Results and Discussion

After i.v. administration of 1 mg/kg cobimetinib to WT (FVB), Cyp3a^−/−Tg-3A4_Hep, Cyp3a^−/−Tg-3A4_Int, or Cyp3a^−/−Tg-3A4_Hep/Int mice, CL ranged from 26 to 35 ml/min/kg, representing 35–50% of liver blood flow (blood-to-plasma ratio in mice, approximately 1.2) (Table 1 and Figure 1B). The data suggested that i.v. CL were not significantly different in WT, CYP3A4 transgenic, and Cyp3a^−/− animals (Figure 1B). In Cyp3a^−/− animals the CL of cobimetinib was 20.8 ml/min/kg (Table 1), suggesting the involvement of non-Cyp3a metabolizing enzymes in the CL of cobimetinib (based on ¹⁴C metabolite identification studies in rat, dog, and human, cobimetinib is also likely to be extensively metabolized in mouse) (Takahashi et al., manuscript in preparation). In contrast to the i.v. study where similar CL was observed across mice, after oral administration of 5 mg/kg cobimetinib the area under the curve (AUC) of cobimetinib in WT, Cyp3a^−/−Tg-3A4_Hep, Cyp3a^−/−Tg-3A4_Int, or Cyp3a^−/−Tg-3A4_Hep/Int was significantly different: 1.35, 3.39, 1.04, and 0.701 μM⋅h, respectively (Table 1 and Figure 1C). The up to 80% lower exposures of cobimetinib observed in animals with intestinal expression of CYP3A4 compared with mice with only hepatic expression suggested that intestinal CYP3A4 played a major role in the oral disposition of cobimetinib. This finding was similar to the observation with docetaxel (van Herwaarden et al., 2007) and triazolam (van Waterschoot et al., 2009a) where intestinal CYP3A4 had a major impact on oral exposure of these drugs (7.5- and 2-fold lower AUC in mice with intestine only versus liver only expression of CYP3A4 for docetaxel and triazolam, respectively).

In Cyp3a^−/−Tg-3A4_Hep/Int mice, the fraction escaping intestinal extraction (F_g) was estimated to be 0.23, where F = F_a⋅F_g⋅F_h F_h is the fraction escaping hepatic extraction and F = 0.14, F_h = 0.61, and F_a was assumed to be 1 (Table 1). In the clinical absolute F study, F was reported to be 46.2%, F_h = 0.87 (based on a CL of 11.7 l/h and liver blood flow of 20.7 ml/min/kg). In mouse, assuming a value of F_a of 1 (the F_a value in human was estimated to be 0.88) (manuscript in preparation), F_g is estimated to be 0.63 (Musib et al., 2013). Thus, the F, F_g, and F_h values observed in the CYP3A4 transgenic mice were quantitatively different from that in human. This was not unexpected because there are physiologic differences, e.g., blood flow, between species. However, if taking into account the involvement of the non-Cyp3a metabolism observed in CYP3a KO animals (subtracting the hepatic extraction in CYP3a KO mice, which was approximately 0.2), the F value in the transgenic animals would be higher and it would be in the range observed in human.

From in vitro incubations of cobimetinib in human liver microsomes, the WT, Cyp3a^−/−, and Cyp3a^−/−Tg-3A4_Hep/Int mouse liver microsome CL_int values were 173 ± 55, 248 ± 57, 129 ± 33, and 305 ± 60 μl/min/mg protein, respectively. The scaled CL_int values from WT and Cyp3a^−/−Tg-3A4_Hep/Int mouse liver microsome (incorporating plasma protein and microsomal binding; f_u plasma 0.036 and f_u microsome 0.48; determine by the Austin calculation) (Austin et al., 2002) followed the rank order of the in vivo CL observed in the respective mice (Table 1). These in vitro data infer that the 9-fold difference of PO exposure in WT, Cyp3a^−/−, and Cyp3a^−/−Tg-3A4_Hep/Int transgenic mice are likely to be from intestinal metabolism (Table 1) and not from hepatic CL.

Recently, the use of CYP3A4 transgenic mice to predict the CL of CYP3A4 substrates in human has been reported (Mitsui et al., 2014). Using 6 CYP3A4 substrates, the in vitro intrinsic CL from human and CYP3A4 transgenic mouse liver microsomes were reported to be within 2-fold of each other. However, the in vivo CL observed in transgenic mice could not be extrapolated directly to predict human CL. A prediction within 2-fold was only obtained from the regression of hepatic CL_int (back-calculated from the in vivo CL using the dispersion model) between human and transgenic mouse (Mitsui et al., 2014). Thus, a priori, a regression would have to be established for model compounds before human CL predictions could be made using data from transgenic mice. Although not discussed by the authors, it is worth noting that hepatic extraction derived from transgenic mouse and human were within 2-fold. It appears at this point that the direct utility of using transgenic mice in predicting human CL remains unclear.

