Effects of experimental setup on the apparent concentration dependency of active efflux transport in in vitro cell permeation experiments

Can We Predict Clinical Pharmacokinetics of Highly Lipophilic Compounds by Integration of Machine Learning or In Vitro Data into Physiologically Based Models? A Feasibility Study Based on 12 Development Compounds

While high lipophilicity tends to improve potency, its effects on pharmacokinetics (PK) are complex and often unfavorable. To predict clinical PK in early drug discovery, we built human physiologically based PK (PBPK) models integrating either (i) machine learning (ML)-predicted properties or (ii) discovery stage in vitro data. Our test set was composed of 12 challenging development compounds with high lipophilicity (mean calculated log P 4.2), low plasma-free fraction (50% of compounds with fu,p < 1%), and low aqueous solubility. Predictions focused on key human PK parameters, including plasma clearance (CL), volume of distribution at steady state (V_ss), and oral bioavailability (%F). For predictions of CL, the ML inputs showed acceptable accuracy and slight underprediction bias [an average absolute fold error (AAFE) of 3.55; an average fold error (AFE) of 0.95]. Surprisingly, use of measured data only slightly improved accuracy but introduced an overprediction bias (AAFE = 3.35; AFE = 2.63). Predictions of V_ss were more successful, with both ML (AAFE = 2.21; AFE = 0.90) and in vitro (AAFE = 2.24; AFE = 1.72) inputs showing good accuracy and moderate bias. The %F was poorly predicted using ML inputs [average absolute prediction error (AAPE) of 45%], and use of measured data for solubility and permeability improved this to 34%. Sensitivity analysis showed that predictions of CL limited the overall accuracy of human PK predictions, partly due to high nonspecific binding of lipophilic compounds, leading to uncertainty of unbound clearance. For accurate predictions of %F, solubility was the key factor. Despite current limitations, this work encourages further development of ML models and integration of their results within PBPK models to enable human PK prediction at the drug design stage, even before compounds are synthesized. Further evaluation of this approach with more diverse chemical types is warranted.

A Permeability- and Perfusion-based PBPK model for Improved Prediction of Concentration-time Profiles

To improve predictions of concentration‐time (C‐t) profiles of drugs, a new physiologically based pharmacokinetic modeling framework (termed ‘PermQ’) has been developed. This model includes permeability into and out of capillaries, cell membranes, and intracellular lipids. New modeling components include (i) lumping of tissues into compartments based on both blood flow and capillary permeability, and (ii) parameterizing clearances in and out of membranes with apparent permeability and membrane partitioning values. Novel observations include the need for a shallow distribution compartment particularly for bases. C‐t profiles were modeled for 24 drugs (7 acidic, 5 neutral, and 12 basic) using the same experimental inputs for three different models: Rodgers and Rowland (RR), a perfusion‐limited membrane‐based model (K_p,mem), and PermQ. K_p,mem and PermQ can be directly compared since both models have identical tissue partition coefficient parameters. For the 24 molecules used for model development, errors in V_ss and t_1/2 were reduced by 37% and 43%, respectively, with the PermQ model. Errors in C‐t profiles were reduced (increased EOC) by 43%. The improvement was generally greater for bases than for acids and neutrals. Predictions were improved for all 3 models with the use of parameters optimized for the PermQ model. For five drugs in a test set, similar results were observed. These results suggest that prediction of C‐t profiles can be improved by including capillary and cellular permeability components for all tissues.

In Vitro–In Silico Modeling of Caffeine and Diclofenac Permeation in Static and Fluidic Systems with a 16HBE Lung Cell Barrier

Static in vitro permeation experiments are commonly used to gain insights into the permeation properties of drug substances but exhibit limitations due to missing physiologic cell stimuli. Thus, fluidic systems integrating stimuli, such as physicochemical fluxes, have been developed. However, as fluidic in vitro studies display higher complexity compared to static systems, analysis of experimental readouts is challenging. Here, the integration of in silico tools holds the potential to evaluate fluidic experiments and to investigate specific simulation scenarios. This study aimed to develop in silico models that describe and predict the permeation and disposition of two model substances in a static and fluidic in vitro system. For this, in vitro permeation studies with a 16HBE cellular barrier under both static and fluidic conditions were performed over 72 h. In silico models were implemented and employed to describe and predict concentration–time profiles of caffeine and diclofenac in various experimental setups. For both substances, in silico modeling identified reduced apparent permeabilities in the fluidic compared to the static cellular setting. The developed in vitro–in silico modeling framework can be expanded further, integrating additional cell tissues in the fluidic system, and can be employed in future studies to model pharmacokinetic and pharmacodynamic drug behavior.

