Mailys De Sousa Mendes, Matthew Harwood, Howard Burt, Sibylle Neuhoff
Certara UK, Simcyp Division, Sheffield, UK
Objectives:
Transporter inhibition can have an impact on the disposition of a drug as well as on its safety and efficacy. Being able to have reliable estimates of inhibition parameters for use in PBPK models is key to evaluating the DDI potential. However, the conventional analysis of the standard in vitro inhibition assays makes several assumptions that can impact the quality of the in-vivo prediction. For example, it assumes that sink conditions are maintained, which can be difficult to achieve experimentally, especially for highly permeable compounds. Moreover, it is sometimes challenging to robustly distinguish between the passive permeability of the substrate from the active transport. It also assumes that the driving concentration for the transporter inhibition is the nominal concentration. In addition, for efflux transporters the intracellular concentration typically perpetrates the inhibition. It has been shown for the substrates that using modelling to estimate the intracellular concentration decreases the inter-laboratory variability and tends to give lower and more consistent Km estimates [1]. The similar conclusions were recently made for inhibition parameters [2] and could explain the overestimation of Ki values frequently observed. We developed a model that mechanistically describes the efflux transport across Caco-2 cells for digoxin and quinidine, two P-gp substrates. The KiP-gp value for quinidine was also estimate using the in-vitro drug-drug interaction (DDI) with digoxin.
Methods:
In-vitro assays
Data for the bidirectional transport of quinidine and digoxin across Caco-2 monolayers were previously generated [3]. Briefly, Caco-2 cells were seeded at a density of 1 x 105 cells/well onto 12-well Transwell® inserts and grown for 23±1 days prior to permeability experiments. Experiments were performed at 37°C, with apical and basolateral volumes of 0.5 and 1.5 mL, respectively, and was stirred at 450 rpm (calibrated plate shaker (BMG LabTechnologies GmbH, Offenburg, Germany). The basolateral and apical compartment were buffered to a pH of 7.4. Digoxin disposition was characterised at concentrations of 0.059, 1, 10, and 100 µM. Quinidine disposition was characterised at concentrations of 0.001, 0.05, 1, 10, and 100 µM. Samples were collected at 5,15,25,50,80, and 120 min for both. For the DDI assays, concentrations of 0.059 µM vs 100 µM, 0.059 µM vs 10 µM and 0.02 µM vs 50 µM for digoxin and quinidine, respectively were used and samples were collected at 5, 15, 25, and 50 min. Sampling of A-B experiments was conducted by moving the Transwell insert to a new well containing blank buffer and retaining the previous well, thereby representing complete removal of drug from basolateral buffer. Sampling of B-A experiments was conducted by removal of 400 µl of apical buffer and replacement with an equal volume of blank buffer.
Data analysis (modelling)
A mechanistic model was developed in R software (version 3.5.1) and included 3 compartments, representing apical and basolateral media in addition to the cell monolayer for the substrate and the inhibitor. No assumption about sink conditions was done and the passive diffusion (CLPD) was estimated. The driving concentration for P-gp as well as the perpetrating concentration for P-gp inhibition was assumed to be the intracellular concentration. The impact of sampling on the concentrations measured was accounted for in the model.
Results:
The model was able to describe the disposition of digoxin and quinidine alone, and digoxin disposition in presence of quinidine .The geometric mean fold error (GMFE) between observed and model predicted digoxin concentrations was 1.29 and the geometric fold bias (GMFB) was 1.15. For quinidine, the GMFE was 1.16 and GMFB was 1.002. And finally the GMFE was 1.18 and GMFB was 1.06. Digoxin Km, Jmax and CLPD were estimated to 18 µM (relative standard error (RSE%): 41%), 252.8 pmol/min (RSE%: 34%), and 41 x 10-6 cm/sec (RSE%: 3%) respectively. Quinidine Km, Jmax, CLPD and Ki were estimated to 0.278 µM (RSE%: 44%), 11.3 pmol/min (RSE%: 37%), 201.2 x 10-6 cm/sec (RSE%: 6%) and 3.45 µM (RSE%: 21%) respectively.
Conclusions:
The model was able to estimate Jmax, Km, and CLPD for digoxin and quinidine with reasonable accuracy. The present data set would have not allowed to estimate a KiP-gp value using the conventional approach, however we were able to estimate a KiP-gp value for quinidine.
References:
[1] Korzekwa K, Nagar S. Compartmental models for apical efflux by P-glycoprotein: part 2–a theoretical study on transporter kinetic parameters. Pharm Res. 2014 Feb;31(2):335–46.
[2] Chaudhry A, Chung G, Lynn A, Yalvigi A, Brown C, Ellens H, O’Connor M, Lee C, Bentz J. Derivation of a System-Independent Ki for P-glycoprotein Mediated Digoxin Transport from System-Dependent IC50 Data. Drug Metab Dispos. 2018 Mar 1;46(3):279–90.
[3] Neuhoff S, Ungell A-L, Zamora I, Artursson P. pH-dependent bidirectional transport of weakly basic drugs across Caco-2 monolayers: implications for drug-drug interactions. Pharm Res. 2003 Aug;20(8):1141–8.
Reference: PAGE 28 (2019) Abstr 8808 [www.page-meeting.org/?abstract=8808]
Poster: Drug/Disease Modelling - Absorption & PBPK