistic approaches for mixture risk
assessments-present capabilities with simple mixtures and future directions.
Environ. Tox. Pharm. 16, 1–11.
Bliss, C.I., 1939. The toxicity of poisons applied jointly. Ann. Appl. Biol. 26, 585–615.
Chipuk, J.E., Green, D.R., 2008. How do BCL-2 proteins induce mitochondrial outer
membrane permeabilization? Trends Cell Biol. 18, 157–164.
Cory, S., Huang, D.C., Adams, J.M., 2003. The Bcl-2 family: roles in cell survival and
oncogenesis. Oncogene 24, 8590–8607.
Cottarel, G., Wierzbowski, J., 2007. Combination drugs, an emerging option for
antibacterial therapy. Trends Biotechnol. 25, 548–554.
Dell’Eva, R., Ambrosini, C., Minghelli, S., Noonan, D.M., Albini, A., Ferrari, N., 2007. The
Akt inhibitor deguelin, is an angiopreventive agent also acting on the NF-kappaB
pathway. Carcinogenesis 28, 404–413.
Fabian, M.A., Biggs 3rd, W.H., Treiber, D.K., Atteridge, C.E., Azimioara, M.D., Benedetti,
M.G., Carter, T.A., Ciceri, P., Edeen, P.T., Floyd, M., Ford, J.M., Galvin, M., Gerlach,
J.L., Grotzfeld, R.M., Herrgard, S., Insko, D.E., Insko, M.A., Lai, A.G., Lélias, J.M.,
Mehta, S.A., Milanov, Z.V., Velasco, A.M., Wodicka, L.M., Patel, H.K., Zarrinkar,
P.P., Lockhart, D.J., 2005. A small-molecule kinase interaction map for clinical
kinase inhibitors. Nat. Biotechnol. 23, 329–336.
Fitzgerald, J.B., Schoeberl, B., Nielsen, U.B., Orger, P.K., 2006. Systems biology and
combination therapy in the quest for clinical efficacy. Nat. Chem. Biol. 2,
458–466.
Galluzzi, L., Vitale, I., Abrams, J.M., Alnemri, E.S., Baehrecke, E.H., Blagosklonny, M.V.,
Dawson, T.M., Dawson, V.L., El-Deiry, W.S., Fulda, S., Gottlieb, E., Green, D.R.,
Hengartner, M.O., Kepp, O., Knight, R.A., Kumar, S., Lipton, S.A., Lu, X., Madeo,
F., Malorni, W., Mehlen, P., Nuñez, G., Peter, M.E., Piacentini, M., Rubinsztein,
D.C., Shi, Y., Simon, H.U., Vandenabeele, P., White, E., Yuan, J., Zhivotovsky, B.,
Melino, G., Kroemer, G., 2012. Molecular definitions of cell death subroutines:
 J. Boyd et al. / Toxicology 313 (2013) 113–121
recommendations of the Nomenclature Committee on Cell Death 2012. Cell
Death Differ. 19, 107–120.
Green, D.R., Kroemer, G., 2004. The pathophysiology of mitochondrial cell death.
Science 305, 626–629.
Haque, M.R., Bradbury, J.H., 2002. Total cyanide determination of plants and foods
using the picrate and acid hydrolysis methods. Food Chem. 77, 107–114.
Jannsen-Heininger, Y.M.W., Mossman, B.T., Heintz, N., Forman, H.J., Kalyanaraman,
B., Finkel, T., Stamler, J.S., Rhee, S.G., van der Vliet, A., 2008. Redox-based reg-
ulation of signal transduction principles, pitfalls, and promises. Free Rad. Biol.
Med. 45, 1–17.
Jonker, D.M., VIsser, S.A.G., van der Graaf, P.H., Voskuyl, R.A., Danhof, M., 2005.
Towards a mechanism-based analysis of pharmacodynamic drug–drug inter-
actions in vivo. Pharmacol. Ther. 106, 1–18.
Kholodenko, B.N., 2006. Cell signaling dynamics in time and space. Nat. Rev. Mol.
Cell Biol. 7, 165–176.
Lee, H-Y., Oh, S-H., Woo, J.K., 2005. Chemopreventive effects of deguelin, a novel
Akt inhibitor, on tobacco-induced lung tumorigenesis. J. Natl. Cancer Inst. 97,
1695–1699.