To verify the relevance of the CYP3A4 transgenic mouse model we sought to determine if the circulating oxidative metabolites of cobimetinib (albeit minor; each <5% of total radioactivity in plasma) (Takahashi et al., manuscript in preparation) in human were also observed in the transgenic mice. Indeed, in the CYP3A4 transgenic mice, oxidative metabolites observed in human circulation (M12, M18, and M19) were observed (for structures, see Supplemental Fig. 1A). This provided supportive information that the metabolic pathways in the transgenic mice reflected those in human (which was CYP3A4 mediated) (R. Takahashi, manuscript in preparation). M12 and M19 were also observed in plasma of FVB and Cyp3a^−/− mice (see Supplemental Fig. 1B), the presence of human metabolites in the FVB and Cyp3a^−/− mice was not unexpected per se because mouse Cyp isoforms (Cyp3a or other Cyps) have the capability to metabolize and produce the same metabolites as human. However, it should be noted that in the transgenic mice with differential liver/intestinal expression of CYP3A4 there appeared to be quantitative and selective differences in metabolite profiles. The M12 levels were higher in transgenic mice expressing intestinal CYP3A4, and M18 was only observed in transgenic mice expressing liver CYP3A4. In addition, M18 was only observed where human CYP3A4 is expressed. Collectively, these data suggest that the transgenic mice provide a unique metabolite profile (compared with WT and Cyp3a^−/−), representative (at least qualitatively) of human metabolites (see Supplemental Fig. 1).

To further assess the utility of the transgenic mice, DDI studies were conducted in Cyp3a^−/−Tg-3A4_Hep/Int and in humanized PXR-CAR-CYP3A4/3A7 mice. In Cyp3a^−/−Tg-3A4_Hep/Int mice, inhibition of CYP3A4 by itraconazole resulted in an 8.3-fold increase in cobimetinib AUC and an approximately 3-fold increase in C_max after oral administration of cobimetinib (Table 2 and Figure 2A). In addition, i.v. administration of cobimetinib with itraconazole (oral) resulted in an approximately 4-fold increase in cobimetinib AUC and half-life; this is in keeping with a similar magnitude decrease in CL (Table 2 and Figure 2B). The greater DDI effect (approximately 8-fold versus approximately 4-fold) observed after oral dosing of cobimetinib was consistent with the contribution of intestinal CYP3A4 to the CL of cobimetinib. This finding was similar to the clinical observation reported where a greater DDI effect was observed after oral versus i.v. administration of midazolam with the CYP3A inhibitor clarithromycin, a 7- and 2.7-fold increases in the exposure (AUC) of midazolam were observed after oral and i.v. administration, respectively (Gorski et al., 1998).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Cobimetinib concentration-time profiles in Cyp3a^−/−Tg-3A4_Hep/Int mice after (A) oral administration of cobimetinib (5 mg/kg, n = 3) with or without co-administration with itraconazole; (B) after i.v. administration of cobimetinib (1 mg/kg, n = 3) with or without co-administration with itraconazole (log scale) and (C) after vehicle and rifampin treatment in PXR-CAR-CYP3A4/3A7 mice.

Itraconazole did not significantly alter the exposure of cobimetinib in Cyp3a KO animals, suggesting that in mice only the Cyp3a isoenzyme was inhibited by itraconazole (Table 3). In addition, itraconazole co-administration to FVB (WT) mice resulted in a 5.8-fold increase in oral exposure of cobimetinib (within the range of fold increase observed in transgenic mice) (Table 3). This is consistent with the suggestion that Cyp3a enzymes in mice may be the primary enzyme that metabolizes cobimetinib (Table 3). Collectively, these data suggest that the magnitude increase in CL of cobimetinib from co-administration with itraconazole to Cyp3a^−/−Tg-3A4_Hep/Int mice was higher than expected based on the apparent low CYP3A4 contribution to CL of cobimetinib (approximately 30% difference in CL between Cyp3a^−/− versus transgenic animals) (Table 1) and the specificity of itraconazole to inhibition of only Cyp3a; the reason for this higher than expected decrease in cobimetinib CL (when co-administered with itraconazole) is unclear. It should be noted that systemic exposure of itraconazole in these studies was similar to clinical exposures from a dose of 200 mg Sopranox solution (see the Sopranox Product Insert and Supplemental Fig. 2); however, the intestinal/hepatic exposures of itraconazole are unknown and may be higher than clinical tissue exposures based on the higher mg/kg dose.

