Predicting the intrinsic membrane permeability of Caco-2/MDCK cells by the solubility-diffusion model.

Membrane permeability is one of the main determinants for the absorption, distribution, metabolism and excretion of compounds and is therefore of crucial importance for successful drug development. Experiments with artificial phospholipid membranes have shown that the intrinsic membrane permeability (P0) of compounds is well-predicted by the solubility-diffusion model (SDM). However, using the solubility-diffusion model to predict the P0 of biological Caco-2 and MDCK cell membranes has proven unreliable so far. Recent publications revealed that many published P0 extracted from Caco-2 and MDCK experiments are incorrect. In this work, we therefore used a small self-generated set as well as a large revised set of experimental Caco-2 and MDCK data from literature to compare experimental and predicted P0. The P0 extracted from Caco-2 and MDCK experiments were systematically lower than the P0 predicted by the solubility-diffusion model. However, using the following correlation: log P0,Caco-2/MDCK = 0.84 log P0,SDM - 1.85, P0 of biological Caco-2 and MDCK cell membranes was well-predicted by the solubility-diffusion model.

Effects of Aqueous Boundary Layers and Paracellular Transport on the Efflux Ratio as a Measure of Active Transport Across Cell Layers.

The efflux ratio (ER), determined by Caco-2/MDCK assays, is the standard in vitro metric to establish qualitatively whether a compound is a substrate of an efflux transporter. However, others have also enabled the utilisation of this metric quantitatively by deriving a relationship that expresses the ER as a function of the intrinsic membrane permeability of the membrane (P0) as well as the permeability of carrier-mediated efflux (Ppgp). As of yet, Ppgp cannot be measured directly from transport experiments or otherwise, but the ER relationship provides easy access to this value if P0 is known. However, previous derivations of this relationship failed to consider the influence of additional transport resistances such as the aqueous boundary layers (ABLs) and the filter on which the monolayer is grown. Since single fluxes in either direction can be heavily affected by these experimental artefacts, it is crucial to consider the potential impact on the ER. We present a model that includes these factors and show both mathematically and experimentally that this simple ER relationship also holds for the more realistic scenario that does not neglect the ABLs/filter. Furthermore, we also show mathematically how paracellular transport affects the ER, and we experimentally confirm that paracellular dominance reduces the ER to unity and can mask potential efflux.

Pitfalls in evaluating permeability experiments with Caco-2/MDCK cell monolayers.

When studying the transport of molecules across biological membranes, intrinsic membrane permeability (P0) is more informative than apparent permeability (Papp), because it eliminates external (setup-specific) factors, provides consistency across experiments and mechanistic insight. It is thus an important building block for modeling the total permeability in any given scenario. However, extracting P0 is often difficult, if not impossible, when the membrane is not the dominant transport resistance. In this work, we set out to analyze Papp values measured with Caco-2/MDCK cell monolayers of 69 literature references. We checked the Papp values for a total of 318 different compounds for the extractability of P0, considering possible limitations by aqueous boundary layers, paracellular transport, recovery issues, active transport, a possible proton flux limitation, and sink conditions. Overall, we were able to extract 77 reliable P0 values, which corresponds to about one quarter of the total compounds analyzed, while about half were limited by the diffusion through the aqueous layers. Compared to an existing data set of P0 values published by Avdeef, our approach resulted in a much higher exclusion of compounds. This is a consequence of stricter compound- and reference-specific exclusion criteria, but also because we considered possible concentration-shift effects due to different pH values in the aqueous layers, an effect only recently described in literature. We thus provide a consistent and reliable set of P0, e.g. as a basis for future modeling.

Revisiting the pKa-Flux method for determining intrinsic membrane permeability.

