Computational experiments reveal plausible mechanisms for changing patterns of hepatic zonation of xenobiotic clearance and hepatotoxicity

https://doi.org/10.1016/j.jtbi.2010.06.011Get rights and content

Abstract

No concrete, causal, mechanistic theory is available to explain how different hepatic zonation patterns of P450 isozyme levels and hepatotoxicity emerge following dosing with different compounds. We used the synthetic method of modeling and simulation to discover, explore, and experimentally challenge concrete mechanisms that show how and why biomimetic zonation patterns can emerge and change within agent-based analogues, expecting that those mechanisms may have counterparts in rats. Mobile objects map to compounds. One analogue represents a cross-section through a lobule. It is comprised of 460 identical, quasi-autonomous functional units called sinusoidal segments (SSs). SSs detect and respond to compound-generated response signals and the local level of an endogenous gradient. Each SS adapts by using those signals to adjust (or not) the probability that it will clear a detected compound during the next simulation cycle. The adjustment decision is based on the value of a biomimetic algorithm that is based on an assumed, evolution imposed, genetic mandate that normal hepatocytes resist increasing the cost of their actions. The algorithm estimates the long-term, discounted cost to a given SS of continuing to use its current clearance effort. Upon compound exposure, lobular analogues developed a variety of clearance and hepatotoxicity patterns that were strikingly similar to those reported in the literature. A degree of quantitative validation was achieved against data on hepatic zonation of CYP1A2 mRNA expression caused by three different doses of TCDD (2,3,7,8-tetracholorodibenzo-p-dioxone).

Introduction

Hepatic zonation is a conspicuous periportal (afferent) to perivenous (efferent) gradient of an attribute within lobules. Zonal differences occur in the clearance of a variety of endogenous compounds and xenobiotics (Gebhardt, 1992, Ierapetritou et al., 2009 1995, Jungermann and Kietzmann, 2000). Zonation is also evident for a number of normal hepatic functions, absent xenobiotic or toxin exposure. There is also differential sensitivity to the induction of cytochrome P450 isozymes (Oinonen and Lindros, 1998). Toxin caused hepatic injury can also exhibit zonal patterns. Such phenomena are most often ascribed to having a multifactorial basis, in which oxygen gradients, other blood-borne signals, and blood flow itself may play prominent roles (Camp and Capitano, 2007, Christoffels et al., 1999, Jungermann and Kietzmann, 2000, Liu et al., 2000). Recent evidence supports the hypothesis that components of the Wnt-β-catenin pathway may play an important role (Benhamouche et al., 2006, Hailfinger et al., 2006, Sheikh-Bahaei et al., 2009, Burke et al., 2009). Braeuning (2009) reviews the role of several pathways including Ras–Raf–MAPK (mitogen-activated, protein kinase) and Wnt-β-catenin. However, no concrete, causal, mechanistic theory has yet been offered for how different types of hepatic zonation phenomena emerge following dosing with different compounds. For this study, we focused on zonation patterns of P450 isozymes and the hepatic damage that can develop following treatment of rats with xenobiotics. We used the synthetic method of modeling and simulation (Hunt et al., 2009) to discover, explore, and experimentally challenge concrete mechanisms that show how and why biomimetic zonation patterns can emerge and change within an agent-based analogue of a hepatic lobule in response to compound dosing. The in silico mechanism may have counterparts in rats.

Christoffels et al. (1999) demonstrated the plausibility of a molecular level mechanism for periportal-to-perivenous gradients of gene expression. Expanding upon the zonation ideas offered in (Gebhardt, 1992), they hypothesized that interaction between two or more, different signal gradients is necessary to enable development of periportal-to-perivenous gene expression patterns that mimic those gradients and are stable under different conditions. They provided support for the hypothesis using both an inductive, mathematical model and a transgenic mouse model into which hepatocyte-specific DNA-response units had been integrated. They stated that, “if a hypothesis like the one presented … is to be tested, we need simple and strictly defined model systems.” They discuss the formidable issues of constructing such model systems using transgenic mice.

Ohno et al. (2008) constructed sophisticated, single-hepatocyte based lobular models that focus on ammonia metabolism with the long-range objective of elucidating how molecular and cellular level properties modify higher-level phenomena. Xenobiotic metabolism and enzyme induction mechanisms were not a focus. They posit that heterogeneous gene expression evolved to optimize energy efficiency. They specify histological structure and zone-specific gene expression of major enzymes, and include the biochemical kinetics of enzymes and transporters. Ierapetritou et al. (2009) recently reviewed the models of Ohno et al. along with a variety of additional computational liver models. Several specified features of zonation but not how those features may emerge. All but one of the models reviewed was an inductive mathematical model. The computational modeling and simulation (M&S) approach used herein (Fig. 1a) and the resulting models are fundamentally different from those inductive mathematical models and so are not directly comparable. Hunt et al. (2009) explain those differences and how the two different M&S approaches complement each other. The approach used herein was developed specifically to enable construction of biomimetic mechanisms that are real (not conceptual) and strictly defined. They are designed for use under conditions that are less supportive of inductive modeling methods (Fig. 1b). Even though abstract, the mechanisms and their spatial context are flexible and sufficiently concrete to instantiate mechanistic hypotheses and test their plausibility experimentally.

