The Toxicologist - Society of Toxicology

146 COMPUTATIONAL MODELING FOR QT
PROLONGATION: A DRUG CARDIOVASCULAR
SAFETY ENDPOINT OF PARAMOUNT IMPORTANCE.
L. G. Valerio 1 , J. Prous Blancafort 2 , A. Valencia 2 and X. Mensa 2 . 1 CDER/OPS,
U.S. FDA, Silver Spring, MD and 2 Prous Institute for Biomedical Research,
Barcelona, Spain. Sponsor: N. Sadrieh.
Because of the potentially catastrophic nature of unanticipated “off-target” drug-related
cardiovascular effects, adequate assessment of cardiac safety is of paramount
importance. Torsades de pointes (TdP), which is a potentially lethal cardiac arrhythmia,,has
been observed with a number of drugs, after they were approved and
marketed. Prediction of such a lethal adverse event, prior to approval and marketing
of the product, would have allowed for a more thorough regulatory assessment
of clinical trial data, and potentially could have halted the clinical trials and terminated
the drug’s development. QT interval prolongation is a precursor to TdP, and
thus predicting this effect would certainly allow for better clinical development of
candidate drug products. The measurement of QTc interval effects has been
adopted into the guidance set by the FDA and international health authorities to
assess pro-arrhythmic potential for new drugs. In recognizing this public health
concern and the goal of developing better predictive methods to minimize pro-arrhythmic
liability, our group has deployed a computational modeling approach
based on new high quality human clinical trial data. The clinical data on QT prolongation
during clinical cardiovascular studies will support the development of innovative
qualitative and quantitative structure-activity models that can help predict
QT prolongation and TdP events, based on chemical structure. This present effort
is the first that we know of, where human clinical trial data will be used to construct
predictive computational toxicology models to help screen drugs for QT prolongation
and subsequently TdP, and will serve as a decision support tool for risk assessment
of new drugs.
147 TRANSLATABILITY OF DRUG INDUCED
CARDIOTOXICITY MOLECULAR MECHANISMS FROM
IN VIVO TO IN VITRO.
A. Enayetallah 1 , D. Puppala 1 , D. Ziemek 2 , N. Greene 1 , Y. Will 1 and M.
Pletcher 1 . 1 Compound Safety Prediction-PGRD, Pfizer Inc., Groton, CT and
2 Computational Sciences CoE, Pfizer Inc., Cambridge, MA.
One of the major hurdles in developing in vitro assays to assess cardiac safety of
pharmaceutical compounds is the lack of understanding of how reflective the cellbased
in vitro models are of in vivo toxicity. The goal of this study is to identify
drug-induced cardiotoxicity mechanisms that are translatable from in vivo to in
vitro. In vivo gene expression data collected from rat hearts after treatment with
known cardiotoxic compounds were obtained from the Iconix DrugMatrix database
for 10 different compounds. To generate comparable in vitro gene expression
datasets, two cell lines of cardiomyocyte origin, H9C2 and 1ry rat cardiomyocytes
were treated with the same 10 compounds from the in vivo studies at their cell toxicity
IC20 and IC50 concentrations for 24 hours before RNA was extracted for microarray
gene expression analysis. A Causal Reasoning Engine, an internally developed
platform that infers upstream molecular causes of experimental gene
expression data in the context of prior biological knowledge, was deployed to analyze
both the in vivo and in vitro datasets. Our results show that the mechanistic
profiles in the two cell lines were similar for some compounds (Dobutamine and
Ergocalciferol) but significantly different for others (Amiodarone and Amitriptylin).
Interestingly, regardless of the similarities or differences in the cell lines we identified
mechanisms that are translatable from in vivo to both cell lines (altered Rho/ROCK
signaling) and mechanisms translatable to only one cell line (endoplasmic reticulum
stress in H9C2 only). We also identified potentially measurable toxicity mechanisms
for compounds with in vivo cardiotoxicity but failed to induce any obvious
cellular injury in vitro that otherwise would have allowed for a prediction of toxicity
(Dexamethasone). In conclusion, this study enhances our ability to develop truly
predictive in vitro cardiac safety assessment assays by pursuing mechanisms that reflect
in vivo biology in the proper candidate cell-based model.
148 ASSOCIATION NETWORKS AND VISUALIZATION
TOOLS TO CORRELATE PRECLINICAL SIGNALS TO
DRUG-INDUCED NAUSEA IN MAN.
J. Glab 1 , M. Clark 2 , J. Valentin 1 and L. Ewart 1 . 1 Safety Pharmacology,
AstraZeneca, Alderley Park, United Kingdom and 2 Discovery Information,
AstraZeneca, Wilmington, DE.
