The Toxicologist - Society of Toxicology

934 THE FDA’S LIVER TOXICITY KNOWLEDGE BASE
(LTKB) PROJECT.
W. Tong 1 , M. Chen 1 , V. Vikrant 1 , Q. Shi 1 , M. Zhang 1 , L. Guo 2 , Z. Liu 1 , J.
Zhang 1 and E. Bearden 1 . 1 Division of Systems Biology, U.S. FDA’s NCTR, Jefferson,
AR and 2 Division of Biochemical Toxicology, U.S. FDA’s NCTR, Jefferson, AR.
A significant amount of drugs to be withdrawn from the market and to fail during
clinical trial stage of development is due to liver toxicity. These cases of drug-induced
liver injury (DILI) are reported as the leading cause of acute liver failure in
the U.S. Thus, DILI has become one of the most important concerns in the drug
development and approval process. DILI has also been identified by the FDA
Critical Path Initiatives as a key area of focus in a concerted effort to broaden the
agency’s knowledge for better evaluation tools and safety biomarkers. In order to
understand liver toxicity at the mechanistic level and develop novel tools for identifying
liver toxicity issues along the various stages of drug development, there is a
need to develop content-rich resources to improve our basic understanding of liver
toxicity and facilitate the efficient development of novel knowledge and tools for
utilization by research, industry and regulatory groups. The FDA’s National Center
for Toxicological Research is developing a liver toxicity knowledge base (LTKB).
The project is drug-specific with aim to identify potential risk factors distinguishing
drugs in DILI. Thus, the project involves a collection of diverse data (e.g., DILI
mechanisms, drug metabolism, histopathology, therapeutic use, targets, side effects,
etc) associated with individual drugs and systems biology analysis to integrate these
data for DILI assessment and prediction. Importantly, both conventional and highthroughput
molecular biomarker assays will be conducted for selected drugs to develop
novel biomarkers based on knowledge accumulated from the project. The
project will provide a wealth of focused knowledge and data mining tools for hepatotoxicity
in the form of networks between chemicals (and drugs), molecular signatures,
liver specific biomarkers, genes/proteins functions, pathways and liver diseases
(and pathology). The preliminary results and early progress of LTKB will be
discussed.
935 VALIDATION OF CELLCIPHR CELLULAR SYSTEMS
BIOLOGY (CSB) TECHNOLOGY. CELLCIPHR2 -
BUILDING ON EXPERIENCE OF IN VIVO RAT LIVER
INJURY PREDICTION TO DEVELOP THE HUMAN
DILI PREDICTION.
S. Thomas 2 , J. Mein 2 , R. Annand 1 , P. Walker 2 , M. Jacewicz 1 , D. Steen 3 , C.
Chesne 3 , J. Gilbert 1, 2 and K. Tsaioun 1, 2 . 1 Apredica, Watertown, MA, 2 Cyprotex,
Macclesfield, United Kingdom and 3 Biopredic, Rennes, France.
Drug-induced liver injury (DILI) is the most frequent reason for the withdrawal of
an approved drug from the market, and it also accounts for up to 50% of cases of
acute liver failure. Nominating pre-clinical candidates is one of the most important
activities in drug development, that clears the path for a compound to be tested in
human clinical trials. Hence, it is of crucial importance for drug development
teams to reduce the failures in preclinical toxicity studies. CellCiphr Cellular
Systems Biology (CSB) technology is an in vitro discovery toxicology platform that
has been validated by top large pharmaceutical companies and EPA, FDA, and
NIH for identifying the risks of rat liver toxicity in new chemical entities. The most
common uses of this technology currently are for lead series selection, lead optimization,
and animal toxicity studies optimization. CellCiphr is built on highcontent
imaging and currently only predicts rat liver injury. As part of building of
the human DILI prediction model, a sub-set of drugs with documented human
DILI or lack thereof has been assessed in a new model in a number of rat and
human transformed and primary liver models including HepG2, HepaRG and primary
cells. Comparison of these models and their advantages and limitations in
building classifiers for human prediction models will be presented for the first time.
The utility of rat liver classifiers for prediction of human DILI will be assessed and
new classifiers presented. CellCiphr technology is a mission-critical tool in drug development,
and addition of human DILI model to already validated rat model is a
significant advanced new tool allowing drug discovery teams to save significant
funds by reducing the late attrition.
936 INVESTIGATING THE ROLE OF MITOCHONDRIAL
TOXICITY IN DRUG-INDUCED LIVER INJURY VIA
CONGRUENT STRUCTURE-ACTIVITY
RELATIONSHIPS.
M. L. Patel 1 , L. Fisk 1 , N. Greene 2 and R. T. Naven 2 . 1 Lhasa Limited, Leeds,
United Kingdom and 2 Worldwide Medicinal Chemistry, Pfizer Inc., Groton, CT.
Drug-induced mitochondrial dysfunction may represent a key mechanistic pathway
in the development of organ-related toxicities, including those affecting the liver.
