A Liver-Centric Multiscale Modeling Framework for Xenobiotics

James P. Sluka, Xiao Fu, Maciej Swat, Julio M. Belmonte, Alin Cosmanescu, Sherry G. Clendenon, John F. Wambaugh, James A. Glazier, Edward E Schmidt
2016 PLoS ONE  
We describe a multi-scale, liver-centric in silico modeling framework for acetaminophen pharmacology and metabolism. We focus on a computational model to characterize whole body uptake and clearance, liver transport and phase I and phase II metabolism. We do this by incorporating sub-models that span three scales; Physiologically Based Pharmacokinetic (PBPK) modeling of acetaminophen uptake and distribution at the whole body level, cell and blood flow modeling at the tissue/organ level and
more » ... olism at the sub-cellular level. We have used standard modeling modalities at each of the three scales. In particular, we have used the Systems Biology Markup Language (SBML) to create both the wholebody and sub-cellular scales. Our modeling approach allows us to run the individual submodels separately and allows us to easily exchange models at a particular scale without the need to extensively rework the sub-models at other scales. In addition, the use of SBML greatly facilitates the inclusion of biological annotations directly in the model code. The model was calibrated using human in vivo data for acetaminophen and its sulfate and glucuronate metabolites. We then carried out extensive parameter sensitivity studies including the pairwise interaction of parameters. We also simulated population variation of exposure and sensitivity to acetaminophen. Our modeling framework can be extended to the prediction of liver toxicity following acetaminophen overdose, or used as a general purpose pharmacokinetic model for xenobiotics. expression levels of transporter proteins and different metabolic enzymes profiles resulting in regional differences in metabolism across the lobule and sinusoid [8] . In addition, the highly interconnected network of sinusoids results in complex blood flow within each lobule [9] . APAP is extensively metabolized in the liver via both Phase I and Phase II pathways [10]. Phase II pathways convert APAP into sulfate and glucuronide conjugates. Phase I metabolism, primarily via Cytochrome P450s (CYP) 2E1 and 1A2, generates N-acetyl-p-benzoquinone imine (NAPQI), an alkylating agent capable of covalently modifying a number of cellular components. The majority of NAPQI reacts with cellular glutathione (GSH) giving a thiol adduct NAPQI-GSH. In cases of APAP overdose, depletion of cellular GSH is thought to be causative in hepatocyte death (necrosis). Acute overdose of APAP results in extensive necrosis of hepatocytes in the centrilobular regions of the liver [11] . Although it is not fully understood how the depletion of GSH progresses to necrosis, it's believed that reactive oxygen species play a significant role [12] . The combination of Phase I and II metabolism is a common feature of xenobiotic metabolism and applies, to varying degrees, to most xenobiotics. Computational models of liver function and toxicity: Modeling a toxicological or pharmacological event requires the integration of processes that spread across different spatial and temporal scales as shown in Fig 1. At the whole body level, compounds are absorbed from the GI tract, distributed among the tissues and organs, and filtered and excreted by the kidneys. The compound diffuses within the blood, and partitions between the serum, serum proteins (such as albumin) and blood cells. At the liver tissue level, compounds enter the liver via the hepatic artery and portal vein, flow through the network of sinusoids within the hepatic lobules, and empties into the hepatic central vein. The compound transports, generally via both passive (diffusive) and active transport pathways, into and out of the hepatocytes lining the sinusoids. At the sub-cellular level, metabolic pathways within the hepatocytes convert the compound into metabolites. The metabolites are transported (either actively or passively) back into the blood circulation or into the bile pathway. Hepatocytes may proliferate, migrate or die, in response to the compound or its metabolites. Gene expression patterns orchestrate the spatial distribution of the metabolic processes within the lobule. To computationally model the multiple scales as well as their interactions we must integrate temporally and spatially across the relevant tissues. Many studies have used whole-body PBPK models to describe APAP ADME [13] [14] [15] [16] [17] [18] [19] . These models are able to reproduce the ADME data for APAP and its metabolites and are validate against in vivo data. Some of these models elaborate the description of APAP metabolism and toxicity through integration of a more detailed subcellular genome-scale flux balance metabolism model [14] or kinetic model of GSH depletion [15, 18] . Other studies focused on systems biology models of APAP metabolism and GSH homeostasis, such as [20] . developed an extended PBPK model of APAP metabolism in humans. Their model includes transport and metabolism of APAP, to the glucuronide and sulfate metabolites, as well as GSH metabolism. This model was calibrated using in vitro human data for APAP and the two major metabolites and models the toxicity of APAP as well as the effectiveness of clinical interventions for APAP overdoses. One important aspect missing from these models is the spatial architecture of the metabolizing organ and the spatial distribution of metabolic enzymes such as CYP450 2E1 at different lobular locations. A model of APAP metabolism that treats the liver as a "well stirred" compartment may miss or underestimate the heterogeneity of the liver's responses. With advancement of imaging techniques, recent modeling studies increasingly incorporate organ-specific structural and spatial features, in addition to biochemical properties of endogenous or exogenous compounds, into their representation of the liver and in understanding metabolism and toxicity [21] [22] [23] . Perhaps the most complete attempt at reconstructing the The sensitivities are from Fig 13. Values below the diagonal are correlations coefficients and above the diagonal R 2 's. The sub-table labeled "All Sensitivities" is the correlation for comparisons between the parameter sensitivities across all parameters and all model outputs (comparisons of 38x15 matrices). The sub-table labeled "Average Sensitivities" is the correlation for comparisons between the parameter sensitivities across all parameter sets and the model output"Average" rows (comparisons of 1x15 vectors). The "w/o RMSEs. . ." sub-table is similar to the "All" table but excludes the RMSEs (APAP ADME specific model outputs) and the "Average" rows from the sensitivity matrices (comparisons of 38x10 matrices).
doi:10.1371/journal.pone.0162428 pmid:27636091 pmcid:PMC5026379 fatcat:uow7debu5bdo5b7dkukqzwxmcm