The therapeutic value of many drugs can be limited by gastrointestinal (GI) adverse
effects such as nausea and vomiting. These events can limit drug absorption, cause
dose-limiting toxicity, affect patients’ compliance and result in labeling restriction
or ultimately drug discontinuation / withdrawal. Drug-induced nausea in Phase I
studies has been quantified in 113 candidate drugs and was present in ~30% (Ewart
et al, 2010). It is a multi-system reflex and a subjective human sensation and there
may not be an analogous experience in animals. Hence little is known about preclinical
biomarkers that may accurately and effectively predict drug-induced nausea
in man. The aim of this analysis was to data mine published documents from
MEDLINE and FDA Adverse Event Reporting System to identify preclinical GI
effects that may be associated with nausea and that could be of potential use in its
prediction. A total of 160 marketed drugs were found that caused nausea in man
and were also reported to have plausible GI observations in animals. A visual representation
was used to display the vast network of associations between the drugs
and observations in the pre-clinical studies. Collectively, gastrointestinal ulcers were
the most common observations in animals occurring in 40% of the drugs that
caused nausea. Within this group, observations of stomach ulcers accounted for
59%. Other common GI effects observed in animals included: gastric lesions
(17%), diarrhea (15%), changes in gastric acid secretion (14%) and emesis (12%).
This investigation demonstrates the feasibility of data-mining approaches to facilitate
discovery of novel, plausible associations that can be used to understand druginduced
adverse events. Utilizing existing public and/or proprietary data sets
negates the need for additional animal usage. The next step would be to test this approach
for its ability to predict drug-induced nausea.
Ewart L.et al(2010) 10th Annual Meeting Safety Pharmacology Society
149 DDT AND TCDD INTERACT WITH PROTEINS
INVOLVED IN THE INSULIN RECEPTOR SIGNALING.
M. Cabarcas-Montalvo, D. Montes-Grajales and J. Olivero-Verbel.
Environmental and Computational Chemistry Group, University of Cartagena,
Cartagena, Colombia.
TCDD (2,3,7,8-Tetrachlorodibenzo-p-dioxin) and DDT (1,1,1-Trichloro-2,2bis(4-chlorophenyl)ethane)
are toxic and persistent environmental pollutants that
have been proposed as etiologic factors for diabetes. However, the existence of target
proteins belonging to the insulin receptor signaling remains elusive.
Accordingly, it has been hypothesized that some proteins involved in this pathway
could be theoretical targets for DDT and its derivatives, as well as for TCDD.
Reported proteins to be part of transduction mechanisms activated by insulin were
obtained from Protein Data Bank (PDB), and prepared employing Sybyl 8.1.
Autodock Vina was utilized for docking calculations between the energy-minimized
geometries (ab initio) of organochlorine compounds and proteins, and
LigandScout was used to identify ligand-residue interactions. The proteins with the
highest binding affinities for both groups of compounds were different.
Phosphodiesterase 10A (PDB 2wey) and the eukaryotic translation initiation factor
4E (eIF4E, PDB 1wkw) presented the greatest binding affinity scores for DDT derivatives
(8.3-8.6 Kcal/mol), whereas in the case of TCDD, best values were computed
for glycogen synthase kinase 3 beta (PDB 1uv5, -8.10 Kcal/mol) and glucocorticoid-regulated
kinase 1 (PDB 3hdn; -8.10 Kcal/mol). These theoretical target
proteins for DDT or TCDD are involved in different actions regarding insulin receptor
activation, such lipolysis, protein synthesis, apoptosis, glycogen and fatty
acids synthesis, and sodium transport, among others. Most protein-ligand complexes
included hydrophobic and aromatic interactions, and in the case of eIF4E,
the binding pocket is shared by several DDT-related compounds. In short, in silico
studies have shown that DDT and TCDD have the theorethical ability to bind proteins
involved in insulin signaling, and this could eventually influence the development
of diabetes. Colciencias-Universidad de Cartagena (Grant 110745921616).
150 COMPUTATIONAL BASED APPROACHES FOR
IDENTIFYING AND UNDERSTANDING KINASES
ASSOCIATED WITH CARDIOTOXICITY.
D. Puppala 1 , V. Bonato 2 , K. Leach 1 , S. Louise-May 1 , K. J. McConnell 3 , M.
Gosink 4 and M. T. Pletcher 1 . 1 Compound Safety Prediction, Pfizer Inc., Groton, CT,
2 Biostatistics Group, Pfizer Inc., Groton, CT, 3 Computational Science CoE, Pfizer
Inc., Groton, CT and 4 DSRD, Pfizer Inc., Groton, CT.
Kinase biology is a complex network which offers promising drugs targets but also
has been associated with significant adverse events like cardiotoxicity. The goal of
this project was to identify kinases that have the greatest potential to induce cardiotoxicity
if targeted and to understand the mechanisms by which they would induce
that toxicity so that these adverse events could be predicted and avoided.
Towards this end we first used a combination of different bioinformatics approaches;
mining the mouse knockout, OMIM, and Toxreporter databases and employing
an automated literature search tool. We then collated these results in order
SOT 2011 ANNUAL MEETING 31