phenotyping drug polypharmacology via eicosanoid profiling ... · phenotyping drug polypharmacology...
TRANSCRIPT
1
Phenotyping Drug Polypharmacology via Eicosanoid Profiling of Blood
Jiao Song1#
, Xuejun Liu2#
, Tadimeti S. Rao2, Leon Chang
2, Michael J. Meehan
3, Jonathan M. Blevitt
1,
Jiejun Wu1, Pieter C. Dorrestein
3 and Marcos E. Milla
1*
1Discovery Sciences and
2Immunology Research, Janssen Research & Development, LLC;
3Skaggs
School of Pharmacology and Pharmaceutical Sciences, University of California – San Diego
*Corresponding author. 3210 Merryfield Row, San Diego, CA 92121; 858-784-3298;
#These authors contributed equally.
Running title: lipidomics for drug profiling
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
2
Abstract
It is widely accepted that small molecule drugs, despite their selectivity at primary targets, exert
pharmacological effects (and safety liabilities) through a multiplicity of pathways. As such, it has proven
extremely difficult to experimentally assess polypharmacology in an agnostic fashion. Profiling of
metabolites produced as part of physiological responses to pharmacological stimuli provides a unique
opportunity to explore drug pharmacology. 122 eicosanoid lipids in human whole blood were monitored
from 10 different donors upon stimulation with several inducers of immunological responses and
treatment with modulators of prostaglandin (PG) and leukotriene (LT) biosynthesis, including clinical and
investigational molecules. Such analysis revealed differentiation between drugs nominally targeting
different eicosanoid biosynthetic enzymes, or even those designed to target the same enzyme. Profiled
agents –some of them marketed products– affect eicosanoid biosynthesis in ways that cannot be predicted
from information on their intended targets. As an example, we employed this platform to discriminate
drugs based on their ability to silence PG biosynthesis in response to bacterial lipopolysaccharide (LPS),
resulting in differential pharmacological activity in an in vivo model of endotoxemia. Some of the
observed effects are subject to variability among individuals, indicating a potential application of this
methodology to the patient stratification, based on their responses to bench mark drugs and experimental
compounds read on the eicosanome via a simple blood test.
Supplementary key words
Eicosanoid biosynthesis, metabolic phenotyping, drug differentiation, polypharmacology, patient
profiling, LC/MS
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
3
Introduction
The era of personalized and precision medicine for drug development has been heralded, with the lofty
goal of utilizing a patient’s unique genetic profile and disease status to find the right drug match. Several
pharmaceutical and biotechnology companies are placing substantial effort in the development of such
tailored therapeutics, with the prospects of significantly improved clinical outcomes: efficacy associated
with reductions in adverse events (1, 2).
While precision medicine shows already high potential for the discovery of highly efficacious, safer
medicines, targeted drug development requires in-depth understanding of the mechanism of action
(MOA) of experimental compounds, as well as robust systems for accurate profiling (and prediction) of
therapeutic response. The systematic analysis of drugs that interact with more than one molecular target
offers an opportunity to dissect the complexities of drug action, and constitutes the first step for the
construction of computational models to predict polypharmacology (3, 4). Advances in omics platforms
over the past decade have enabled early exploration of global physiological perturbations (genome,
proteome, metabolome, epigenome and lipidome) in health and disease, thus ushering the era of systems
biology. Several approaches drawing from this emerging discipline have been utilized to build drug-target
networks for complex diseases, involving multiple drugs acting on distinct targets. Those efforts have
aided in the formulation of combinatorial therapy hypotheses (5, 6). A less desirable, yet a central aspect
to polypharmacology is ever-present off-target effects. One major reason for de-prioritization of candidate
drugs is the discovery of toxicities, often during late pre-clinical profiling and clinical development
stages. Emerging computational methodologies are starting to provide more efficient means to assess
compound attributes and “drug-likeness” (7). With the increasing availability of large pharmacological
databases (8, 9), in silico prediction algorithms are starting to provide a foundation for the analysis of
interaction networks involving drugs, targets and anti-targets. Complementing in silico approaches,
advances in analytical chemistry, particularly in the field of mass spectrometry, are enabling larger scale
quantification of multiple analytes with unprecedented specificity, sensitivity and accuracy. In spite of
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
4
this progress, methodologies applicable to clinical studies are still lacking the simplicity and throughput
to become widely used. Once these hurdles are overcome, efficient, disease-focused platforms applicable
for patient phenotyping will become essential for stratification based on biochemical differences in their
responses to inducers of disease, and drug action.
Here we describe a molecular phenotyping study of human whole blood, by analysis of eicosanoid lipids
derived from the conversion of membrane phospholipids to arachidonic acid (AA), an early and critical
event in the initiation and establishment of multiple inflammatory responses. The biosynthesis of
eicosanoids occurs well within an hour upon exposure of cells to a host of stimuli, providing an
opportunity to take quick “snapshots” of metabolic pathways driving the production of lipid messengers.
We employed a high-throughput platform including an informatics workflow that allows unsupervised,
simultaneous quantification of 122 eicosanoids in human whole blood, tissues, and other matrices (10).