Loewe, S., 1928. Die quantitativen probleme der pharmakologie. Ergebn. Physiol. 27,
47–187.
Ma, J., Waxman, D.J., 2008. Combination of antiangiogenesis with chemotherapy for
more effective cancer treatment. Mol. Cancer Ther. 7, 3670–3684.
Natarajan, M., Lin, K.M., Hsueh, R.C., Sternweis, P.C., Ranganathan, R., 2006. A global
analysis of cross-talk in a mammalian cellular signalling network. Nat. Cell Biol.
8, 571–580.
Oh, S.H., Woo, J.K., Jin, Q., 2008. Identification of novel antiangiogenic anticancer
activitieso f deguelin targeting hypoxia-inducible factor-1␣. Int. J. Cancer 122,
5–14.
Oh, S.H., Woo, J.K., Yazici, Y.D., Myers, J.N., Kim, W.-Y., Jin, Q., Hong, S.S., Park, H.-J.,
Suh, Y.-G., Kim, K.-W., Hong, W.K., Lee, H.-Y., 2007. Structural basis for depletion
of heat shock protein 90 client proteins by deguelin. J. Natl. Cancer Inst. 99,
949–961.
Omura, S., Sasaki, Y., Iwai, Y., Takeshima, H., 1995. Staurosporine, a potentially
important gift from a microorganism. J. Antibiot. (Tokyo) 48, 535–548.
121
Peng, X.H., Karna, P., O’Regan, R.M., 2007. Down-regulation of inhibitor of apoptosis
proteins by deguelin selectively induces apoptosis in breast cancer cells. Mol.
Pharmacol. 71, 101–111.
Sauro, H.M., Kholodenko, B.N., 2004. Quantitative analysis of signaling networks.
Prog. Biophys. Mol. Biol. 86, 5–43.
Sung, H.J., Ma, W., Wang, P.Y., Hynes, J., O’Riordan, T.C., Combs, C.A., McCoy, J.P., Bunz,
F., Kang, J.G., Hwang, P.M., 2010. Mitochondrial respiration protects against
oxygen-associated DNA damage. Nat. Commun. 1, 1–8.
Tyson, J.J., Chen, K.C., Novak, B., 2003. Sniffers, buzzers, toggles and blinkers: dynam-
ics of regulatory and signaling pathways in the cell. Curr. Opin. Cell Biol. 15,
221–231.
Tallarida, R.J., 2006. An overview of drug combination analysis with isobolograms.
J. Pharmacol. Exp. Ther. 319, 1–7.
Tallarida, R.J., 2007. Interactions between drugs and occupied receptors. Pharma-
col.Ther. 113, 197–209.
Tuttle, S.W., Maity, A., Oprysko, P.R., Kachur, A.V., Ayene, I.S., Bigalow, J., Koch, C.J.,
2007. Detection of reactive oxygen species via endogenous oxidative pentose
phosphate cycle activity in response to oxygen concentration: implications for
the mechanism of HIF-1alpha stabilization under moderate hypoxia. J. Biol.
Chem. 282, 36790–36796.
U.S. Dept of Health and Human Services, Food and Drug Administration, Center for
Drug Evaluation and Research, Center for Biologics Evaluation and Research.
Guidance for Industry: Drug Interaction Studies – Study Design, Data Analysis,
and Implications for Dosing and Labeling. September 2006.
U.S. EPA. Supplementary Guidance for Conducting Health Risk Assessment of Chem-
ical Mixtures (2000).
Wu, Z., Zhao, X.-M., Chen, L., 2010. A systems biology approach to identify effective
cocktail drugs. BMC Syst. Biol. 4, S2–S7.
Zhang, L., Zhang, Y., Huang, S.M., 2009a. Scientific and regulatory perspectives on
metabolizing enzyme-transporter interplay and its role in drug interactions:
challenges in predicting drug interactions. Mol. Pharm. 6, 1766–1774.
Zhang, L., Zhang, Y.D., Zhao, P., Huang, S.M., 2009b. Predicting drug–drug interac-
tions: an FDA perspective. AAPS J. 11, 300–306.