Intrinsic membrane permeability is one of several factors that critically determine the intestinal absorption of a chemical. The intrinsic membrane permeability of a chemical is usually extracted from transwell experiments with Caco-2 or MDCK cells, preferably by the pKa-Flux method, which is considered the method of choice when aqueous boundary layer effects need to be excluded. The pKa-Flux method has two variants, the iso-pH method, where apical and basolateral pH are equal, and the gradient-pH method, where apical and basolateral pH are different. The most commonly used method is the gradient-pH method, as it is intended to reflect the pH-conditions in the gastrointestinal tract. However, concentration-shift effects caused by the applied pH-difference between apical and basolateral compartment in the gradient-pH method have not been considered in the evaluation of the experimental data in the past. Consequently, incorrect intrinsic membrane permeabilities have been determined. In this work, we present a revised method for extracting the intrinsic membrane permeability from gradient-pH data that considers concentration-shift effects in the basolateral aqueous boundary layer and filter as well as in the cytosol. Furthermore, we propose the use of the iso-pH method, where only concentration-shift effects in the cytosol need to be considered, as an alternative to the gradient-pH method. We use the five lipophilic bases amantadine, chloroquine, propranolol, venlafaxine and verapamil as examples to compare gradient-pH method and iso-pH method with regard to the extractability of the intrinsic membrane permeability. For lipophilic bases, the iso-pH method proves to be advantageous. All intrinsic membrane permeabilities determined in this work were substantially higher than the intrinsic membrane permeabilities reported in literature.

Mechanistic modeling of the bioconcentration of (super)hydrophobic compounds in Hyalella azteca.

Bioconcentration tests using the freshwater amphipod Hyalella azteca as an alternative to conventional fish tests have recently received much attention. An appropriate computational model of H. azteca could help in understanding the mechanisms behind bioconcentration, in comparison to the fish as test organism. We here present the first mechanistic model for H. azteca that considers the single diffusive processes in the gills and gut. The model matches with the experimental data from the literature quite well when appropriate physiological information is used. The implementation of facilitated transport was essential for modeling. Application of the model for superhydrophobic compounds revealed binding to organic matter and the resulting decrease in bioavailable fraction as the main reason for the observed counterintuitive decrease in uptake rate constants with increasing octanol/water partition coefficient. Furthermore, estimations of the time needed to reach steady state indicated that durations of more than a month could be needed for compounds with a log Kow > 8, limiting the experimental applicability of the test. In those cases, model-based bioconcentration predictions could be a preferable approach, which could be combined with in vitro biotransformation measurements. However, our sensitivity analysis showed that the uncertainty in determining the octanol/water partition coefficients is a strong source of error for superhydrophobic compounds.

Impact of cholesterol and sphingomyelin on intrinsic membrane permeability.

Transwell experiments with Caco-2 or MDCK cells are the gold standard for determining the intestinal permeability of chemicals. The intrinsic membrane permeability (P0), that can be extracted from these experiments, might be comparable to P0 measured in black lipid membrane (BLM) experiments and P0 predicted by the solubility-diffusion model. Unfortunately, the overlap between experimental P0,Caco-2/MDCK and P0,BLM data is very small. So far, differences between both approaches have been attributed to the cholesterol and sphingomyelin content of cell membranes, but the database is too sparse to thoroughly test this theory. To create a diverse dataset, we measured P0,BLM of ten chemicals in BLM experiments using DPhPC and DPhPC/cholesterol/sphingomyelin membranes. The results were compared to predicted BLM data and experimental Caco-2/MDCK data obtained from literature. While P0,BLM of all chemicals was well predicted by the solubility-diffusion model, P0,Caco-2/MDCK was only predictable for rather hydrophilic compounds with logarithmic hexadecane/water partition coefficients below -0.5. The effect of cholesterol and sphingomyelin on P0,BLM was negligibly small.

Screening of 6000 Compounds for Uncoupling Activity: A Comparison Between a Mechanistic Biophysical Model and the Structural Alert Profiler Mitotox.

Protonophoric uncoupling of phosphorylation is an important factor when assessing chemicals for their toxicity, and has recently moved into focus in pharmaceutical research with respect to the treatment of diseases such as cancer, diabetes, or obesity. Reliably identifying uncoupling activity is thus a valuable goal. To that end, we screened more than 6000 anionic compounds for in vitro uncoupling activity, using a biophysical model based on ab initio COSMO-RS input parameters with the molecular structure as the only external input. We combined these results with a model for baseline toxicity (narcosis). Our model identified more than 1250 possible uncouplers in the screening dataset, and identified possible new uncoupler classes such as thiophosphoric acids. When tested against 423 known uncouplers and 612 known inactive compounds in the dataset, the model reached a sensitivity of 83% and a specificity of 96%. In a direct comparison, it showed a similar specificity than the structural alert profiler Mitotox (97%), but much higher sensitivity than Mitotox (47%). The biophysical model thus allows for a more accurate screening for uncoupling activity than existing structural alert profilers. We propose to use our model as a complementary tool to screen large datasets for protonophoric uncoupling activity in drug development and toxicity assessment.

Could chemical exposure and bioconcentration in fish be affected by slow binding kinetics in blood?