Analysis of Non-linear Pharmacokinetics of P-Glycoprotein Substrates in a Microfluidic Device Using a Mathematical Model that Includes an Unstirred Water Layer (UWL) Compartment

PURPOSE

The purpose of this research is to analyze non-linear pharmacokinetics of P-glycoprotein (P-gp) substrates in a cell based assay of a microfluidic device, which might be affected by hydrodynamic barrier (unstirred water layer, UWL).

RESULTS

Apparent permeability (P_app) were obtained using non-P-gp substrates (propranolol, metoprolol, and atenolol) and P-gp substrates (quinidine and talinolol) in a commercially available microfluidic device, organoplate ® of Caco-2 cell based assay. The previous UWL resistance model was well fitted to P_app of static and flow condition by assuming UWL including and negligible condition, while P-gp substrates of higher passive permeability (quinidine) was apart from the fitting curve. The concentration dependent non-linear kinetics of P-gp substrates, quinidine and talinolol, was more analyzed in detail, and apparent V_max discrepancy between static and flow assay condition in the quinidine assay was observed, while that was not observed in talinolol, the lower permeable substrate. Based on the experimental results, a mathematical model for P-gp substrates including UWL compartment on the previous 3-compartment model was developed, and it indicated that the apparent V_max was variable along with the ratio between passive permeability and UWL permeability.

CONCLUSIONS

The mathematical model adding UWL compartment well explained non-linear pharmacokinetics of apparent permeability of P-gp substrate in the microfluidic device. The model also has a potential to be applied to P-gp substrate permeability analysis in vivo.

P-glycoprotein Substrate Assessment in Drug Discovery: Application of Modeling to Bridge Differential Protein Expression Across In Vitro Tools

P-glycoprotein (P-gp) efflux assay is an integral part of discovery screening, especially for drugs requiring brain penetration as P-gp efflux ratio (ER) inversely correlates with brain exposure. However, significant variability in P-gp ER generated across cell lines can lead to misclassification of a P-gp substrate and subsequently disconnect with brain exposure data. We hypothesized that the ER depends on P-gp protein expression level in the in vitro assay. Quantitative proteomics and immunofluorescence staining were utilized to characterize P-gp protein expression and localization in four recombinant cell lines, over-expressing human or mouse P-gp isoforms, followed by functional evaluation. Efflux data generated in each cell line was compared against available rodent brain distribution data. The results suggested that the cell line with highest P-gp expression (hMDCK-MDR1 sourced from NIH) led to greatest dynamic range for efflux; thus, proving to be the most sensitive model to predict brain penetration. Cell lines with lower P-gp expression exhibited the greatest tendency for compound-dependent in vitro efflux saturation leading to false negative results. Ultimately, P-gp kinetics were characterized using a compartmental model to generate system-independent parameters to resolve such discrepancy. This study highlights the need for careful choice of well characterized P-gp in vitro tools and utility of modeling techniques to enable appropriate interpretation of the data.

Characterization and Validation of Canine P-Glycoprotein-Deficient MDCK II Cell Lines for Efflux Substrate Screening

PURPOSE

We characterized three canine P-gp (cP-gp) deficient MDCKII cell lines. Their relevance for identifying efflux transporter substrates and predicting limitation of brain penetration were evaluated. In addition, we discuss how compound selection can be done in drug discovery by using these cell systems.

