Denise Feick (1), Donato Teutonico (2), and Henrik Cordes (1)
(1) Sanofi R&D, Drug Metabolism and Pharmacokinetics, Frankfurt am Main, Germany; (2) Sanofi R&D, Pharmacokinetics Dynamics and Metabolism Department, Vitry-sur-Seine, France
Introduction/Objectives: Oral drug bioavailability (Foral) can be expressed as the product of the (soluble) drug fraction absorbed from the gut lumen (Fa), the fraction of drug that escaped gut metabolism (Fg) and the fraction of drug that escaped hepatic metabolism (Fh) [1]. Fa is the net result between membrane permeation (passive and/or active) and the active efflux.
The Caco-2 permeability assay is a valuable tool in drug research and development to assess human intestinal drug permeation, where the in situ measurement is complex and expensive [3]. Amongst others, the apparent permeability coefficient (Papp), efflux ratio, and recovery can be extracted from the Caco-2 permeability assay [2]. In vitro to in vivo extrapolation (IVIVE) – the qualitative and quantitative translation of experimental data into physiological processes – is challenging, as the absorption via the human intestine is much more complex than it could be represented by a cell monolayer in Caco-2 assays. Additionally, drug permeation from the gut lumen into the intestinal blood circulation is affected by physicochemical compound properties (e.g., solubility, lipophilicity, and molecular weight), luminal properties (e.g., pH, osmolarity, food intake, and effective surface area) and active processes such as metabolism and transport.
The overall objective of this work is to extrapolate human Fa from Caco-2 derived Papp values by systematic investigation of various covariates. This was performed by:
- Establishment of a diverse benchmark dataset for human drug absorption parameters including Foral, Fa, Fg, Fh, physicochemical properties, metabolic and active transporter liability in the gut and the liver, Caco-2 determined in vitro properties such as Papp, and effective human intestinal permeability (Peff)
- Assessment of the predictivity for human Fa using Peff, Papp and calculated transcellular intestinal permeability (Pint,physchem) [5] in PK-Sim®
- Determination of the optimal transcellular intestinal permeability in PK-Sim (Pint,opt) to match reported Fa
- Evaluation of potential patterns related to compound covariates to translate Papp values into Pint,opt
Methods: Data acquisition and literature search was performed using Pubmed [6]. Calculation of Pint,physchem and Pint,opt was performed using PK-Sim® Version 11 [7] and R 4.2.0 [8] with the OSPsuite-R package [9].
Results: The compiled dataset comprised 138 compounds including their physicochemical properties, Papp and Peff values, ADME liabilities as well as information about human Foral, Fa, Fg and Fh [1]. Enzyme and transporter expression in gut and liver was informed by the PK-Sim expression database. The most prominent gut enzyme and transporter were CYP3A4 and P-glycoprotein with assignment to 38% and 39% of compounds, respectively. For further data analysis, we excluded compounds with reported values for Fa of 100%.
Correlation of Peff with Papp values improved for compounds that are only passively absorbed (R2 = 0.89) compared to correlation for the whole dataset (R2 = 0.62) as well as correlation of Fa with Caco-2 derived Papp (R2 = 0.64 vs. 0.43). Simulation of Fa in PK-Sim® by (1) using Pint,physchem or inserting (2) Peff values and (3) Caco-2 derived Papp values for transcellular intestinal permeability does not result in a good correlation with reported Fa values while methods 2 and 3 showed an overestimation of Fa for all compounds. Relationship of Fa and Pint,opt can be described by a sigmoidal Emax model. Calculation of the permeability ratio (Pint,opt/Papp) revealed a trend towards lower permeability ratios for metabolized and lipophilic compounds, i.e., overestimation of permeability with Caco-2 assays. For compounds where no gut metabolism or transport has been reported, no trend could be observed, implying that further compound properties impact intestinal permeability and need to be considered.
Conclusions: Translation of Caco-2 determined Papp values to human intestinal permeability is challenging, as Fa is dependent on passive as well as active processes in the body. Experimental procedures and reporting need to take active processes for each tested drug product into account. Evaluation of compound covariates on their contribution to the observed experimental readouts could help to improve IVIVE, also in combination with mechanistic modeling approaches. As an outlook, the benchmark dataset can help to inform feature selection by artificial intelligence (AI) models.
References:
[1] Varma et al. J Med Chem. 2010;53(3):1098-108. doi: 10.1021/jm901371v
[2] Hubatsch et al. Nat Protoc. 2007;2(9):2111-9. doi: 10.1038/nprot.2007.303
[3] Avdeef et al. J Med Chem. 2010 May 13;53(9):3566-84. doi: 10.1021/jm901846t
[4] Sun et al. Pharm Res. 2002;19(10):1400-16. doi: 10.1023/a:1020483911355
[5] Thelen et al. J Pharm Sci. 2011;100(12):5324-45. doi: 10.1002/jps.22726
[6] National Library of Medicine. https://pubmed.ncbi.nlm.nih.gov/
[7] Open Systems Pharmacology Suite. https://www.open-systems-pharmacology.org/
[8] R Core Team 2022. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/
[9] Sevestre et al. 2023. ospsuite: R package to manipulate OSPSuite Models. https://github.com/open-systems-pharmacology/ospsuite-r
Reference: PAGE 32 (2024) Abstr 10941 [www.page-meeting.org/?abstract=10941]
Poster: Drug/Disease Modelling - Absorption & PBPK