The possible implications of slow binding kinetics on respiratory uptake, bioconcentration and exposure of chemicals were evaluated in the present study. Most physiological and chemical information needed for such an evaluation is already known from the literature or can be estimated. However, data for binding kinetics of chemicals in fish plasma have not been reported in the literature yet. In the first part of this study, we therefore experimentally investigated the plasma binding kinetics for ten chemicals, including pollutants like polycyclic aromatic hydrocarbons and a pesticide. The determined desorption rate constants were in the range of 0.4 s-1 to 0.1 s-1. In the second part of this study, we present a comparative modeling analysis of generic predictions with binding kinetics of different velocities. For doing so, a model that explicitly represents binding kinetics in blood was developed and applied for different hypothetical scenarios. The evaluation showed that slow sorption kinetics only limits respiratory uptake and thus influences the levels of bioaccumulation for extreme and, by that, rather unlikely parameter combinations (i.e. for strongly sorbing chemicals with very slow binding kinetics). It can therefore be assumed that limitations on respiratory uptake due to slow binding kinetics in blood are rather unlikely for most chemicals.

Relevance of desorption kinetics and permeability for in vitro-based predictions of hepatic clearance in fish.

The impact of desorption kinetics and permeation kinetics on in vitro-based predictions of in vivo hepatic blood clearances is investigated in the present study. Most commonly, possible limitations due to slow desorption of chemicals from albumin or slow permeation of chemicals through cellular membranes are not considered when in vivo clearances are predicted from in vitro biotransformation rate constants. To evaluate whether the most commonly used extrapolation models might thus overlook important kinetic limitations, we compare predictions of in vivo clearance that explicitly consider desorption and permeation kinetics with predictions of in vivo clearance that neglect these aspects. Our results show that strong limitations due to slow permeation kinetics are possible depending on the assumed permeability value. While permeability values estimated with a mechanistic approach are fast enough to avoid significant limitations, other experimentally derived permeability values lead to dramatically decreased in vivo clearance predictions. These latter values lead to unrealistically low in vivo biotransformation estimates. Furthermore, we also evaluated the implications of desorption kinetics using experimentally determined desorption rate constants. These evaluations show that slow desorption kinetics are unlikely to limit in vivo clearance.

Prediction of Unbound Fractions for in Vitro-in Vivo Extrapolation of Biotransformation Data.

For in vitro-in vivo extrapolation of biotransformation data, the different sorptive environments in vitro and in vivo need to be considered. The most common approach for doing so is using the ratio of unbound fractions in vitro and in vivo. In the literature, several algorithms for prediction of these unbound fractions are available. In this study, we present a theoretical evaluation of the most commonly used algorithms for prediction of unbound fractions in S9 assays and blood and compare prediction results with empirical values from the literature. The results of this analysis prove a good performance of "composition-based" algorithms, i.e. algorithms that represent the inhomogeneous composition of in vitro assay and in vivo system and describe sorption to the individual components (lipids, proteins, water) in the same way. For strongly sorbing chemicals, these algorithms yield constant values for the ratio of unbound fractions in vitro and in vivo. This is mechanistically plausible, because in these cases, the chemicals are mostly bound, and the ratio of unbound fractions is determined by the volume ratio of sorbing components in both phases.

Exploring the octanol-water partition coefficient dataset using deep learning techniques and data augmentation.

Today more and more data are freely available. Based on these big datasets deep neural networks (DNNs) rapidly gain relevance in computational chemistry. Here, we explore the potential of DNNs to predict chemical properties from chemical structures. We have selected the octanol-water partition coefficient (log P) as an example, which plays an essential role in environmental chemistry and toxicology but also in chemical analysis. The predictive performance of the developed DNN is good with an rmse of 0.47 log units in the test dataset and an rmse of 0.33 for an external dataset from the SAMPL6 challenge. To this end, we trained the DNN using data augmentation considering all potential tautomeric forms of the chemicals. We further demonstrate how DNN models can help in the curation of the log P dataset by identifying potential errors, and address limitations of the dataset itself.

Environmental Sorption Behavior of Ionic and Ionizable Organic Chemicals.

