Elsevier

Science of The Total Environment

Volume 615, 15 February 2018, Pages 150-160
Science of The Total Environment

High-throughput in-silico prediction of ionization equilibria for pharmacokinetic modeling

https://doi.org/10.1016/j.scitotenv.2017.09.033Get rights and content

Highlights

  • We have replaced a proprietary human variability model with an open-source EPA tool for high throughput risk prioritization.

  • Introduced the ionizable atom type (IAT), a high-throughput method for assessing the effects of ionization on compound PK.

  • Identified broad differences in the ionization of chemicals intended for pharmaceutical use, near-, and far-field sources.

  • pKa was estimated for 8132 pharmaceuticals and 24,281 other compounds to which humans might be exposed in the environment.

  • Explored the pKa prediction uncertainty for 22 NHANES chemicals using IATs and how errors in predictions impact PK models.

Abstract

Chemical ionization plays an important role in many aspects of pharmacokinetic (PK) processes such as protein binding, tissue partitioning, and apparent volume of distribution at steady state (Vdss). Here, estimates of ionization equilibrium constants (i.e., pKa) were analyzed for 8132 pharmaceuticals and 24,281 other compounds to which humans might be exposed in the environment. Results revealed broad differences in the ionization of pharmaceutical chemicals and chemicals with either near-field (in the home) or far-field sources. The utility of these high-throughput ionization predictions was evaluated via a case-study of predicted PK Vdss for 22 compounds monitored in the blood and serum of the U.S. population by the U.S. Centers for Disease Control and Prevention National Health and Nutrition Examination Survey (NHANES). The chemical distribution ratio between water and tissue was estimated using predicted ionization states characterized by pKa. Probability distributions corresponding to ionizable atom types (IATs) were then used to analyze the sensitivity of predicted Vdss on predicted pKa using Monte Carlo methods. 8 of the 22 compounds were predicted to be ionizable. For 5 of the 8 the predictions based upon ionization are significantly different from what would be predicted for a neutral compound. For all but one (foramsulfuron), the probability distribution of predicted Vdss generated by IAT sensitivity analysis spans both the neutral prediction and the prediction using ionization. As new data sets of chemical-specific information on metabolism and excretion for hundreds of chemicals are being made available (e.g., Wetmore et al., 2015), high-throughput methods for calculating Vdss and tissue-specific PK distribution coefficients will allow the rapid construction of PK models to provide context for both biomonitoring data and high-throughput toxicity screening studies such as Tox21 and ToxCast.

Introduction

Regulatory agencies worldwide are tasked with characterizing the safety of tens of thousands of commercial chemicals, yet only a small subset have been fully characterized with respect to hazard and exposure (Egeghy et al., 2012, Judson et al., 2009, USGAO, 2009, USGAO, 2013). As thousands of new chemicals are introduced into commerce each year (Judson et al., 2009, USGAO, 2009, USGAO, 2013, Wilson and Schwarzman, 2009), it becomes much more challenging to set research priorities for determining what risk, if any, these chemicals in our environment pose to human and ecological populations (Thomas et al., 2013).

High throughput, in vitro testing programs such as Tox21 (Tice et al., 2013) and ToxCast (Kavlock et al., 2012) have been screening thousands of chemicals for potential bioactivity. However, interpretation of these data relies on nominal tested concentration unless the results can be extrapolated to in vivo conditions (e.g., Wetmore et al., 2015). The Centers for Disease Control and Prevention (CDC) National Health and Nutrition Examination Survey (NHANES) includes measurements of hundreds of xenobiotic chemical concentrations in blood and serum in the U.S. population (CDC, 2012). But, without knowing how these chemicals distribute within the body, blood concentrations cannot be related to potential concentrations in tissues that might be targets of toxic effects. Further, without knowing tissue distribution, neither the total body burden of the chemical nor the rate of exposure can be estimated.

