Predicting human oral bioavailability in vitro: a combined primary human gut/liver microphysiological system and mechanistic modelling approach

Evaluating oral bioavailability (F) is crucial for understanding the pharmacokinetics of new drug candidates. However, animal models often fail to accurately predict human oral bioavailability, as shown in a seminal study involving 184 compounds (R2 = 0.34).

To enhance the translation of drug efficacy and safety data from in vitro to in vivo, microphysiological systems (MPS) have emerged, capable of fluidically connecting organs essential for modeling human drug ADME.

Here, we introduce a Dual-organ MPS with six wells, each containing two compartments – a Transwell®-based primary intestinal epithelial monolayer (RepliGut® Planar) and a 3D liver microtissue (primary human hepatocytes).

This Gut/Liver MPS, which simulates first-pass metabolism, is combined with a mechanistic mathematical model to estimate human oral bioavailability and its components: the fraction absorbed (Fa), the fraction escaping gut wall elimination (Fg), and the fraction escaping hepatic elimination (Fh).

We investigated the pharmacokinetics of midazolam because both the gut and liver play well-established roles in its metabolism. We then compared this to a Gut/Liver MPS model that uses Caco-2 cells in the gut model.

After combining and fluidically coupling the gut and liver tissues in the Dual-organ MPS, we simulated oral and intravenous (IV) drug delivery routes. We quantified midazolam and its primary metabolite, 1’-hydroxymidazolam, by performing LCMS on the media samples collected from the liver and gut apical compartments over a 72-hour period.

A 2-compartment mechanistic model was developed to describe the changes in parent and metabolite concentrations. We fitted the model to the LCMS data and used Bayesian inference to investigate the parameter space, determining the upper and lower bounds for various ADME parameters, including intrinsic clearance of the gut and liver, and apparent permeability.

We estimated Fa, Fg, Fh, and F from model simulations of experimental data and established scaling methods using model parameter estimates.

The model estimation of F was found to be 0.53 for the primary Gut/Liver MPS, which aligns with the range observed in humans (0.096 to 0.68, with a mean of 0.31(2)). In contrast, F was estimated to be 0.66 for the Caco-2/Liver MPS.

Using a primary gut model improved the estimation of F, as no metabolism was observed in Caco-2 cells, leading to a higher prediction of Fg compared to the primary Gut/Liver MPS.

Here, we present a Dual-organ MPS capable of simulating oral and IV drug delivery routes. When combined with mathematical modelling, using midazolam, we demonstrate the ability to mechanistically profile and to extract multiple ADME parameters including oral bioavailability and its components, Fa, Fg and Fh from a single experiment.