Several researchers have reported previously on the systematic measurement of eicosanoid production in
whole blood upon stimulation (11, 12), yet the platform that we describe here offers higher resolution,
sensitivity and throughput, making its application to clinical sample analysis, and compound screening or
profiling for drug discovery, more practical. Covered lipids include PGs, TXs, LTs, eoxins, and lipoxins,
shown to be sensitive immunological markers (13-15). When used in survey mode, this platform can
discern human donors based on their responses to stimuli, and silencing with clinically relevant
compounds. Distinct responses against compounds with the same nominal target protein allow
visualization of global pharmacology of individuals with unprecedented resolution, including off-target
responses.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
5
Methods
Compound treatments. Fresh heparinized human whole blood samples from donors that met the
predefined election criteria (no anti-inflammatory treatments) were obtained from the Normal Blood
Donor Service at the Scripps Research Institute on the day of experiment, and used immediately within
30min. HWB samples were processed as previously described (10). Briefly, HWB samples were diluted
1:1 with RPMI-1640 medium and dispensed to 96-well plates at 200 μL/well. Each well was then
pretreated with or without clinical and experimental molecules targeting PG and LT biosynthesis
(Supplemental Table 3), in concentration-response mode (0.03, 0.3, 3, 30 μM) at 37 ͦC for 15 min,
followed by a 30-min stimulation with calcium ionophore, A23187 (final concentration 30 μM) or LPS
from E. coli 0111:B4 (Sigma, St Louis, MO, final concentration 1 μg/ml) at 37 °C. Plates were then
centrifuged at 1300 rpm at 4 ͦC for 10 min and 100 μl supernatants were immediately transferred to 96-
well plate containing 1ml 10% MeOH in water containing a mixture of deuterated internal standards (1ng
each).
Mouse endotoxemia model. Female, Balb/C mice (n=5 /group, 6-8 weeks, ~20 g; Charles River)
acclimated for at least 7 days with free access to food and water with 12 h light/dark cycle. Indomethacin
and rofecoxib were prepared in 20% HPCD at concentrations of 1 and 0.1 mg/mL each and administered
at doses of 1 and 10 mg/Kg, (po) 30 min before ip injection of LPS (20 μg/mouse). Blood samples were
collected 2 h after LPS dose and processed for serum.
Lipidomics profiling. An in-house developed UPLC-MS based lipidomics platform has been described
previously(10). Briefly, human whole blood samples from healthy donors subject to different treatments
were processed employing 96-well format solid phase extraction. Extracted lipids were then subjected to
simultaneous quantitative UPLC-MS analysis in scheduled MRM (sMRM) mode.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
6
TNFα quantification in mouse blood. Mice were pre-dosed with compounds followed by intraperitoneal
administration of LPS. Blood samples were drawn into after the animals were sacrificed. Levels of TNFα
were then quantified employing ELISA kits (R&D Systems, Minneapolis, MN).
Data analysis and visualization. The absolute concentrations (pg/μL) of 122 eicosanoid lipids in
individual sample were quantified, based on a calibration curve generated from serial diluted standard
samples in each experiment, using our customized R-based quantification software. Data analysis,
including principal component analyses (PCA)- based quality control (QC) process to detect outlier
samples, differential production analysis of eicosanoid lipids in response to various biological
perturbations and hierarchical clustering, were conducted using ArrayStudio V7.0 (www.omicsoft.com).
For visualization purposes, both the absolute concentrations of lipids under various biological conditions,
and the ratios of lipids in response to various biological perturbations were log2 transformed, and
presented in heat maps to reflect the expression profiles of multiple lipids under different conditions.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
7
Results
Metabolic Phenotyping. Human whole blood samples from healthy donors were stimulated with the
calcium ionophore A23187 to induce activation of cPLA2, a cellular phospholipase that converts
membrane phospholipids to AA and other fatty acids in response to intracellular calcium mobilization.
Depending on the cell type, AA and other long fatty acyl chain lipids can be converted through enzymatic
and non-enzymatic reactions into multiple species that can be grouped into two main classes: (1) PGs
derived from the initial production of PGA by cyclooxygenases 1 & 2 (COX1/2) and (2) LTs derived
from the initial intermediate LTA4, produced through the combined action of 5-lipoxygenase (5-LO) and
its accessory protein, 5-LO activating protein (FLAP). The heat map views that we present provide an
efficient, facile way to survey individual variations in lipid profiles derived from COX, LOX as well as
non-enzymatic pathways. The absolute amount of each lipid was transformed into log2 values in order to
accommodate a larger dynamic range, and the colors (from blue to red) represent the changes in log2
transformed values. A23187 stimulation produced dramatic increases in multiple eicosanoids
(Supplemental Figure 1). Consistent with previous observations, approximately half of the lipids in the
panel showed significant increase in concentration upon A23187 stimulation for 30 min (10) and many of
them remained elevated above the baseline levels for up to 24 hr. These findings contrast with short half-
lives, typically in the order of minutes, reported for exogenously added eicosanoids in cellular systems.
This remarkable persistency allows testing of serum samples to capture polypharmacological effects of
drugs without undue demands on sample collection and preservation.
Effect of Disease Relevant Stimuli on Eicosanoids. Stimulation of the AA pathway with A23187 offers
a consistent way of profiling multiple donor samples (and drugs) with a standard, yet non-physiological
stimulus. Correlations between eicosanome signatures and their causal physiology can only be found from
examination of putative inducers of inflammation, and their modulation by experimental and marketed
compounds. We employed a set of agents that trigger inflammatory responses through the activation of
neutrophils, macrophages, mast cells and other cell types involved in innate, early immune responses,
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
8
including bacterial LPS, LTE4 and PGE2. LTE4 was chosen as it is a more stable Cys-LT with known
immunomodulatory effects (16) and can drive neutrophilia in humans (17). The changes in eicosanome
in 2 donors induced by A23187 (included as a comparator), LPS, PGE2 and LTE4 are summarized in
Figure 1. In contrast to A23187, the LPS-induced eicosanome occurred with a delayed onset.
Upon application of LTE4 (up to 0.25 μg/mL) and PGE2 (up to 1 μg/ml) to whole blood, in addition to
the added lipids, metabolites of both (such as EXE4, 11t-LTE4 of LTE4 and PGD2, PGJ2, PGE1 of
PGE2) were detected throughout the time course study indicating metabolic conversion of the added
eicosanoid. Surprisingly, LTE4 and PGE2 application also led to detectable changes in eicosanome
perhaps consistent with their immunomodulatory activities (16). Noteworthy are marked increases in
PGD2 following application of PGE2. While no literature precedence exists for the conversion of PGE2
to PGD2, or induction of production of PGD2 by increased PGE2, our evaluation of experimental
procedures has ruled out analytical artifacts (Supplemental Figure 2). When we spiked d4-PGE2 (1μg/ml)
in HWB, no d4-PGD2 was observed. But we noticed a significant increase in the concentrations of PGE2
metabolites previously detected such as PGD2, PGJ2, LXB4 and 11b-PGE2. Additional studies are
needed to elucidate mechanisms of PGE2 dependent PGD2 production in human whole blood.