Tissue distribution of chemicals remains an important aspect of pharmacokinetics (PK) that is not rapidly measured using in vitro or in vivo techniques. Tissue PK methodologies exist in the PK literature for the prediction of chemical distribution into specific tissues or the whole body (e.g., volume of distribution at steady-state or Vdss) but require specific information on physico-chemical behavior. In silico prediction of such chemical tissue distribution is heavily influenced by three key parameters: binding to tissue and plasma, hydrophobicity, and ionization (Peyret et al., 2010, Schmitt, 2008). Hydrophobicity (quantified by the octanol-water partition coefficient, logP) drives distribution of neutral compounds; however, a neutral compound at one pH can become ionized, for example, at a physiological pH. Thus, chemical ionization is key in estimating distribution (illustrated in Fig. 1). For predicting tissue distribution, tissues can be broadly described as consisting of components with differing affinities for chemicals depending on the charged state of the organic chemical molecule, as shown in Fig. 1 (Peyret et al., 2010, Schmitt, 2008). The resulting ratio between the total concentration (ionized and un-ionized) of chemical in the tissue and the plasma is the distribution coefficient (logD) (Manners et al., 1988). In PK, logD is described through tissue-specific partition coefficients (PC) (Peyret et al., 2010, Schmitt, 2008).

At a given pH, some atoms of a compound can donate (dissociation) to or receive (association) protons from one or more atoms or sites within the compound (Fig. 2). Chemical association/dissociation changes the overall molecular charge, with the potential for coexistence of multiple microspecies (i.e., different charge states of the same parent molecule). The chemical association/dissociation equilibrium constant (pKa) characterizes the pH at which concentrations of protonated or deprotonated chemical microspecies associated with an ionizable atom or site are in equilibrium. The aim of the present work was to generate ionization profiles of chemicals at an atomic level using a rapid approach suitable for thousands of chemicals.

pKa is often reported in scientific literature as a single numerical value, sometimes categorized as “acid” or “base”. This is sufficient for a compound that undergoes a single ionization, but in many cases there are multiple ionizations, and each pKa needs to be characterized in the range of 0 < pH < 14, as shown in Fig. 2. This information is vital for PK because the overall charge and the fraction extant at a certain pH follows the Henderson-Hasselbalch equation (Hasselbalch, 1916, Henderson, 1908), which has a different behavior for acidic (negative to neutral) and basic (neutral to positive) events as pH is increased. Therefore, ionization cannot be characterized by a scalar pKa value only, nor is it possible to compare predictions of quantitative structure–activity relationship models without further characterizing the ionization kinetics.

Understanding chemical-specific ionization properties is critical for predicting tissue distribution. As new data sets of chemical-specific information on metabolism and excretion for hundreds of chemicals are being made available (e.g., Wetmore et al., 2015), high-throughput methods for calculating Vdss and tissue-specific PK distribution coefficients will allow the rapid construction of compartmental and physiologically-based PK models. PK distribution describes how chemicals can accumulate preferentially in certain tissues, producing higher concentrations in that tissue, as characterized by tissue-specific PC. Much PK literature has been devoted to prediction of tissue PCs (Haddad et al., 2000, Peyret et al., 2010, Poulin and Krishnan, 1996a, Poulin and Krishnan, 1996b, Poulin and Theil, 2000, Rodgers et al., 2005, Rodgers and Rowland, 2006, Schmitt, 2008). These models provide the context for use and interpretation of both biomonitoring data (e.g., NHANES) and high-throughput toxicity screening studies (e.g., Tox21 and ToxCast). When appropriate PC and metabolism/physiological information are used, dynamic simulation of physiologically based pharmacokinetic (PBPK) models allows prediction of chemical concentrations in specific tissues at different times (Caldwell et al., 2012, Mumtaz et al., 2012, Pearce et al., 2017, Yoon et al., 2012).

