Using the LeiCNS-PK3.0 Physiologically-Based Pharmacokinetic Model to Predict Brain Extracellular Fluid Pharmacokinetics in Mice

Abstract

Introduction: The unbound brain extracelullar fluid (brain_ECF) to plasma steady state partition coefficient, K_p,uu,BBB, values provide steady-state information on the extent of blood-brain barrier (BBB) transport equilibration, but not on pharmacokinetic (PK) profiles seen by the brain targets. Mouse models are frequently used to study brain PK, but this information cannot directly be used to inform on human brain PK, given the different CNS physiology of mouse and human. Physiologically based PK (PBPK) models are useful to translate PK information across species.

Aim: Use the LeiCNS-PK3.0 PBPK model, to predict brain extracellular fluid PK in mice.

Methods: Information on mouse brain physiology was collected from literature. All available connected data on unbound plasma, brain_ECF PK of 10 drugs (cyclophosphamide, quinidine, erlotonib, phenobarbital, colchicine, ribociclib, topotecan, cefradroxil, prexasertib, and methotrexate) from different mouse strains were used. Dosing regimen dependent plasma PK was modelled, and Kpuu,BBB values were estimated, and provided as input into the LeiCNS-PK3.0 model to result in prediction of PK profiles in brain_ECF.

Results: Overall, the model gave an adequate prediction of the brain_ECF PK profile for 7 out of the 10 drugs. For 7 drugs, the predicted versus observed brain_ECF data was within two-fold error limit and the other 2 drugs were within five-fold error limit.

Conclusion: The current version of the mouse LeiCNS-PK3.0 model seems to reasonably predict available information on brain_ECF from healthy mice for most drugs. This brings the translation between mouse and human brain PK one step further.

Keywords: brain; leiCNS-PK3.0; mouse; physiologically-based pharmacokinetics (PBPK).

Fig. 1

Visual predictive check plots evaluating the predictive accuracy of the mouse version of the LeiCNS-PK3.0 model. Ten drugs with different physicochemical properties and affinities to active transporters were used to evaluate the model predictions. The solid lines and colored band represent the median and 95% prediction interval, respectively, of the model’s prediction of the unbound pharmacokinetic profile at the plasma (red), brain extracellular fluid (yellow). The black dots represent the unbound drug concentrations measured in mice. Drugs were simulated with various routes of administrations: cyclophosphamide, quinidine and phenobarbital were intraperitoneal; erlotinib and ribociclib were orally; prexasertib was subcutaneous; colchicine, topotecan, cefadroxil and methotrexate were intravenously administered. Please note the different axes scales

Fig. 2

Box plot of the relative accuracy errors to evaluate the prediction accuracy of the current mouse version of the LeiCNS-PK3.0 model. The predictions of the ten drugs in plasma (red) and brain extracellular fluid (yellow) were evaluated using the relative accuracy errors. The green and yellow vertical lines represent two- and five-fold error limit, respectively. The predictions of methotrexate and prexasertib were beyond the two-fold errors but were within the five-fold error