Human Drug Response Differentiation. We monitored changes in the lipidomics signature of A23187
in human blood with or without pre-incubation with several marketed or experimental COX and LOX
pathway modulators developed for the treatment of multiple inflammatory conditions (Figure 2A;
signature changes for all individuals are shown on Supplemental Figure 3). A23187 stimulation alone
induced changes in 33 lipids relative to untreated condition (Figure 2A, leftmost column). Depending on
inhibition of COX or LOX pathway, observed effects on the lipidome were biased towards the PGs or
LTs, respectively (Figure 2A-C). In addition, off-pathway activities were also observed. The first
generation anti-inflammatory drug discovered over half a century ago, Indomethacin, a dual COX1/2
inhibitor, potently blocked the synthesis of multiple PGs, with modest effects on other branches of the
eicosanome. We observed two distinct differences, one between Indomethacin vs COX-2 inhibitors and
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
9
the second between celecoxib/valdecoxib vs rofecoxib, all known COX-2 inhibitors (Figure 2B). The
selective COX2 inhibitors celecoxib (Celebrex™) and valdecoxib (Bextra™; (18)) displayed inhibitory
effects on certain LTs on top of their expected PG pharmacology (19, 20). In susceptible individuals,
aspirin can trigger asthma, a clinical phenotype shared by several COX-inhibiting NSAIDs. The flux of
AA through COX/LOX pathway generates mediators that have either bronchoconstrictive (CysLTs,
PGD2) or bronchodilatory (PGE2, PGI2) functions. While the exact etiology of aspirin-sensitive asthma
is unknown, shifting of AA metabolism into preponderance of bronchoconstrictor eicosanoids may
contribute to the bronchospasm. Only recently, the safety of COX-2 inhibitors in asthmatic subjects was
established (19, 21) implicating a role for COX-1 pathway/pathway-derived mediators in aspirin-sensitive
asthma. Equally surprising is our observation of only tenuous pharmacological activity across the
surveyed eicosanome for Rofecoxib (Vioxx™(22); Figures 2B and 3B-D), a drug now retired from the
market because of cardiovascular safety liability and mortality (23). It has been suggested that this
liability may stem from inhibition of PGI2 biosynthesis (24). Based on the lack of strong
pharmacological activity across the eicosanome at concentrations well above its Ki, we suggest that
alternative mechanisms may be operative. Consistent with this, it has been suggested that Rofecoxib
conversion to an ionized species under physiological conditions mediates accumulation of maleic
anhydride, which in turn may promote formation of oxidized lipids and increases in low-density
lipoprotein (LDL). LDL could serve as the trigger for the cardiac liabilities seen with this drug. Other
COX1/2 modulators are not ionizable in a similar way to Rofecoxib, and do not trigger the resulting
adverse effect (25-27).
On the LT branch (Figure 2C), the 5-LO inhibitor Zileuton (Zyflo™ (28)) displayed a relatively broad
profile at the two higher concentrations, with moderate pharmacological effects across the eicosanome.
Two experimental drugs, MK591 (Quiflapon (29)) and its predecessor MK886 (30), targeting the 5-LO
activating protein FLAP, showed a cleaner profile, blocking only the production of LTs and leaving the
COX pathway intact. Montelukast (SingulairTM (31)
), a selective antagonist of the LTD4 receptor CysLT1,
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
10
had a relatively circumscribed pharmacological footprint on the eicosanome, affecting exclusively
cysteinyl LT levels. This is somewhat surprising and may indicate the existence of feedback loops
connecting signaling from LT-sensing G protein-coupled receptors to their biosynthetic enzymes. These
observations are consistent with the inhibition of CysLT production in ragweed and mite-stimulated
peripheral blood mononuclear cells from patients with asthma (32) and correlation of clinical response to
Montelukast treatment in asthmatics with decreases in urinary LTE4 levels, a marker for systemic CysLT
tone (33). These findings also raise the question on the relative merits of targeting LTC4 synthase
pathway as an alternative to CysLT receptor antagonism for treating CysLT driven pulmonary conditions.
We also performed a comparison of two 5-LO inhibitors, Zileuton (Zyflo™), a marketed drug that is
efficacious in severe asthma, yet is not widely prescribed because of idiosyncratic elevations in liver
enzymes (34), and MK-0633 (35, 36), an experimental molecule developed by Merck that did not show
efficacy in asthma and chronic obstructive pulmonary disease (COPD) trials (40). We used these two
molecules to illustrate the remarkable differences that can be revealed between two inhibitory molecules
targeting the same enzyme, 5-LO. MK-0633 has much higher potency relative to Zileuton in blocking the
production of LOX pathway eicosanoids, and is more selective as well: Zileuton's off-target effects on the
COX pathway lipids are more significant at the higher concentration range (Figure 2C). Given that MK-
0633, a more selective 5-LOX inhibitor molecule with a superior physiochemical profile relative to
Zileuton, did not meet efficacy end points in proof of concept studies in asthma and COPD asthma trial
(36), the data reveals that mechanisms in addition to 5-LOX inhibition, such a COX inhibition,
contributes to the efficacy of Zileuton.