In this study, estimates of pKa were generated for 32,413 compounds to which humans might be exposed. This included 8,132 pharmaceuticals and 24,281 pesticidal, industrial and consumer use compounds. A high-throughput method for assessing the effects of ionization on compound PK, the ionizable atom type (IAT), was used. IATs are specific configurations of atoms within a chemical that has the propensity to protonate or deprotonate. Using IATs, a probability distribution of pKa values and therefore the probability of an atom to become ionized were estimated for 13 IATs based on predictions for all 32,413 chemicals. Broad differences were identified in the ionization of chemicals intended for pharmaceutical use and chemicals with both near-field (in the home) and far-field sources. The utility of these high-throughput ionization predictions was evaluated by assessing the impact of chemical ionization on predicted Vdss for 22 NHANES chemicals using Monte Carlo sampling to explore the impact of uncertainty in the predictions for each IAT in each compound.

Section snippets

Materials and methods

This study uses existing ionization prediction tools (ChemAxon, SPARC, and ADMET Predictor) for a library of 32,413 chemicals. The predictions are organized by IAT to characterize probability distributions for certain types of ionization events in order to perform sensitivity analysis of the predicted Vdss.

For clarification in this manuscript, the word “distribution” is used in three ways: 1) with respect to PK, the concentration of chemical in different tissues of the body (e.g., a lipophilic

Results

pKa predictions were generated for the 32,413 chemicals. Fig. 3 illustrates the number of association and dissociation actions of these chemicals using the predicted pKa values derived from ChemAxon versus the pH at which the ionization occurs. This also shows the acidic/basic pKa predictions as a function of chemical class. The ionization profiles of pharmaceutical compounds versus compounds with potential near- and far-field environmental sources were distinct: Far-field chemicals had the

Discussion

A major challenge in assessing any risk posed by thousands of untested and unmeasured chemicals is cost-efficient predictive models (National Academies of Sciences and Medicine, 2017). PK methods are needed to keep pace with expanded biomonitoring data (i.e., exposure) and high-throughput screening (i.e., hazard). High-throughput methods allow for prediction of key PK properties such as Vdss and tissue-specific PK distribution coefficients (Haddad et al., 2000, Peyret et al., 2010, Poulin and

Acknowledgments

The Oak Ridge Institute for Science and Education provided funding for K. Mansouri and initially for C. Strope (EPA-ORD/NCCT-2012-05) before continuing at the Hamner Institutes for Health Sciences. The U.S. Environmental Protection Agency through its Office of Research and Development funded and managed a portion of the research described here, with support for J.F. Wambaugh, J.R. Rabinowitz, and C. Stevens.

We would like to thank Thomas Peyret for sharing the Excel spreadsheets with the code

Disclaimer

The U.S. Environmental Protection Agency (U.S. EPA) through its Office of Research and Development funded and conducted the research described here. It has been reviewed by the U.S. EPA and approved for publication. The views expressed in this publication are those of the authors and do not necessarily represent the views or policies of the Agency. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

References (61)

  • T. Rodgers et al.

    Physiologically based pharmacokinetic modelling 2: predicting the tissue distribution of acids, very weak bases, neutrals and zwitterions

    J. Pharm. Sci.

    (2006)
  • T. Rodgers et al.

    Physiologically based pharmacokinetic modeling 1: predicting the tissue distribution of moderate-to-strong bases

    J. Pharm. Sci.

    (2005)
  • W. Schmitt

    General approach for the calculation of tissue to plasma partition coefficients

    Toxicol. in Vitro

    (2008)
  • J.A. Arnot et al.

    Screening level risk assessment model for chemical fate and effects in the environment

    Environ. Sci. Technol.

    (2006)
  • A.D. Bochevarov et al.

    Multiconformation, density functional theory-based pKa prediction in application to large, flexible organic molecules with diverse functional groups

    J. Chem. Theory Comput.

    (2016)
  • J.C. Caldwell et al.

    Cutting edge PBPK models and analyses: providing the basis for future modeling efforts and bridges to emerging toxicology paradigms

    J. Toxicol.

    (2012)
  • CDC

    National Health and Nutrition Examination Survey

    (2012)
  • F.S. Collins et al.

    Transforming environmental health protection

    (2008)
  • D.J. Dix et al.

