Using Simcyp Designer to extend a PBPK-QSP Model of Warfarin to Target Binding
Tariq Abdulla (1), Masoud Jamei (1), Frederic Bois (1)
(1) Certara UK, Simcyp Division
Objectives: CYP2C9 and VKORC1 variant genotypes are associated with increased sensitivity to warfarin. CYP2C9 mediates the majority of the clearance of the more potent enantiomer S-warfarin. The high affinity of warfarin for its target VKORC1 is a cause of target mediated drug disposition (TMDD). Higher warfarin sensitivity in patients with VKORC1 AA and AG genotypes seem to be caused by differences in PD rather than PK [1].
The effect of variant genotypes have been estimated as covariates in a number of population PK/PD models; for example with the effect of CYP2C9 genotype on clearance, and the effect of VKORC1 genotype on EC50 or concentration-effect slope [1,2,3].
A published Simcyp PBPK model of S- and R-warfarin explains population variability in exposure due to CYP2C9 activity and the OAT2 hepatic uptake transporter [4]. Differences in exposure emerge from the enzyme abundance for different genotypes, and genotype frequencies, in a population.
Here, we extend the PBPK model for R- and S- warfarin to include binding models, first within hepatocytes, and then within multiple tissue compartments. The aim was to evaluate whether an extended model could simultaneously account for TMDD observed with a microdose (0.1mg) of rac-warfarin [5], the differential potency of the enantiomers, and increased warfarin sensitivity of CYP2C9 and VKORC1 variant genotypes.
Methods: The racemic warfarin PBPK model was based on [4] and represents the enantiomers as separate compounds. The enantiomer models are distinguished by their intrinsic clearance rates by hepatic enzymes, and by uptake into hepatocytes by the OAT2 transporter.
The PBPK model was imported into Simcyp Designer. The binding model was developed and initially connected to the liver intracellular concentrations for R- and S-warfarin. A two binding sites dynamic binding model was added, with R-warfarin having a 6-fold higher Kd for site 1 [6]; consistent with the 2 to 5 fold greater potency of the S- enantiomer.
The inital model was able to capture the TMDD observed for a rac-warfarin microdose. However, this required scaling the Kd parameters to picomolar. It was thought that this might reflect processes not included in the model, such as binding in other tissues. While VKORC1 expression is highest in the liver, collectively other tissues could have a significant impact on the TMDD. Therefore, the model was further extended to include dynamic binding in the brain, lungs, kidney, spleen, muscle, gut, and pancreas.
Finally, receptor occupancy in hepatocytes was used as input to a turnover PD model of prothrombin complex activity (PCA). International normalised ratio (INR) was taken to be 1/PCA, and used as the PD endpoint. Total hepatocyte receptor occupancy was assumed to inhibit the synthesis of PCA.
Results: The extended model was able to capture the observed TMDD effect. VKORC1 concentration was set 5 fold higher in the liver than in other tissues (an approximation based on mRNA expression). A concentration of 25uM in the liver was selected providing a reasonable fit to observed data. Predicted Cmax and AUClast were within 2-fold of observations, both for microdose and therapeutic dose data. Predicted AUC of the microdose was ~3-fold higher than observed when the binding models were not included.
The PD model was able to reproduce single dose and steady state changes in INR [1]. The model was sensitive to VKORC1 concentrations. A 3-fold reduction in VKORC1 concentration (based on relative mRNA in the A/A genotype [7]) increased the mean predicted INR at 20 days from 2.41 to 2.79 for a CYP2C9 1*/1* genotype. This suggests a potential explanation for the higher sensitivity in this genotype, with a lower VKORC1 expression. VKORC1 expression did not have a significant effect on warfarin PK, consistent with observed data.
Conclusion: By incorporating target binding within multiple tissues, the extended model was able to capture the TMDD of warfarin, that is especially significant at a low dose. Further, using target occupancy within hepatocytes as an input to a PD model allowed for the same model to be linked to effects on both PK and PD.
The model would benefit from further verification for the effect of each enantiomer separately, and for the predicted dose response of various genotype combinations. The Simcyp Designer framework for extending PBPK models with mechanistic tissue binding and pharmacodynamic effects is readily applicable to other compounds.
References:
[1] Ferrari, M. et al. (2017). Eur J Clin Pharmacol, 73(6), 699-707.
[2] Hamberg, A. et al. (2007). CPT, 81(4), 529–538.
[3] Xue, L et al. BJCP, 83(4), 823–835.
[4] Bi Y. A. et al. (2018). Molecular Pharmaceutics, 15(3), 1284–1295.
[5] Lappin G. et al. Clin Pharmacol Ther. 2006 Sep;80(3):203-15.
[6] Lewis, B. C. et al. (2016). Pharmacogenetics and Genomics, 26(1), 44–50.
[7] Rieder, M. J. et al. (2005). NEJM, 352(22), 2285–2293.