Inter-subject Variability. Blood samples from 23 human donors were analyzed to understand inter-
individual variability in A23187-induced eicosanoids (Supplemental Figure 3). The results show
remarkable variability in the levels of lipid production. The observed qualitative and quantitative
differences in levels of specific PGs and LTs both in basal and stimulated conditions indicate that there is
an individual “inflammatory response potential”, which is readable through the eicosanome. To rule out
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
11
the possible variations caused by cellular activations or sample handling artifacts, we also demonstrated
that eicosanoid generation is not dependent upon time of sample processing after the blood was collected
from donor. Over two hours of incubation at 37 °C, the primary eicosanoid levels remain relatively
unchanged and no new metabolite was observed in the blood spiked with deuterated AA (Supplemental
Figure 4). Additionally, we took into account the donors’ smoking status (37, 38) and dietary supplements
intake in order to avoid the external inference with the eicosanoid levels. With 2 smokers and 1 person
taking fish oil out of our 23 donors, the data showed no bias due to these two factors. The lower detection
limits (LLODs) and upper detection limits (ULODs) of all the lipids in 10 donors were summarized in the
Supplemental Table 1, as well as their quantification ranges (pg/μL). The measured amounts of
eicosanoids under stimulated conditions in human whole blood are consistent with reported levels (39).
Most of the lipids exhibited inter-donor variation at the baseline measurements, which is intrinsic to the
human blood studies. Because of the interest in baseline eicosanoid quantification, the standard curves
were prepared in RPMI medium rather than blood; this could be a potential source of error/variation due
to differential suppression from media and human blood. Even though no two individuals were identical
in their eicosanome signatures in response to calcium ionophore stimulation and modulations of the
stimulation by various compounds, there are clear evidences for coordinate changes in clusters of
eicosanoids, suggesting the possibility of phenotypic profiling of targeted populations via their eicosanoid
signatures.
Predicting in vivo Pharmacology from Lipidomics Signatures. LPS is a bacterial product and a TLR4
agonist, and LPS activation of monocytes/macrophages leads to elaboration of various cytokines such as
tumor necrosis factor-α, interleukin-1β, interleukin-6 and interleukin-10, and pro-inflammatory lipid
mediators such as PGs, LTs and platelet-activating factor (PAF). Under the conditions of the experiment,
LPS induced the production of mainly COX pathway metabolites (Figure 3A), including PGE2, TXB2,
and 12-HHT (40, 41) with delayed kinetics relative to the calcium ionophore control, starting at 4 h post-
addition. The overall LPS induction profiles of the lipids are summarized in Supplemental Table 2,
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
12
illustrating donor variability. We examined the effects COX1/2 and selective COX-2 inhibitors on LPS
induced lipidomics signature in human whole blood as a model of inflammation. Average changes on this
signature from eight donors are shown on Figure 3B (changes for all individuals are shown on
Supplemental Figure 5). Indomethacin (42) displayed higher potency relative to Celecoxib (43) and
Rofecoxib (22) in inhibiting the production of COX pathway eicosanoids, Rofecoxib exhibited an
inhibitory effect only at the top concentrations tested (3 and 30 μM).
PG metabolites such as TXB2 and PGE2 have modulatory effects on TNFα synthesis and release.
Inhibition of thromboxane (TX) biosynthesis attenuates endotoxin-induced TNFα production (44)
whereas PGE2 suppresses TNFα release, and blockade of PGE2 biosynthesis not only stimulates TNFα
release, but prevents its breakdown (45, 46). Therefore, it would not be possible to predict a priori the
outcome of inhibiting COX pathway. The effect of combinatorial inhibition TXB2 and PGE2, on LPS
induced release of cytokines such as TNFα: would need to be addressed in vivo. We set out to investigate
the outcome of combinatorial inhibition of TXB2 and PGE2 on TNFα release in an endotoxemia model in
vivo. We examined the effects of Indomethacin and Rofecoxib in LPS induced inflammation model in
mouse with serum levels of TNFα as a biomarker of inflammatory response and, PGE2 and TXB2 levels
as measures of COX pathway inhibition. For this, a sub-lethal dose of LPS was given to mice pre-dosed
with Indomethacin, Rofecoxib or saline solution. Figure 4A shows that a limited lipidomics signature
resembling the one detected with human whole blood was observable in blood samples from those
animals, including elevations in TXB2 and PGE2. Again something that could not be predicted a priori,
Indomethacin, but not Rofecoxib, induced a dramatic elevation in TNFα levels (Figure 4B). The serum
levels of PGE2 (Figure 4C) showed strong negative correlation (r = -0.727) to those of TNFα, with a p
value of 0.00006 (Figure 4D). This observation indicated that, in the mouse, up-regulation of TNFα
production in response to LPS by inhibition of PGE2 biosynthesis may over-ride down-regulation
expected from silencing of TXB2 biosynthesis. While NSAIDs are successfully used to treat several
inflammatory conditions in clinical practice, the observations from the present study shed light on
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
13
potential side effects induced by NSAIDs and selective COX-2 inhibitors. For example, literature reports
indicate that NSAIDs such as indomethacin increased LPS induced elaboration of IL-6 and TNFα in
human whole blood, and exogenously added PGE2 attenuated LPS-induced cytokine response and
reversed the effects of Indomethacin. Interestingly, selective COX-2 inhibition (NS398), but not selective
COX-1 inhibition (SC-560), exerted a stimulatory effect on the expression of pro-inflammatory cytokines
(47). In a more translational study, Endres and coworkers showed that short term use of either aspirin or
ibuprofen results in a 'rebound' increase in cytokine (eg., IL-1β) -induced cytokine synthesis (TNFα) in
humans (48). These results collectively support homoeostatic functions of arachidonic metabolites. The
lipidomics platform combined with protein arrays to interrogate cytokine pathways may help identify
novel, differentiated anti-inflammatory agents. Our observations provide a plausible hypothesis to
explain seemingly contradictory outcomes from the use of NSAIDs for the treatment of inflammatory
bowel disease (IBD). NSAIDs have been inconsistently implicated in cases of de novo IBD and disease
exacerbations (49). While there is conflicting data regarding NSAID use and IBD exacerbations, with
subsets of patients in which NSAIDs use worsens disease activity and others in which selective COX-2
inhibitors lower disease activity (50). The identification of those distinct patient populations may impact
optimal clinical management. Our work that shows that indomethacin dramatically increased TNFα
levels, while the selective COX-2 inhibitor, Rofecoxib did not. This may explain instances of de novo
IBD and disease exacerbation in certain patients exposed to non-selective NSAIDs. Reductions in TNFα
levels in IBD patients exposed to some COX-2 inhibitors may also explain the better tolerance of
selective COX-2 inhibitors versus NSAIDs, and the occasional improvement in disease activity (51).
