Elena Righetti1,2, Aleksandr Petrov3,4, Ralf Engbert5, Daniel Schad6, Charlotte Kloft3,7, Andreas Reichel8, Wilhelm Huisinga3,4
1Fondazione the Microsoft Research-University of Trento Centre for Computational and Systems Biology (COSBI), 2Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, 3Graduate Research Training Program PharMetrX, 4Institute of Mathematics, University of Potsdam, 5Department of Psychology, University of Potsdam, 6Institute for Mind, Brain and Behavior, HMU Health and Medical University, 7Department of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universität Berlin, 8Preclinical Modeling and Simulation, Preclinical Development, Bayer, Bayer AG
Objectives: Unbound brain exposure is a critical factor in central nervous system (CNS) drug development, which requires information on the total brain exposure and the fraction unbound in the brain. Mechanistic models such as the Rodgers & Rowland (R&R) methods [1,2] allow to predict brain interstitial and cellular partitioning a priori, ultimately predicting fractions unbound in the tissue and its subspaces. Unlike other tissues, the brain is separated from circulation by the blood-brain barrier (BBB) and the brain-cerebrospinal fluid barrier (BCSFB), which restrict rapid exchange between interstitial water and plasma or cerebrospinal fluid (CSF), respectively. As a result, quantifying experimentally tissue partitioning often involves the fraction unbound in brain tissue (fuT) rather than the tissue-to-plasma water partition coefficient. Reliable fuT values should reflect in vivo brain conditions, where membranes and pH gradients are present. However, the parameter is often estimated by in vitro brain homogenate measurements. Standard a-priori methods, in turn, do not account for biases introduced by experimental protocols: in brain homogenate protocols, compounds are typically spiked into homogenized tissue, leading to disruption of membranes and pH gradients. By addressing these factors, our work aims to provide the means for a more realistic assessment of a-priori predictions of the fraction unbound in brain homogenates of small drug molecules. Methods: We adapted the R&R methods to explicitly account for cell membrane and pH gradient disruption. Brain homogenates were assumed to comprise of the same brain tissue components as the intact tissue, but without the membrane barrier and pH gradient between interstitial and cellular spaces; homogenate pH was calculated as a weighted sum of interstitial, cellular, and buffer pH values, depending on the dilution factor of the experimental protocol. Physiological and anatomical parameters, including brain tissue composition and compartment volumes, were sourced from the literature [3–6]. The method was evaluated against fraction unbound values [7] measured by equilibrium dialysis on brain homogenates [8,9]. Despite some experimental pitfalls, this is a standard technique in drug discovery settings due to its cost-effectiveness and high-throughput capability [3]. We included equilibrium dialysis data from human and rodent brain homogenates, assuming the commonly accepted species-independence of brain fraction unbound [9]. We selected ~20 small molecules that primarily cross the BBB via passive diffusion, ensuring a balanced representation of acidic, basic, and neutral compounds. Drug physicochemical parameters were sourced from GastroPlusTM software and the literature [1,2,9–11]. Results: In our homogenate R&R method derivation, we defined the fraction unbound in tissue homogenate (fuH) as the appropriate quantity for comparison with experimental homogenate data, fundamentally differing from the standard PBPK parameter fuT. Our fuH predictions fell within a 3-fold deviation margin from experimental values for ~80% of compounds – compared to ~68% from the original R&R method. Comparing fuT and fuH predictions further highlighted compounds most affected by homogenization, including metoclopramide and citalopram, with relative deviations from fuT of 0.85 and 0.50, respectively, and a mean relative deviation of 0.26 across all compounds. Key factors influencing measured values in brain homogenate assay predictions are buffer pH and dilution factor. At sufficiently high dilution, homogenate pH approaches buffer pH (typically ~7.4), resembling plasma pH. This shift significantly impacts drug distribution into cellular components, particularly for moderate-to-strong bases and type I zwitter ions. Deviations from experimental data likely stem from uncertainties in input parameters, assay variability, or species-dependent differences in brain tissue composition. Additionally, residual blood can bias a-priori predictions, particularly for compounds with strong plasma protein binding. To correct for overestimation, we performed a sensitivity analysis on residual blood fractional volume and incorporated its impact on fraction unbound predictions. Conclusion: Our homogenate R&R-based approach improved alignment with experimental measurements. More extensive validation against brain homogenate data is a necessary step to increase confidence or improve fraction unbound predictions in intact brain tissue for CNS small molecules.
1. Rodgers, T., Leahy, D. & Rowland, M. J Pharm Sci 94, 1259–1276 (2005). 2. Rodgers, T. & Rowland, M. J Pharm Sci 95, 1238–1257 (2006). 3. Reichel, A. Blood-Brain Barrier in Drug Discovery 5–41 (Wiley, 2015). 4. ICRP Publication 89. Ann ICRP 32, 5–265 (2002). 5. Poulin, P. & Theil, F. J Pharm Sci 91, 129–156 (2002). 6. Gaohua, L. et al. Drug Metab Pharmacokinet 31, 224–233 (2016). 7. Kalvass, J. C., Maurer, T. S. & Pollack, G. M. Drug Metabolism and Disposition 35, 660–666 (2007). 8. Summerfield, S. G. et al. Xenobiotica 38, 1518–1535 (2008). 9. Di, L. et al. Drug Metabolism and Disposition 39, 1270–1277 (2011). 10. Mamada, H. et al. Mol Divers 25, 1261–1270 (2021). 11. Rodgers, T. & Rowland, M. Pharm Res 24, 918–933 (2007).
Reference: PAGE 33 (2025) Abstr 11644 [www.page-meeting.org/?abstract=11644]
Poster: Drug/Disease Modelling - CNS