    The ToxCast program for prioritizing toxicity testing of environmental chemicals

    Toxicol. Sci.

    (2007)
  • W.J. Doucette

    Quantitative structure-activity relationships for predicting soil-sediment sorption coefficients for organic chemicals

    Environ. Toxicol. Chem.

    (2003)
  • Environment and Climate Change Canada

    Domestic Substances List

    (2016)
  • P.A. Fenner-Crisp et al.

    The endocrine disruptor screening program developed by the US Environmental Protection Agency

    Ecotoxicology

    (2000)
  • A. Franco et al.

    Estimation of the soil–water partition coefficient normalized to organic carbon for ionizable organic chemicals

    Environ. Toxicol. Chem.

    (2008)
  • A. Franco et al.

    Influence of soil pH on the sorption of ionizable chemicals: modeling advances

    Environ. Toxicol. Chem.

    (2009)
  • S. Geidl et al.

    How does the methodology of 3D structure preparation influence the quality of pKa prediction?

    J. Chem. Inf. Model.

    (2015)
  • K.A. Hasselbalch

    Die Berechnung der Wasserstoffzahl des Blutes aus der freien und gebundenen Kohlensäure desselben, und die Sauerstoffbindung des Blutes als Funktion der Wasserstoffzahl

    (1916)
  • Health Canada

    Final Integrated Framework for the Health-Related Components of Categorization of the Domestic Substances List Under CEPA 1999

    (2009)
  • H.E. Helson

    Structure diagram generation

    Rev. Comput. Chem.

    (2007)
  • L.J. Henderson

    Concerning the relationship between the strength of acids and their capacity to preserve neutrality

    Am. J. Phys.

    (1908)
  • B.L. Ingle et al.

    Informing the human plasma protein binding of environmental chemicals by machine learning in the pharmaceutical space: applicability domain and limits of predictability

    J. Chem. Inf. Model.

    (2016)
  • Cited by (20)

    • In silico pharmacology

      2023, Computational Approaches in Drug Discovery, Development and Systems Pharmacology
    • Prioritization of mixtures of neurotoxic chemicals for biomonitoring using high-throughput toxicokinetics and mixture toxicity modeling

      2023, Environment International
      Citation Excerpt :

      data of the httk package. A list of all relevant input parameters for the 13,573 chemicals with data sources is summarized in Table B2 (Brown et al., 2007; Gülden et al., 2002; Honda et al., 2019; Ito and Houston, 2004; Jones et al., 2017; Linakis et al., 2020; Lombardo et al., 2018; McGinnity et al., 2004; Obach, 1999; Paixão et al., 2012; Shibata et al., 2002; Sipes et al., 2017; Strope et al., 2018; Tonnelier et al., 2012; Uchimura et al., 2010; Wambaugh et al., 2014, 2019; Wetmore, 2015; Wetmore et al., 2012b; Wood et al., 2017; Jones et al., 2002; Asturiol et al., 2020). Bioassay data.

    • Using chemical structure information to develop predictive models for in vitro toxicokinetic parameters to inform high-throughput risk-assessment

      2020, Computational Toxicology
      Citation Excerpt :

      The HTTK package LogP was obtained from the DSSTox database [51] or predicted using the EPA’s estimation program interface (EPI) suite (http://www.epa.gov/tsca-screening-tools/epi-suitetmestimation-program-interface) or using OPERA [71]. pKa was taken from the literature when available or taken from predictions produced by Strope et al [72]. The in vitro and in silico Css calculations were compared to get an estimate of variance in Css values based on the models developed in this work.

    • New approach methodologies for exposure science

      2019, Current Opinion in Toxicology
      Citation Excerpt :

      The parent compound's physical––chemical properties are the first determinant of how rapidly it will be transformed. These transformation processes parallel absorption, distribution, metabolism, and excretion that take place inside an organism, albeit at different rates [90,113,144]. Environmental transport and fate processes involve the inter-relationships between and within food webs, including humans [53].

    View all citing articles on Scopus
    View full text