Interestingly both NSAIDs and TNFα antagonists are mainstays of therapy for many inflammatory and
rheumatologic diseases. Our findings call for comprehensive clinical pharmacology studies to explore
potential unintended antagonism as a result of dual therapy, if the NSAID is not chosen carefully.
Discussion
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
14
Two compelling lessons emerge from our analysis. (1) Lipidome-focused profiling provides a unique
scope for observation of the biochemical mechanisms of action of experimental and physiological
inducers of inflammatory responses and anti-inflammatory mechanisms impinging directly or indirectly
on arachidonic pathway. This is not surprising, as the metabolome is the spoken language of proteins as
they interact with each other directly through physical contact, or indirectly through modulatory effects
mediated by biosynthetic intermediaries and second messengers. (2) Even for the best characterized
physiological agents and pharmacological tools, responses are complex. Thus, traditional approaches that
rely on target proximal or distal signaling events to assess dose effect relationships (target biology) seem
inappropriate to capture the true (poly) pharmacology of essentially all therapeutics. We need to depart
from traditional target-by-target selectivity panels common in pharmaceutical industry, in favor of more
comprehensive analytical readouts and finger prints for unbiased exploration target pharmacology. The
highly efficient platform for the profiling of drug polypharmacology described here exemplifies such a
platform.
Our studies offer a great example of applying lipidomics profiling to drug discovery. Firstly we have a
comprehensive, highly efficient eicosanoid panel that enables large-scale compound screening. Secondly,
this is the first time to employ this eicosanoid panel to differentiate clinical molecules and demonstrate
their distinct poly-pharmacology. Last but not least, this is also the first in the field to employ the
platform to reveal TNFα pharmacology and investigate the individual differences in immune responses.
We envision the broad utility of the agonistic lipidomics profiling platform described here to complex
biological problems, for example to defining lipidomics networks in health and disease,
pharmacogenomics, hypotheses generation, pathway based drug discovery and re-purposing existing
pharmacopoeia. In turn, we expect that information obtained from such profiling work will provide much
needed complementary data to enable building more efficient networks that are today dominated by
expression array data, lacking predictive value with regard to the actual production of proteins and
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
15
modulation of their specific functions. These lessons, taken as a whole, provide an illustration for how
precision medicine may become a reality, hopefully in the not-so-distant future.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
16
Acknowledgements
We thank Tatiana Koudriakova for generously sharing Janssen Discovery Sciences-La Jolla mass
spectrometry instrumentation that are essential for this work. We also thank Xiaohua Xue, Alec Lebsack
and Jim Karras from Janssen Immunology Discovery for helpful discussions and additions to the
manuscript, and the laboratory of Prof. Ed Dennis at UCSD for help in method development and
troubleshooting. We also wish to acknowledge Drs. Philip E. Needleman, Peter C. Isakson, Henry P.
Halushka, James P. Edwards and Anne M. Fourie, for their extensive review of draft versions of the
manuscript, with many insightful critiques, and Laura Pace for thoughtfully pointing out the relevance of
our findings to the identification of IBD patient subpopulations with differential responses to NSAIDs.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
17
References
1. Miller MC, 3rd, Mohrenweiser HW, Bell DA. Genetic variability in susceptibility and response to toxicants. Toxicology letters. 2001 Mar 31;120(1-3):269-80. PubMed PMID: 11323185. 2. Deloitte. The Changing Face of R&D in the Future Pharmaceutical Landscape2007. 3. Roth BL, Sheffler DJ, Kroeze WK. Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders and schizophrenia. Nature reviews Drug discovery. 2004 Apr;3(4):353-9. PubMed PMID: 15060530. 4. Hopkins AL. Drug discovery: Predicting promiscuity. Nature. 2009 Nov 12;462(7270):167-8. PubMed PMID: 19907483. 5. Yildirim MA, Goh KI, Cusick ME, Barabasi AL, Vidal M. Drug-target network. Nature biotechnology. 2007 Oct;25(10):1119-26. PubMed PMID: 17921997. 6. Boran AD, Iyengar R. Systems approaches to polypharmacology and drug discovery. Current opinion in drug discovery & development. 2010 May;13(3):297-309. PubMed PMID: 20443163. Pubmed Central PMCID: 3068535. 7. Bickerton GR, Paolini GV, Besnard J, Muresan S, Hopkins AL. Quantifying the chemical beauty of drugs. Nature chemistry. 2012 Feb;4(2):90-8. PubMed PMID: 22270643. Pubmed Central PMCID: 3524573. 8. Serra V, Eichhorn PJ, Garcia-Garcia C, Ibrahim YH, Prudkin L, Sanchez G, et al. RSK3/4 mediate resistance to PI3K pathway inhibitors in breast cancer. The Journal of clinical investigation. 