In addition to studying the effect of CYP3A4 inhibition, the effect of induction was investigated using the PXR-CAR-CYP3A4/3A7 mice. It has been reported that the PXR-CAR-CYP3A4/3A7 model displays human-like CYP3A4 induction to various PXR inducers (Hasegawa et al., 2011). In addition to an increase in CYP3A4 mRNA levels after rifampin induction in PXR-CAR-CYP3A4/3A7 mice, in vivo exposures of triazolam were reported to be decreased by up to 90% after rifampin treatment (Hasegawa et al., 2011). Consistent with the role of CYP3A4 in the metabolism of cobimetinib, CYP3A4 induction resulted in an approximately 80% decrease in the exposure of cobimetinib (Table 2 and Figure 2C). As a reference, clinical induction of CYP3A from 600 mg of rifampin has been reported to decrease the exposure of the prototypical CYP3A substrate, midazolam by approximately 90%–98% (Backman et al., 1996; Gorski et al., 2003; University of Washington Drug Interaction Database). Thus, the observed effect of rifampin induction on cobimetinib in PXR-CAR-CYP3A4/3A7 mice was on the lower end of the scale compared with a sensitive CYP3A4 substrate such as midazolam. It is worth noting that the exposure of rifampin in this study was consistent with clinical concentrations observed from a dose of 600 mg (Kwara et al., 2014) (see Supplemental Fig. 2).

The utility of a specific transgenic model is dependent on the disposition and metabolism of the compound/drug of interest. For instance, since cobimetinib was predominantly metabolized by CYP3A4, and was metabolized by both intestinal and liver CYP3A, the CYP3A4 transgenic mice with differential expression in each tissue were used as an ideal model. It appeared that unlike midazolam, where mice Cyp2c metabolized midazolam and confounded the interpretation of data from Cyp3a KO mice (van Waterschoot et al., 2008), cobimetinib disposition and metabolism appeared to be mainly driven by Cyp3a/CYP3A in mice and human. However, some contribution from CYP (as evidenced by the Cyp3a KO data) and non-CYP enzymes such as UGT (f_m likely to be low) on the metabolism of cobimetinib cannot be excluded. One consideration when using this model is that for some CYP3A substrates, the contribution of CYP3A5 to the metabolism of a compound can be substantial (Lamba et al., 2002). While cobimetinib appears to be mainly metabolized by CYP3A4 (data on file), for midazolam, vincristine, and atanazavir, CYP3A5 has been reported to contribute to approximately 50%–80% of the overall f_m CYP3A (Walsky et al., 2012; Tseng et al., 2014). Therefore, data generated in the CYP3A4 transgenic mice, while representative of the *3/*3 genotype (with low/no CYP3A5 expression) may not reflect the population expressing CYP3A5 (*1/*3 or *1/*1). The allelic frequencies of CYP3A5*1/*3 in Caucasians and African Americans were reported to be approximately 17% and 40%, respectively, and for CYP3A5*1/*1, they were reported to be approximately 1% and 45% in Caucasian and African American subjects, respectively. Significant ethnic variability in allelic frequency has been reported (Xie et al., 2004). In addition, many compounds/drugs are metabolized by multiple enzymes and may be substrates of transporters; thus, compensation from deletion of one enzyme may result in a compensatory increase in another metabolizing enzyme or transporter. For instance, an increase in Cyp2c55 expression is reported with deletion of Cyp3a (van Waterschoot et al., 2009b). We had characterized some of the Cyp, Ugt, and transporter changes by reverse-transcription polymerase chain reaction and observed some changes in Cyp, transporter, and Ugt expression (compared with WT animals) in both hepatic and intestinal/duodenal tissue (see Supplemental Figs. 3 and 4). In most cases these changes were modest and unlikely to affect the overall interpretation of data for cobimetinib.

Overall, the findings from this study, albeit with one compound, suggest that the transgenic mouse model under specific conditions may be a useful tool to predict the relative contribution of hepatic and intestinal metabolism. Recently, the clinical interaction data of cobimetinb with itraconazole have been reported, a 6.7- and 3.2-fold increase in cobimetinib AUC and C_max, respectively, were observed (American Society of Clinical Pharmacology and Therapeutics Annual Conference, New Orleans, 2015). In addition, based on the observed DDI data with itraconazole, a PBPK model was developed and used to simulate the effect of rifampin induction (600 mg daily). An 83% decrease in cobimetinib exposure was predicted with rifampin induction. Comparison of the observed/predicted clinical DDI data to the effect of itraconazole and rifampin in the transgenic mouse models suggests that the transgenic model provided a good prediction of the clinical DDI effect. In the future, as more data using transgenic mice become available, particularly for drugs in which clinical DDI and or clinical disposition is well understood, it is anticipated that in tandem with in vitro data, PBPK modeling, and simulations, a clearer picture of the utility and limitations of the transgenic model will become evident.