Traditionally our tools for environmental risk assessment of organic chemicals have been developed for neutral chemicals. However, many commercial chemicals are ionic or ionizable and require different tools and approaches for their assessment. In recent years this task starts to obtain increasing attention but our understanding for their environmental fate is still far behind that for neutral chemicals. This review first gives an overview on the principles that govern ionic partitioning in environmental systems which are more complex than the simple partition processes of neutral chemicals. Second, a summary of our current knowledge on various topics such as bioaccumulation, sorption in soils, and nonspecific-toxicity reveals that ionic species can actually be quite hydrophobic contrary to commonly held beliefs. Eventually, we discuss existing models for the quantitative prediction of organic ions' sorption in soils and biota. We have to assert that the available model tools are quite restricted in their application range compared to neutral chemicals which is due to the higher complexity of the various ionic sorption processes. In order to further advance our understanding more high-quality sorption data are needed with a focus on multivalent and zwitterionic ions in all partition systems as well as cations in biological matrices.

Estimating the Equilibrium Distribution of Perfluoroalkyl Acids and 4 of Their Alternatives in Mammals.

Perfluoroalkyl acids (PFAAs) mostly exist as ionic compounds that are of major concern because of their accumulative behavior. The discussion about their risk is ongoing considering the increasing production of structurally similar alternatives. We conducted model calculations based on equilibrium distribution coefficients that allow studying the distribution of PFAAs and their alternatives in various mammalian organs through comparison to empirical measurements in humans and rats. The calculations rely on experimentally determined distribution coefficients of a series of PFAAs and 4 of their alternatives to physiological matrices such as structural proteins, storage lipids, membrane lipids, albumin, and fatty acid binding protein (FABP). The relative sorption capacities in each organ were calculated from the combination of distribution coefficients and physiological data. The calculated distribution of PFAAs and alternatives within the organs showed that albumin and membrane lipids and, to a lesser extent, structural proteins have the highest relative sorption capacities for the compounds. Sorption to FABP is only relevant in the distribution of short-chain PFAAs. Storage lipids play a minor role in the distribution of all studied compounds. Our calculated distribution of PFAAs was evaluated by comparison to reported PFAA concentrations in various organs. Environ Toxicol Chem 2021;40:910-920. © 2020 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Yolk Sac of Zebrafish Embryos as Backpack for Chemicals?

The zebrafish embryo (Danio rerio) has developed into one of the most important nonsentient animal models for the hazard assessments of chemicals, but the processes governing its toxicokinetics (TK) are poorly understood. This study compares the uptake of seven test compounds into the embryonic body and the yolk sac of the zebrafish embryo using TK experiments, a dialysis approach, thermodynamic calculations, and kinetic modeling. Experimental data show that between 95% (4-iodophenol) and 67% (carbamazepine) of the total internal amount in 26 h post fertilization (hpf) embryos and between 80 and 49% in 74 hpf embryos were found in the yolk. Thus, internal concentrations determined for the whole embryo overestimate the internal concentration in the embryonic body: for the compounds of this study, up to a factor of 5. Partition coefficients for the embryonic body and a one-compartment model with diffusive exchange were calculated for the neutral test compounds and agreed reasonably with the experimental data. For prevalently ionic test compounds at exposure pH (bromoxynil, paroxetine), however, the extent and the speed of uptake were low and could not be modeled adequately. A better understanding of the TK of ionizable test compounds is essential to allow assessment of the validity of this organismic test system for ionic test compounds.

Yolk-Water Partitioning of Neutral Organic Compounds in the Model Organism Danio rerio.

Yolk is the most important temporary biological compartment of the early life stages of fish embryos. The sorption strength of a chemical to yolk components may significantly influence the distribution of that chemical in the fish embryo. We determined yolk-water partition coefficients (Kyolk/water , in liters of water per kilogram of yolk, normalized to dry wt) for 70 neutral organic chemicals. The log Kyolk/water values range from 0.76 to 6.56. On the basis of these values, we developed polyparameter linear free energy relationship models to predict yolk-water partitioning for a broad range of neutral organic chemicals with a root mean squared error of 0.37 and r2 of 0.919. These models can be applied for the prediction of internal concentrations at equilibrium (neglecting biotransformation and active transport) in the zebrafish embryo test system. Environ Toxicol Chem 2020;39:1506-1516. © 2020 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Predicting Uncoupling Toxicity of Organic Acids Based on Their Molecular Structure Using a Biophysical Model.