2014 Mar 3;124(3):1418. PubMed PMID: 24590298. Pubmed Central PMCID: 3934174. 9. Soto A, Martin V, Jimenez E, Mader I, Rodriguez JM, Fernandez L. Lactobacilli and Bifidobacteria in Human Breast Milk: Influence of Antibiotherapy and Other Host and Clinical Factors. Journal of pediatric gastroenterology and nutrition. 2014 Feb 28. PubMed PMID: 24590211. 10. Song J, Liu X, Wu J, Meehan MJ, Blevitt JM, Dorrestein PC, et al. A highly efficient, high-throughput lipidomics platform for the quantitative detection of eicosanoids in human whole blood. Anal Biochem. Feb 15;433(2):181-8. PubMed PMID: 23103340. Epub 2012/10/30. eng. 11. Fradin A, Zirrolli JA, Maclouf J, Vausbinder L, Henson PM, Murphy RC. Platelet-activating factor and leukotriene biosynthesis in whole blood. A model for the study of transcellular arachidonate metabolism. Journal of immunology. 1989 Dec 1;143(11):3680-5. PubMed PMID: 2555415. 12. Reilly IA, Knapp HR, Fitzgerald GA. Leukotriene B4 synthesis and neutrophil chemotaxis in chronic granulocytic leukaemia. Journal of clinical pathology. 1988 Nov;41(11):1163-7. PubMed PMID: 2850300. Pubmed Central PMCID: 1141723. 13. Lundstrom SL, Yang J, Kallberg HJ, Thunberg S, Gafvelin G, Haeggstrom JZ, et al. Allergic asthmatics show divergent lipid mediator profiles from healthy controls both at baseline and following birch pollen provocation. PloS one. 2012;7(3):e33780. PubMed PMID: 22438998. Pubmed Central PMCID: 3305349. 14. Quehenberger O, Dennis EA. The human plasma lipidome. The New England journal of medicine. 2011 Nov 10;365(19):1812-23. PubMed PMID: 22070478. Pubmed Central PMCID: 3412394. 15. Dumlao DS, Buczynski MW, Norris PC, Harkewicz R, Dennis EA. High-throughput lipidomic analysis of fatty acid derived eicosanoids and N-acylethanolamines. Biochimica et biophysica acta. 2011 Nov;1811(11):724-36. PubMed PMID: 21689782. Pubmed Central PMCID: 3205334. 16. Xue L, Barrow A, Fleming VM, Hunter MG, Ogg G, Klenerman P, et al. Leukotriene E4 activates human Th2 cells for exaggerated proinflammatory cytokine production in response to prostaglandin D2. Journal of immunology. 2012 Jan 15;188(2):694-702. PubMed PMID: 22174450. Pubmed Central PMCID: 3272457.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
18
17. Laitinen LA, Laitinen A, Haahtela T, Vilkka V, Spur BW, Lee TH. Leukotriene E4 and granulocytic infiltration into asthmatic airways. Lancet. 1993 Apr 17;341(8851):989-90. PubMed PMID: 8096945. 18. Talley JJ, Brown DL, Carter JS, Graneto MJ, Koboldt CM, Masferrer JL, et al. 4-[5-Methyl-3-phenylisoxazol-4-yl]- benzenesulfonamide, valdecoxib: a potent and selective inhibitor of COX-2. Journal of medicinal chemistry. 2000 Mar 9;43(5):775-7. PubMed PMID: 10715145. 19. Daham K, Song WL, Lawson JA, Kupczyk M, Gulich A, Dahlen SE, et al. Effects of celecoxib on major prostaglandins in asthma. Clinical and experimental allergy : journal of the British Society for Allergy and Clinical Immunology. 2011 Jan;41(1):36-45. PubMed PMID: 20880055. 20. Chavez ML, DeKorte CJ. Valdecoxib: a review. Clinical therapeutics. 2003 Mar;25(3):817-51. PubMed PMID: 12852704. 21. Daham K, James A, Balgoma D, Kupczyk M, Billing B, Lindeberg A, et al. Effects of selective COX-2 inhibition on allergen-induced bronchoconstriction and airway inflammation in asthma. The Journal of allergy and clinical immunology. 2014 Jan 22. PubMed PMID: 24461582. 22. Prasit P, Wang Z, Brideau C, Chan CC, Charleson S, Cromlish W, et al. The discovery of rofecoxib, [MK 966, Vioxx, 4-(4'-methylsulfonylphenyl)-3-phenyl-2(5H)-furanone], an orally active cyclooxygenase-2-inhibitor. Bioorganic & medicinal chemistry letters. 1999 Jul 5;9(13):1773-8. PubMed PMID: 10406640. 23. Waxman HA. The lessons of Vioxx--drug safety and sales. The New England journal of medicine. 2005 Jun 23;352(25):2576-8. PubMed PMID: 15972862. 24. Vane JR, Bakhle YS, Botting RM. Cyclooxygenases 1 and 2. Annual review of pharmacology and toxicology. 1998;38:97-120. PubMed PMID: 9597150. 25. Mason RP, Walter MF, McNulty HP, Lockwood SF, Byun J, Day CA, et al. Rofecoxib increases susceptibility of human LDL and membrane lipids to oxidative damage: a mechanism of cardiotoxicity. Journal of cardiovascular pharmacology. 2006;47 Suppl 1:S7-14. PubMed PMID: 16785833. 26. Mason RP, Walter MF, Day CA, Jacob RF. A biological rationale for the cardiotoxic effects of rofecoxib: comparative analysis with other COX-2 selective agents and NSAids. Sub-cellular biochemistry. 2007;42:175-90. PubMed PMID: 17612051. 27. Walter MF, Jacob RF, Day CA, Dahlborg R, Weng Y, Mason RP. Sulfone COX-2 inhibitors increase susceptibility of human LDL and plasma to oxidative modification: comparison to sulfonamide COX-2 inhibitors and NSAIDs. Atherosclerosis. 2004 Dec;177(2):235-43. PubMed PMID: 15530895. 28. Rubin P, Dube L, Braeckman R, Swanson L, Hansen R, Albert D, et al. Pharmacokinetics, safety, and ability to diminish leukotriene synthesis by zileuton, an inhibitor of 5-lipoxygenase. Agents and actions Supplements. 1991;35:103-16. PubMed PMID: 1781415. 