Predictions of mechanism-based drug interactions (MBDDIs) have historically been based on in-vitro parameters derived in human liver microsomes (HLMs). Typically, these parameters are kinact, the maximum inactivation rate constant and KI, the concentration that results in the half maximum inactivation rate constant describing the mechanism based inactivation (MBI) of the enzyme. When predicting these MBDDIs, a critical assumption is made. This assumption states that the intrahepatocyte unbound concentration of the perpetrator drug is approximated by the extracellular unbound plasma concentration of the drug, i.e. minimal net transport of the perpetrator drug into or out of the hepatocytes. Since drug transporters are expressed at the sinusoidal and canalicular membrane of the hepatocytes, this assumption is unlikely to be true. When cultured appropriately, sandwich-cultured human hepatocytes (cSCHHs) are able to reestablish bile canalicular networks and to express many of the transporters (e.g. P-glycoprotein; P-gp) expressed in the liver. Therefore, we selected 8 CYP3A-based MBIs which have resulted in either over or underprediction of in vivo DDIs, and asked the question “Do cSCHHs better predict these in vivo CYP3A MBDDIs than do HLMs? In addition we asked if the differences in MBI parameters determined in HLMs and cSCHHs is a result of efflux transporters (e.g P-gp). CYP3A inactivation parameters (KI and kinact) were determined in HLMs (n=3) or in cSCHHs, (n=3; in the presences or absence of elacridar) for the following MBIs: ritonavir, nelfinavir, amprenavir, lopinavir, verapamil, troleandomycin, diltiazem or erythromycin using midazolam 1’-hydroxylation as the CYP3A measure. kinact was significantly lower in cSCHHs compared to HLMs for all 8 inactivators, ranging from ~2-fold (diltiazem) to ~130-fold (amprenavir). KI, was ~8-10-fold lower in HLMs compared to cSCHHs for lopinavir and nelfinavir, whereas the KI for verapamil was ~8-fold higher in HLMs compared to cSCHHs. P-gp inhibition by elacridar resulted in more potent inactivation of CYP3A by ritonavir and troleandomycin, but had no effect on amprenavir, lopinavir, verapamil or erythromycin. Previously published MBDDIs between the MBIs listed above and midazolam were predicted using a comprehensive mechanistic static prediction model with kinact and KI parameters generated in HLMs or cSCHHs. Midazolam DDIs were more accurately predicted for ritonavir (high dose only) and nelfinavir using cSCHHs. Diltiazem, verapamil and erythromycin mediated DDIs were in general slightly overpredicted by HLMs but slightly underpredicted by cSCHHs. Troleandomycin mediated DDIs were drastically underpredicted by both HLMs and cSCHHs. CYP3A inactivation was different in HLMs and cSCHHs, and P-gp played a significant role in this discrepancy for ritonavir and troleandomycin. However, for the remaining MBIs, other mechanisms must play a role in differences in CYP3A inactivation between HLMs and cSCHHs. Nevertheless, cSCHHs provide comparable, if not better, prediction accuracy of in vivo DDIs by the MBIs evaluated.
Supported by Simcyp Ltd., an ARCS fellowship and NIH grants GM07550 and GM032165.