We present a purely mechanistic model to predict protonophoric uncoupling activity ECw of organic acids. All required input information can be derived from their chemical structure. This makes it a convenient predictive model to gain valuable information on the toxicity of organic chemicals already at an early stage of development of new commercial chemicals (e.g., in agriculture or pharmaceutical industries). A critical component of the model is the consideration of the possible formation of heterodimers from the neutral and anionic monomer, and its permeation through the membrane. The model was tested against literature data measured in chromatophores, submitochondrial particles, isolated mitochondria, and intact green algae cells with good success. It was also possible to reproduce pH-dependencies in isolated mitochondria and intact cells. Besides the prediction of the ECw, the mechanistic nature of the model allows researchers to draw direct conclusions on the impact of single input factors such as pH- and voltage-gradients across the membrane, the anionic and neutral membrane permeability, and the heterodimerization constant. These insights are of importance in drug design or chemical regulation.

Comparison of a simple and a complex model for BCF prediction using in vitro biotransformation data.

A promising approach for bioaccumulation assessment with reduced animal use is the prediction of bioconcentration factors (BCFs) using in vitro biotransformation data. However, it has been recognized that the BCFs predicted using current models often are in poor agreement with experimental BCFs. Furthermore, extrahepatic biotransformation (e.g. in gill or GIT) is usually not accounted for. Here, we compare two BCF prediction models: a simple one-compartment and a more advanced multi-compartment model. Both models are implemented in a two-in-one calculation tool for the prediction of BCFs using in vitro data. Furthermore, both models were set up in a way that in vitro data for extrahepatic biotransformation can be easily considered, if desired. The models differ in their complexity: the one-compartment model is attractive because its simplicity, while the multi-compartment model is characterized by its refined closeness to reality. A comparison of the results shows that both models yield almost identical results for the presently evaluated cases with plausible physiological data. For regulatory purposes, there is thus no reason not to use the simple one-compartment model. However, if it is desired to represent special in vivo characteristics, e.g. first-pass effects or the direct GIT-to-liver blood flow, the multi-compartment model should be used.

COSMOperm: Mechanistic Prediction of Passive Membrane Permeability for Neutral Compounds and Ions and Its pH Dependence.

We present a new and entirely mechanistic COSMOperm method to predict passive membrane permeabilities for neutral compounds, as well as anions and cations. The COSMOperm approach is based on compound-specific free energy profiles within a membrane of interest from COSMO-RS (conductor-like screening model for realistic solvation) calculations. These are combined with membrane layer-specific diffusion coefficients, for example, in the water phase, the polar head groups, and the alkyl tails of biochemical phospholipid bilayers. COSMO-RS utilizes first-principle quantum chemical structures and physically sound intermolecular interactions (electrostatic, hydrogen bond, and van der Waals). For this reason, it is unbiased toward different application scenarios, such as in cosmetics and industrial chemical or pharmaceutical industries. A fully predictive calculation of passive permeation through phospholipid bilayer membranes results in a performance of r2 = 0.92; rmsd = 0.90 log10 units for neutral compounds and anions, as compared to gold standard black lipid membrane experiments. It will be demonstrated that new membrane types can be generated by the related COSMOplex method and directly used for permeability studies by COSMOperm.

Membrane/Water Partitioning and Permeabilities of Perfluoroalkyl Acids and Four of their Alternatives and the Effects on Toxicokinetic Behavior.

The search for alternatives to bioaccumulative perfluoroalkyl acids (PFAAs) is ongoing. New, still highly fluorinated alternatives are produced in hopes of reducing bioaccumulation. To better estimate this bioaccumulative behavior, we performed dialysis experiments and determined membrane/water partition coefficients, Kmem/w, of six perfluoroalkyl carboxylic acids (PFCAs), three perfluoroalkanesulfonic acids, and four alternatives. We also investigated how passive permeation might influence the uptake kinetics into cells, measuring the passive anionic membrane permeability Pion through planar lipid bilayers for six PFAAs and three alternatives. Experimental Kmem/w and Pion were both predicted well by the COSMO-RS theory (log RMSE 0.61 and 0.46, respectively). Kmem/w values were consistent with the literature data, and alternatives showed similar sorption behavior as PFAAs. Experimental Pion values were high enough to explain observed cellular uptake by passive diffusion with no need to postulate the existence of active uptake processes. However, predicted pKa and neutral permeabilities suggest that also the permeation of the neutral species should be significant in case of PFCAs. This can have direct consequences on the steady-state distribution of PFAAs across cell membranes and thus toxicity. Consequently, we propose a model to predict pH-dependent baseline toxicity based on Kmem/w, which considers the permeation of both neutral and anionic species.