29. Brideau C, Chan C, Charleson S, Denis D, Evans JF, Ford- AW, et al. Pharmacology of MK-0591 (3-[1-(4-chlorobenzyl)-3-(t-butylthio)-5-(quinolin-2-yl-methoxy)- indol-2-yl]-2,2-dimethyl propanoic acid), a potent, orally active leukotriene biosynthesis inhibitor. Canadian journal of physiology and pharmacology. 1992 Jun;70(6):799-807. PubMed PMID: 1330258. 30. Gillard J, Ford-Hutchinson AW, Chan C, Charleson S, Denis D, Foster A, et al. L-663,536 (MK-886) (3-[1-(4-chlorobenzyl)-3-t-butyl-thio-5-isopropylindol-2-yl]-2,2 - dimethylpropanoic acid), a novel, orally active leukotriene biosynthesis inhibitor. Canadian journal of physiology and pharmacology. 1989 May;67(5):456-64. PubMed PMID: 2548691. 31. Jones TR, Labelle M, Belley M, Champion E, Charette L, Evans J, et al. Pharmacology of montelukast sodium (Singulair), a potent and selective leukotriene D4 receptor antagonist. Canadian journal of physiology and pharmacology. 1995 Feb;73(2):191-201. PubMed PMID: 7621356. 32. Frieri M, Therattil J, Wang SF, Huang CY, Wang YC. Montelukast inhibits interleukin-5 mRNA expression and cysteinyl leukotriene production in ragweed and mite-stimulated peripheral blood mononuclear cells from patients with asthma. Allergy and asthma proceedings : the official journal of regional and state allergy societies. 2003 Sep-Oct;24(5):359-66. PubMed PMID: 14619337.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
19
33. Cai C, Yang J, Hu S, Zhou M, Guo W. Relationship between urinary cysteinyl leukotriene E4 levels and clinical response to antileukotriene treatment in patients with asthma. Lung. 2007 Mar-Apr;185(2):105-12. PubMed PMID: 17393242. 34. Watkins PB, Dube LM, Walton-Bowen K, Cameron CM, Kasten LE. Clinical pattern of zileuton-associated liver injury: results of a 12-month study in patients with chronic asthma. Drug safety : an international journal of medical toxicology and drug experience. 2007;30(9):805-15. PubMed PMID: 17722971. 35. Bernstein JA, Liu N, Knorr BA, Smugar SS, Hanley WD, Reiss TF, et al. MK-0633, a potent 5-lipoxygenase inhibitor, in chronic obstructive pulmonary disease. Respiratory medicine. 2011 Mar;105(3):392-401. PubMed PMID: 20970976. 36. Wasfi YS, Villaran C, de Tilleghem Cle B, Smugar SS, Hanley WD, Reiss TF, et al. The efficacy and tolerability of MK-0633, a 5-lipoxygenase inhibitor, in chronic asthma. Respiratory medicine. 2012 Jan;106(1):34-46. PubMed PMID: 21945511. 37. Wu-Wang CY, Wang SL, Lim C, Milles M, Slomiany A, Slomiany BL. Cigarette smoking reduces human salivary eicosanoids. Prostaglandins, leukotrienes, and essential fatty acids. 1992 Oct;47(2):101-4. PubMed PMID: 1461919. 38. Zijlstra FJ, Vincent JE, Mol WM, Hoogsteden HC, Van Hal PT, Jongejan RC. Eicosanoid levels in bronchoalveolar lavage fluid of young female smokers and non-smokers. European journal of clinical investigation. 1992 May;22(5):301-6. PubMed PMID: 1592082. 39. Psychogios N, Hau DD, Peng J, Guo AC, Mandal R, Bouatra S, et al. The human serum metabolome. PloS one. 2011;6(2):e16957. PubMed PMID: 21359215. Pubmed Central PMCID: 3040193. 40. Noguchi K, Shitashige M, Yanai M, Morita I, Nishihara T, Murota S, et al. Prostaglandin production via induction of cyclooxygenase-2 by human gingival fibroblasts stimulated with lipopolysaccharides. Inflammation. 1996 Oct;20(5):555-68. PubMed PMID: 8894718. 41. Matsumoto H, Naraba H, Murakami M, Kudo I, Yamaki K, Ueno A, et al. Concordant induction of prostaglandin E2 synthase with cyclooxygenase-2 leads to preferred production of prostaglandin E2 over thromboxane and prostaglandin D2 in lipopolysaccharide-stimulated rat peritoneal macrophages. Biochemical and biophysical research communications. 1997 Jan 3;230(1):110-4. PubMed PMID: 9020023. 42. Hart FD, Boardman PL. Indomethacin: A New Non-Steroid Anti-Inflammatory Agent. British medical journal. 1963 Oct 19;2(5363):965-70. PubMed PMID: 14056924. Pubmed Central PMCID: 1873102. 43. Penning TD, Talley JJ, Bertenshaw SR, Carter JS, Collins PW, Docter S, et al. Synthesis and biological evaluation of the 1,5-diarylpyrazole class of cyclooxygenase-2 inhibitors: identification of 4-[5-(4-methylphenyl)-3-(trifluoromethyl)-1H-pyrazol-1-yl]benze nesulfonamide (SC-58635, celecoxib). Journal of medicinal chemistry. 1997 Apr 25;40(9):1347-65. PubMed PMID: 9135032. 44. Altavilla D, Squadrito F, Canale P, Ioculano M, Squadrito G, Campo GM, et al. G 619, a dual thromboxane synthase inhibitor and thromboxane A2 receptor antagonist, inhibits tumor necrosis factor-alpha biosynthesis. European journal of pharmacology. 1995 Nov 3;286(1):31-9. PubMed PMID: 8566149. 45. Spatafora M, Chiappara G, D'Amico D, Volpes D, Melis M, Pace E, et al. Effect of indomethacin on the kinetics of tumour necrosis factor alpha release and tumour necrosis factor alpha gene expression by human blood monocytes. Pharmacological research : the official journal of the Italian Pharmacological Society. 1991 Apr;23(3):247-57. PubMed PMID: 2068050. 46. Asari Y, Majima M, Sugimoto K, Katori M, Ohwada T. Release site of TNF alpha after intravenous and intraperitoneal injection of LPS from Escherichia coli in rats. Shock. 1996 Mar;5(3):208-12. PubMed PMID: 8696985.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
20
