Time-dependent inactivation of CYP3A4
To investigate the time-dependent inactivation of CYP3A4 by sunitinib, seven gradient concentrations (0–5 μM) and six time points (0–20 min) were used. The data were then fitted to a linear regression model, which reflected the linear relation between ‘ln remaining activity’ and ‘inactivation concentration’ (I ). The negative slope of this linear relationship reflected the observed inactivation rates (Kobs) value, which could be plotted against I to allow the fitting of inactivation kinetic parametersK I and K inact to the nonlinear least-squares regression based on Eq. 1. using Prism v.6.0 (GraphPad, San Diego, CA, USA).
(1)
Molecular docking simulations
The CYP2J2 crystal structure homology model from the Clustal Omega webserver (https://www.ebi.ac.uk/Tools/msa/clustalo/) was used to conduct docking simulations between TKIs and rivaroxaban in SYBYL (X-1.1) (Ning et al., 2019). The crystal structure of CYP3A4 (PDB: 4D7D) was from the crystal structures that bound to a known inhibitor. The 3D structures of the TKIs were subjected to energy minimisation using the default Tripos force field parameters, and the Gasteiger-Hückel charges were calculated for each compound. The Surflex-Dock mode was used to generate binding conformations of TKIs with CYP2J2. The optimal conformations were determined by their empirical functions ChemScore. The PyMOL Molecular Graphics System v.16.1.0.15350 (DeLano Scientific LLC) was used to visualise the docking results.
Quantitative prediction of DDI risk
Kinetic constants were included in the mechanistic static model to explore reversible inhibition and time-dependent inactivation. This static model was previously developed and refined by Fahmi et al. (Fahmi, Maurer, Kish, Cardenas, Boldt & Nettleton, 2008) and Isoherranen et al. (Isoherranen, Lutz, Chung, Hachad, Levy & Ragueneau-Majlessi, 2012) to account for the inhibition of multiple P450 isoforms. In the present study, this model was designed to explore the contributions of enzyme inhibition in the prediction of DDI risk. The area under the curve ratio (AUC Ratio/AUCR) in the presence of a pharmacokinetic DDI was used as the index, as described by Eq. 2.
(2)
Here, A is the time-dependent inactivation of each P450 isoform that was observed in the liver, as described by Eq. 3.
(3)(4)
Here, B is the reversible inhibition of each P450 isoform that was observed in the liver, as described by Eq. 4. The degradation rates (Kdeg) of CYP2J2 and CYP3A4 were 0.00026 and 0.00032 min–1, respectively (Cheong et al., 2017), where I represented the in vivo concentration of inhibitors in healthy and solid tumour patients. Additionally, the fraction of rivaroxaban metabolised by CYP2J2 or CYP3A4 was input from our previous study (Zhao et al., 2021), which were 0.95 for CYP2J2 and 0.025 for CYP3A4.