The silent accumulation of damage to mitochondria over time has been proposed as
an explanation for observed idiosyncratic toxicity. As part of ongoing work to de-
velop structure-activity relationships (SARs) to predict the toxicity of chemicals, we
investigated the suitability of using existing hepatotoxicity structural alerts, as indicators
of mitochondrial dysfunction. We report the evaluation of a hepatotoxicity
knowledge base, using three published data sets for mitochondrial dysfunction.
Two data sets, comprising of 105 and 282 compounds, respectively, were constructed
based on publications reporting in vivo results for mitochondrial toxicity.
The third data set of 2490 compounds was based on the results of in vitro assays.
The three data sets were processed against a knowledge base of 62 SARs for hepatotoxicity.
Predictions were classed as correct when a hepatotoxicity alert was activated
for a compound reported to cause mitochondrial dysfunction, or, in the absence
of any alerts for a compound, reported to be negative. The results showed an
overall concordance of 55% to 65% across the three data sets, with sensitivity ranging
from 28% to 62% and specificity from 67% to 72%.
We have previously developed alerts for hepatotoxicity based on published in vivo
data. A subset of these alerts performed well against the data sets reporting in vivo
evaluations, indicating that for those chemical classes, the mechanisms of toxicity
may act through the mitochondria. This was in contrast to the performance against
the data set reporting in vitro results. Reasons for this could include the biased coverage
of chemical space of the hepatotoxicity alerts, and the extrapolation of alerts
based on in vivo data to predict in an in vitro scenario.
937 QUANTITATIVE SIMULATION OF INTRACELLULAR
SIGNALING CASCADES IN A VIRTUAL LIVER:
ESTIMATING DOSE DEPENDENT CHANGES IN
HEPATOCELLULAR PROLIFERATION AND
APOPTOSIS.
J. Jack 1 , M. Mennecozzi 2 , C. Haugh 1 , J. Wambaugh 1 and I. Shah 1 . 1 National
Center for Computational Toxicology, U.S. EPA, Research Triangle Park, NC and
2 Joint Research Centre, European Commission, Ispra, Italy.
The US EPA Virtual Liver (v-Liver) is developing an approach to predict dosedependent
hepatotoxicity as an in vivo tissue level response using in vitro data. The
v-Liver accomplishes this using an in silico agent-based systems model that dynamically
integrates environmental exposure, microdosimetry, and individual cellular
responses to simulate the in vivo hepatic lobule. One requirement for the v-Liver
models is inclusion of nuclear receptor (NR) pathways critical to nongenotoxic hepatocarcinogenesis.
Progress to date has focused on signal transduction networks
controlling cell proliferation and apoptosis, which are key events in hepatocarcinogenesis.
First, a literature derived knowledgebase was used to reconstruct a biochemical
signaling network of cell cycle initiation and caspase-mediated apoptotic
signaling events. The network contains 46 proteins and 77 interactions, encompassing
pathways relevant to hepatocytes including the epidermal growth factor
(EGF) and tumor necrosis factor alpha (TNFα). Second, static analysis of the
topology of the crosstalk network revealed key signaling proteins, reducing the
overall complexity of the model. Third, the mechanistic model was translated into
an asynchronous Boolean network ensemble to simulate the quantitative dose-response
of hepatocyte populations to various ligands: growth factors, cytokines, and
mitogenic inhibitors. The model predictions are consistent with the synergistic effects
of EGF and TNFα on early cell proliferation reported in the literature. Finally,
the model was calibrated with ToxCast data on environmental compounds, revealing
the potential impact of chemical perturbations on normal hepatocyte behavior.
This approach of static and dynamic pathway analysis and evaluation using
in vitro data provides a foundation for estimating tissue level effects of chemical exposures
from the simulation of individual cellular responses. This abstract does not
necessarily reflect US EPA policy.
938 PHYSIOLOGICALLY BASED PHARMACOKINETIC
(PBPK) MODELING TO PREDICT PLASMA
PHARMACOKINETICS FOR LIVER TRANSPORTER
SUBSTRATES.
H. A. Barton 1 , H. Jones 2 and Y. Lai 1 . 1 Pharmacokinetics, Pharmacodynamics, and
Metabolism, Pfizer, Inc., Groton, CT and 2 Pharmacokinetics, Pharmacodynamics,
and Metabolism, Pfizer, Inc., Sandwich, United Kingdom.
Predictions of plasma pharmacokinetics for substrates of liver uptake and biliary excretion
transporters can be made using PBPK models. We have implemented a
PBPK model for several drugs including fluvastatin, valsartan, rosuvastatin, and
others. The well stirred liver is replaced with liver blood and tissue compartments
connected by bidirectional passive diffusion and active uptake transport, while active
efflux from liver to blood is assumed not to occur. Liver disposition includes
nonspecific binding (partitioning), metabolism, and active biliary transport. Active
transport processes, assumed to be occurring at concentrations below half maximal
activity, are described by first order processes parameterized as clearances of free
SOT 2011 ANNUAL MEETING 199