47. Hartel C, von Puttkamer J, Gallner F, Strunk T, Schultz C. Dose-dependent immunomodulatory effects of acetylsalicylic acid and indomethacin in human whole blood: potential role of cyclooxygenase-2 inhibition. Scandinavian journal of immunology. 2004 Oct;60(4):412-20. PubMed PMID: 15379866. 48. Endres S, Whitaker RE, Ghorbani R, Meydani SN, Dinarello CA. Oral aspirin and ibuprofen increase cytokine-induced synthesis of IL-1 beta and of tumour necrosis factor-alpha ex vivo. Immunology. 1996 Feb;87(2):264-70. PubMed PMID: 8698389. Pubmed Central PMCID: 1384283. 49. Kefalakes H, Stylianides TJ, Amanakis G, Kolios G. Exacerbation of inflammatory bowel diseases associated with the use of nonsteroidal anti-inflammatory drugs: myth or reality? European journal of clinical pharmacology. 2009 Oct;65(10):963-70. PubMed PMID: 19711064. 50. Mahadevan U, Loftus EV, Jr., Tremaine WJ, Sandborn WJ. Safety of selective cyclooxygenase-2 inhibitors in inflammatory bowel disease. The American journal of gastroenterology. 2002 Apr;97(4):910-4. PubMed PMID: 12008668. 51. El Miedany Y, Youssef S, Ahmed I, El Gaafary M. The gastrointestinal safety and effect on disease activity of etoricoxib, a selective cox-2 inhibitor in inflammatory bowel diseases. The American journal of gastroenterology. 2006 Feb;101(2):311-7. PubMed PMID: 16454836.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from
21
Figure Legends
Figure 1. Human whole blood eicosanoid signatures in response to stimulations of LPS, LTE4 and PGE2.
Heat map of averaged log2 ratios (from two donors) of 26 eicosanoids in human whole blood, in response
to stimulation of calcium ionophore (A23187; Calon), LPA, LPS, and LTE4, relative to untreated
samples. These 26 eicosanoids were induced by either one of the three stimulations (LPS, LTE4, and
PGE2), with fold change greater than 2 and false discovery rate less than 0.05, in both donors in at least
two combinations of time and concentrations. The color index in heatmap reflects the magnititude of
induction (or reduction), with a range from -4 (reflecting 16 fold reduction) to +8 (reflecting 256 fold
induction).
Figure 2. Effect of COX inhibitors, FLAPs, and Montelukast on A23187 induced eicosanoids in human
whole blood. A. Heat map of averaged log2 ratios of 33 eicosanoids in human whole blood, in response to
stimulation of calcium ionophore (A23187, 30 uM), and modulations of the stimulation by four COX
inhibitors (Celecoxib, Indomethacin, Rofecoxib, and Valdecoxib), four 5-LOX/FLAP Inhibitors (MK-
0633, MK591, MK886, and Zileuton), and Montelukast, a LTD4 receptor antagonist, at four
concentrations (0.03 – 30 uM) each. These 33 eicosanoids were induced by calcium ionophore (Calon),
with fold change greater than 1.5 and false discovery rate less than 0.05, in 23 donors collectively; log2
ratios for each compound treatment were averaged among at least 10 donors. Data for individual donor
were presented in Supplement Figure 2. B. Heat map of averaged log2 ratios of 6 eicosanoids in COX
pathway, in response to calcium ionophore stimulation and its modulation by COX inhibitors. B. Heat
map of averaged log2 ratios of 21 eicosanoids in LOX pathway, in response to calcium ionophore
stimulation and its modulation by FLAPs and Montelukast. Condition of CaIon: calcium ionophore vs.
untreated; Condition of individual compound: CaIon + Compound vs. CaIon + Vehicle.
Figure 3. Effect of COX inhibitors on human whole blood LPS eicosanoid signature. A. Heat map of
averaged log2 ratios (from two donors) of five eicosanoids in human whole blood, in response to
stimulation of calcium ionophore (A23187, 30 uM; Calon), or LPS (0.01 – 1.25 ug/mL). These five
eicosanoids were induced by LPS stimulation, with fold change greater than 2 and false discovery rate
less than 0.05, in both donors in at least two combinations of time and concentrations. B. Heat map of
log2 ratios (averaged among 4-8 donors) of the LPS eicosanoid signature, in human whole blood in
response to LPS stimulation and COX inhibitors after 4 hours. Data for individual donor were presented
in Supplement Figure 3. C and D. Box plots of log2 ratios of the TXB2 and PGE2 in individual human
whole blood samples, in response to LPS stimulation and COX inhibitors. Condition of LPS: LPS (1
ug/mL) vs. untreated; Condition of COX inhibitors: LPS + COX inhibitor vs. LPS + Vehicle.
Figure 4. Effect of LPS and COX inhibitors on human LPS eicosanoid signature in mouse. A. Heat map
of human LPS eicosanoid signature in mouse whole blood, after in vivo LPS stimulation, with or without
treatment of COX inhibitors. Log2 transformed lipid levels in mice from six treatment groups were
normalized by the median value of each lipid in all animals. B. TNFα levels in mouse whole blood
measured by ELISA. C. PGE2 levels in mouse whole blood. D. The correlation between the
concentrations of PGE2 and TNFα in whole blood from mice with LPS stimulation. G1: Saline, G2: LPS
(1 mg/Kg) + vehicle, G3: LPS (1 mg/Kg) + 1 mg/Kg Indomethacin, G4: LPS + 10 mg/Kg Indomethacin,
G5: LPS + 1 mg/Kg Rofecoxib, and G6: LPS + 10 mg/Kg Rofecoxib.
by guest, on May 29, 2018
ww
w.jlr.org
Dow
nloaded from