METHOD

hMDR1, hBCRP-transfected, and non-transfected MDCKII ZFN cells (all with knock-down of endogenous cP-gp) were used for measuring permeability and efflux ratios for substrates. The compounds were also tested in MDR1_Caco-2 and BCRP_Caco-2, each with a double knock-out of BCRP/MRP2 or MDR1/MRP2 transporters respectively. Efflux results were compared between the MDCK and Caco-2 models. Furthermore, in vitro MDR1_ZFN efflux data were correlated with in vivo unbound drug brain-to-plasma partition coefficient (K_p,_uu).

RESULTS

MDR1 and BCRP substrates are correctly classified and robust transporter affinities with control substrates are shown. Cell passage mildly influenced mRNA levels of transfected transporters, but the transporter activity was proven stable for several years. The MDCK and Caco-2 models were in high consensus classifying same efflux substrates. Approx. 80% of enlisted substances were correctly predicted with the MDR1_ZFN model for brain penetration.

CONCLUSION

cP-gp deficient MDCKII ZFN models are reliable tools to identify MDR1 and BCRP substrates and useful for predicting efflux liability for brain penetration.

Novel insights into the organic solute transporter alpha/beta, OSTα/β: From the bench to the bedside

Organic solute transporter alpha/beta (OSTα/β) is a heteromeric solute carrier protein that transports bile acids, steroid metabolites and drugs into and out of cells. OSTα/β protein is expressed in various tissues, but its expression is highest in the gastrointestinal tract where it facilitates the recirculation of bile acids from the gut to the liver. Previous studies established that OSTα/β is upregulated in liver tissue of patients with extrahepatic cholestasis, obstructive cholestasis, and primary biliary cholangitis (PBC), conditions that are characterized by elevated bile acid concentrations in the liver and/or systemic circulation. The discovery that OSTα/β is highly upregulated in the liver of patients with nonalcoholic steatohepatitis (NASH) further highlights the clinical relevance of this transporter because the incidence of NASH is increasing at an alarming rate with the obesity epidemic. Since OSTα/β is closely linked to the homeostasis of bile acids, and tightly regulated by the nuclear receptor farnesoid X receptor, OSTα/β is a potential drug target for treatment of cholestatic liver disease, and other bile acid-related metabolic disorders such as obesity and diabetes. Obeticholic acid, a semi-synthetic bile acid used to treat PBC, under review for the treatment of NASH, and in development for the treatment of other metabolic disorders, induces OSTα/β. Some drugs associated with hepatotoxicity inhibit OSTα/β, suggesting a possible role for OSTα/β in drug-induced liver injury (DILI). Furthermore, clinical cases of homozygous genetic defects in both OSTα/β subunits resulting in diarrhea and features of cholestasis have been reported. This review article has been compiled to comprehensively summarize the recent data emerging on OSTα/β, recapitulating the available literature on the structure-function and expression-function relationships of OSTα/β, the regulation of this important transporter, the interaction of drugs and other compounds with OSTα/β, and the comparison of OSTα/β with other solute carrier transporters as well as adenosine triphosphate-binding cassette transporters. Findings from basic to more clinically focused research efforts are described and discussed.

Examining P-gp efflux kinetics guided by the BDDCS - Rational selection of in vitro assay designs and mathematical models

The generation of reliable kinetic parameters to describe P-glycoprotein (P-gp) activity is essential for predicting the impact of efflux transport on gastrointestinal drug absorption. The compound-specific selection of in vitro assay designs and ensuing data analysis methods is explored in this manuscript. We measured transcellular permeability and cellular uptake of five P-gp substrates in Caco-2 and LLC-PK1 MDR1 cells. Kinetic parameters of P-gp-mediated efflux transport (K_m, V_max) were derived from conventional and mechanistic compartmental models. The estimated apparent K_m values based on medium concentrations in the conventional permeability model indicated significant differences between the cell lines. The respective intrinsic K_m values based on unbound intracellular concentrations in the mechanistic compartmental models were significantly lower and comparable between cell lines and assay formats. Non-specific binding or lysosomal trapping were shown to cause discrepancies in the kinetic parameters obtained from different assay formats. A guidance for the selection of in vitro assays and kinetic assessment methods is proposed in line with the Biopharmaceutics Drug Disposition Classification System (BDDCS). The recommendations are expected to aid the acquisition of robust and reproducible kinetic parameters of P-gp-mediated efflux transport.