Following cycles of model construction, evaluation and selection, and refinement, we arrived at a hybrid discrete event, discrete time system that maps to a cross-section through a hepatic lobule having periportal-to-perivenous (P-to-P) flow and a connection to extrahepatic tissue. The hepatic lobule component is comprised of 460 identical, quasi-autonomous functional units called sinusoidal segments (SSs). Each SS maps to a small portion of a sinusoid (Fig. 2). During a simulation cycle, each SS contributes to establishment of an endogenous gradient signal, and it has an opportunity to clear a detected compound administered at the start of the simulation. There is a cost associated with each clearance event. There is also a cost associated with responding to the compound. We consider response signals generated in one of three ways. (1) A compound that exits the lobule enters an extrahepatic tissue space and causes release of a response signal, proportional to potency. (2) Compound causes response signal release within SSs. (3) Or, the compound itself functions as response signal. Herein we focus more on (1). Each simulation cycle, each SS adapts to clearance events and any changes in the two signal types by adjusting (or not) the probability that it will clear a detected compound during the next simulation cycle. That decision is based on an assumed, evolution imposed, genetic mandate to resist increasing the cost of its actions. It can try to reduce future costs in one of the two ways. If few or no response signals were detected, reduce compound clearance during the next simulation cycle. If response signals were detected, reduce the cost of their subsequent detection by increasing compound clearance during the next simulation cycle. Each SS makes use of a biomimetic algorithm to estimate long-term, discounted cost of continuing to use its current clearance effort. The outcome determines which action is selected.

Upon compound exposure, the simulated lobule developed a variety of compound-specific gradients in P-to-P clearance effort. Several gradient patterns were strikingly similar to those reported in the literature for P450 isozymes following xenobiotic dosing (for convenience, examples are provided in Supplementary Fig. S5). Zonal patterns of clearance effort and SS damage changed depending on compound dose and potency. We called the system a zonally responsive lobular analogue (ZoRLA). A ZoRLA was used to achieve a degree of quantitative validation against data on hepatic zonation of CYP1A2 mRNA expression caused by three different doses of TCDD (2,3,7,8-tetracholorodibenzo-p-dioxone).

Section snippets

Methods

To distinguish clearly in silico components and processes from corresponding rat counterparts, we use small caps when referring to the former. Parameter names are italicized.

Results from preZoRLA1

Each SS in preZoRLA1 used the mechanism in Fig. 2c to form and respond to gradients of r- and b-signals (targeted attribute 3). The mechanism also produced and removed proteins. Clearance effort (p) for each SS was proportional to its current number of proteins. The mechanism was capable of forming a variety of P-to-P b-signal and clearance effort gradient patterns that were both dose and potency dependent. Examples are presented in Fig. 4. Various parameterizations (Table 1) enabled achieving

Plausible mappings of SS mechanisms to hepatic counterparts

Braeuning recently presented arguments and supporting evidence for the Ras–Raf–MAPK and Wnt-β-catenin signaling pathways playing roles in both hepatocyte zonation as well as induction of P450 isozymes (2009). There are clear consistencies between the ideas presented in that paper and the more abstract preZoRLA1 mechanisms, and they are identified in Supplementary Material.

Braeuning also presents evidence of an overall deactivating or repressive effect of Ras–Raf–MAPK (mitogen-activated, protein

Acknowledgements

We thank members of the UCSF BioSystems Group (Sean Kim, Jon Tam, Teddy Lam, and Jesse Engelberg), Glen Ropella, Yoav Shoham, and Zbigniew Kmieç for helpful suggestions and discussions. The work was supported in part by a Computational and Systems Biology Fellowship (SSB) provided by the CDH Research Foundation (CDHRP-08-0044). The Foundation played no role in the design or conduct of the research.

References (43)

  • C.J. Broccardo et al.

    Single cell analysis of switch-like induction of CYP1A1 in liver cell lines

    Toxicol. Sci.

    (2004)
  • R. Buhler et al.

    Zonation of cytochrome P 450 isozyme expression and induction in rat liver

    Eur. J. Biochem.

    (1992)
  • Z.D. Burke et al.

    Liver zonation occurs through a β-Catenin-dependent, c-Myc-independent mechanism

    Gastroenterology

    (2009)
  • J.P. Camp et al.

    Induction of zone-like liver function gradients in HepG2 cells by varying culture medium height

    Biotechnol. Prog.

    (2007)
  • V.M. Christoffels et al.

    A mechanistic model for the development and maintenance of portocentral gradients in gene expression in the liver

    Hepatology

    (1999)
  • L. Darden

    Strategies for discovering mechanisms: schema instantiation, modular subassembly, forward/backward chaining

    Philos. Sci.

    (2002)
  • J. Fisher et al.

    Executable cell biology

    Nat. Biotechnol.

    (2007)
  • V. Grimm et al.

    Pattern-oriented modeling of agent-based complex systems: lessons from ecology

    Science

    (2005)
  • S. Hailfinger et al.

    Zonal gene expression in murine liver: lessons from tumors

    Hepatology

    (2006)
  • C.A. Hunt et al.

    At the biological modeling and simulation frontier

    Pharm. Res.

    (2009)
  • C.A. Hunt et al.

    Physiologically based synthetic models of hepatic disposition

    J. Pharmacokinet. Pharmacodyn.

    (2006)
  • Cited by (0)

    View full text