PBTK model-based analysis of CYP3A4 induction and the toxicokinetics of the pyrrolizidine alkaloid retrorsine in man

Abstract

Cytochrome P450 (CYP)3A4 induction by drugs and pesticides plays a critical role in the enhancement of pyrrolizidine alkaloid (PA) toxicity as it leads to increased formation of hepatotoxic dehydro-PA metabolites. Addressing the need for a quantitative analysis of this interaction, we developed a physiologically-based toxicokinetic (PBTK) model. Specifically, the model describes the impact of the well-characterized CYP3A4 inducer rifampicin on the kinetics of retrorsine, which is a prototypic PA and contaminant in herbal teas. Based on consumption data, the kinetics after daily intake of retrorsine were simulated with concomitant rifampicin treatment. Strongest impact on retrorsine kinetics (plasma AUC ${}_{24}$ and $C_{\text{max}}$ reduced to 67% and 74% compared to the rifampicin-free reference) was predicted directly after withdrawal of rifampicin. At this time point, the competitive inhibitory effect of rifampicin stopped, while CYP3A4 induction was still near its maximum. Due to the impacted metabolism kinetics, the cumulative formation of intestinal retrorsine CYP3A4 metabolites increased to 254% (from 10 to 25 nmol), while the cumulative formation of hepatic CYP3A4 metabolites was not affected (57 nmol). Return to baseline PA toxicokinetics was predicted 14 days after stop of a 14-day rifampicin treatment. In conclusion, the PBTK model showed to be a promising tool to assess the dynamic interplay of enzyme induction and toxification pathways.

Keywords: Bioactivation; Drug–drug interaction; Enzyme induction; PBPK model; Rifampicin.

Fig. 1

a Physiologically-based toxicokinetic (PBTK) model of retrorsine. The liver compartment was represented by the extended clearance model accounting for hepatic transport and metabolism. b Empirical 2-compartment model of rifampicin parametrized for liver (intracellular) and small intestine (intracellular), separately. c Retrorsine-rifampicin interaction mediated by CYP3A4 induction and competitive CYP3A4 inhibition by rifampicin. All parameters of the retrorsine-rifampicin interaction model are summarized in Tables S1–S2. The full system of ordinary differential equations is given in Eqs. S1–S15. CL$_{\text{act,in}}^{}$ Active uptake clearance; CL$_{\text{act,ef}}^{}$ Active efflux clearance; CL$_{\text{bile}}^{}$ Biliary clearance; CL$_{\text{met,gut}}^{}$ Gut metabolic clearance; CL$_{\text{met,liv}}^{}$ Liver metabolic clearance; CL$_{\text{met,tis}}^{}$ Liver or gut metabolic clearance; CL$_{\text{r}}^{}$ Renal clearance; $F_{\text{a}}$ Intestinal fraction absorbed; $f_{\text{m},\text{CYP}3A4}$ Fraction of retrorsine metabolized by CYP3A4; k_12,tis Transition rate constant from central to peripheral compartment for rifampicin; k_21,tis Transition rate constant from peripheral to central compartment for rifampicin; $k_{\text{a}}$ Intestinal absorption rate constant; $k_{\text{depot,tis}}$ Absorption rate constant from depot compartment for rifampicin; $k_{\text{e,tis}}$ Elimination rate constant for rifampicin; PS$_{\text{diff}}^{}$ Passive influx diffusion flow rate; tis = liver cellular space or small intestine (both cellular space)

Fig. 2

a Temperature-dependent retrorsine (RET) depletion in incubations of HepaRG cells determined by Enge et al. (2021). Medium loss assays were performed either at 4 $_{}^{\circ }$C or at 37 $_{}^{\circ }$C. Observed data ($n=4$ biological replicates, each assessed with $n=2$ technical replicates) were described with a monoexponential model (Eq. S16). Median (solid lines) and 5–95% interpercentile range (shaded areas) are based on 1000 Monte Carlo simulations. In vivo liver active uptake CL$_{\text{act,in}}^{}$ and passive diffusion PS$_{\text{diff}}^{}$ were predicted for non-saturating conditions (Eqs. S17–S20). b Concentration-dependent RET depletion in human liver microsomal incubations. Microsomal assays were performed with 1, 15, 50 or 200 $\mathrm{\mu }$M of retrorsine (RET$_0^{}$). Observed data ($n=2$ technical replicates using microsomes of pooled livers) were described by an end-product inhibition model (Eqs. S21–S24). Median (solid lines) and 5–95% interpercentile range (shaded areas) are based on 1000 Monte Carlo simulations. In vivo liver metabolic clearance CL$_{\text{met,liv}}^{}=V_{\text{max,liv}}/K_{\text{M,liv}}$ was predicted for the case of linear kinetics (RET $\ll K_{\text{M,liv}}$)

Fig. 3

Predicted effect of a two-week oral intake of rifampicin (600 mg/day) on the kinetics of retrorsine during consumption of herbal tea with a retrorsine dose of 0.019 $\mathrm{\mu }$g/kg body weight/day for 5 weeks. a Concentration-time profiles of retrorsine in liver and gut tissue. b Daily formation of hepatic and intestinal CYP3A4 and non-CYP3A4 metabolites of retrorsine. c Time-dependent formation of hepatic and intestinal CYP3A4 and non-CYP3A4 metabolites of retrorsine given as cumulative amount. Note: Day 1, 3 and 14 after the first dose of rifampicin (CYP3A4 induction) are marked by $\ast$. Day 2, 6 and 14 after the last dose of rifampicin (CYP3A4 de-induction) are marked by #

Fig. 4

Sensitivity analysis of interaction model parameters. The sensitivity coefficient indicates the effect of parameter reduction by 10% ($-$) and parameter increase by 10% (+) on the cumulative amount of retrorsine CYP3A4 metabolites in liver (blue color) and gut (orange color) at 5 weeks of model simulation. Abbreviations: $f_{\text{m,CYP3A4}}$ = fraction of retrorsine metabolized by CYP3A4, EC$_{50}^{}$ = rifampicin concentration at half-maximal CYP3A4 induction in vivo, $E_{\text{max}}$ = maximal CYP3A4 induction effect of rifampicin in vivo, $K_i$ = rifampicin concentration at half-maximal CYP3A4 inhibition (color figure online)