Challenges of using in vitro data for modeling P-glycoprotein efflux in the blood-brain barrier

The efficacy of central nervous system (CNS) drugs may be limited by their poor ability to cross the bloodbrain barrier (BBB). Transporters, such as p-glycoprotein, may affect the distribution of many drugs into the CNS in conjunction with the restricted paracellular pathway of the BBB. It is therefore important to gain information on unbound drug concentrations in the brain in drug development to ensure sufficient drug exposure from plasma at the target site in the CNS. In vitro methods are routinely used in drug development to study passive permeability and p-glycoprotein efflux of new drugs. This review discusses the challenges in the use of in vitro data as input parameters in physiologically based pharmacokinetic (PBPK) models of CNS drug disposition of p-glycoprotein substrates. Experience with quinidine demonstrates the variability in in vitro parameters of passive permeability and active pglycoprotein efflux. Further work is needed to generate parameter values that are independent of the model and assay. This is a prerequisite for reliable predictions of drug concentrations in the brain in vivo.

Drug-transporter interaction testing in drug discovery and development

The human body consists of several physiological barriers that express a number of membrane transporters. For an orally absorbed drug the intestinal, hepatic, renal and blood-brain barriers are of the greatest importance. The ATP-binding cassette (ABC) transporters that mediate cellular efflux and the solute carrier transporters that mostly mediate cellular uptake are the two superfamilies responsible for membrane transport of vast majority of drugs and drug metabolites. The total number of human transporters in the two superfamilies exceeds 400, and about 40-50 transporters have been characterized for drug transport. The latest Food and Drug Administration guidance focuses on P-glycoprotein, breast cancer resistance protein, organic anion transporting polypeptide 1B1 (OATP1B1), OATP1B3, organic cation transporter 2 (OCT2), and organic anion transporters 1 (OAT1) and OAT3. The European Medicines Agency’s shortlist additionally contains the bile salt export pump, OCT1, and the multidrug and toxin extrusion transporters, multidrug and toxin extrusion protein 1 (MATE1) and MATE2/MATE2K. A variety of transporter assays are available to test drug-transporter interactions, transporter-mediated drug-drug interactions, and transporter-mediated toxicity. The drug binding site of ABC transporters is accessible from the cytoplasm or the inner leaflet of the plasma membrane. Therefore, vesicular transport assays utilizing inside-out vesicles are commonly used assays, where the directionality of transport results in drugs being transported into the vesicle. Monolayer assays utilizing polarized cells expressing efflux transporters are the test systems suggested by regulatory agencies. However, in some monolayers, uptake transporters must be coexpressed with efflux transporters to assure detectable transport of low passive permeability drugs. For uptake transporters mediating cellular drug uptake, utilization of stable transfectants have been suggested. In vivo animal models complete the testing battery. Some issues, such as in vivo relevance, gender difference, age and ontogeny issues can only be addressed using in vivo models. Transporter specificity is provided by using knock-out or mutant models. Alternatively, chemical knock-outs can be employed. Compensatory changes are less likely when using chemical knock-outs. On the other hand, specific inhibitors for some uptake transporters are not available, limiting the options to genetic knock-outs.

Use of physiologically based pharmacokinetic modeling for assessment of drug-drug interactions

Interactions between co-administered medicines can reduce efficacy or lead to adverse effects. Understanding and managing such interactions is essential in bringing safe and effective medicines to the market. Ideally, interaction potential should be recognized early and minimized in compounds that reach late stages of drug development. Physiologically based pharmacokinetic models combine knowledge of physiological factors with compound-specific properties to simulate how a drug behaves in the human body. These software tools are increasingly used during drug discovery and development and, when integrating relevant in vitro data, can simulate drug interaction potential. This article provides some background and presents illustrative examples. Physiologically based models are an integral tool in the discovery and development of drugs, and can significantly aid our understanding and prediction of drug interactions.