Attempting to Unmask the Inhibition of Sulfotransferase 1E1 in 17α-Ethinyl Estradiol Drug Interactions.

Drug–drug interactions (DDIs) involving combination oral contraceptives are complex, because the estrogenic (17α-ethinyl estradiol, EE) and progestin (e.g., drospirenone) components can both act as victim or perpetrator.1 In this regard, EE presents as a DDI victim because it is a sulfotransferase 1E1 (SULT1E1), UDP-glucuronosyltransferase 1A1 (UGT1A1), and cytochrome P450 3A4 (CYP3A4) substrate. This means that one needs to consider all 3 enzymes when attempting to understand EE DDI.1 As an example, Helmer et al reported that ziritaxestat (autotaxin inhibitor) significantly impacts the pharmacokinetic (PK) profile of EE.2 The inhibition of CYP3A4 was ruled out, as the DDI with drospirenone, a sensitive CYP3A4 substrate, was weaker (Table 1).3 Moreover, ziritaxestat did not affect plasma bilirubin concentrations and so the same authors also excluded UGT1A1 inhibition. It was concluded that the DDI with EE involved SULT1E1 after showing that it was inhibited by ziritaxestat in vitro (half-maximal inhibitory concentration, IC50 < 0.8 μM) (Table 1). Such an approach is logical, because SULT1E1 is expressed in the intestine and liver and plays a major role in EE first-pass metabolism.1, 4 Of note, ziritaxestat is not the first example of an EE DDI involving SULT1E1 inhibition. Schwartz et al have also presented a similar mechanism for etoricoxib (cyclooxygenase-2 inhibitor) after demonstrating SULT1E1 inhibition in vitro (Table 1).1, 5 Additionally, etoricoxib was considered a weak CYP3A4 inhibitor at the time of the EE DDI study.6, 7 Most importantly, the publications described above illustrate 2 possible approaches supporting the testing of the SULT1E1 inhibition hypothesis. Helmer et al subjected plasma samples to enzyme hydrolysis and determined the impact of ziritaxestat on circulating EE 3-O-glucuronide (±β-glucuronidase) and EE 3-O-sulfate (±arylsulfatase) levels.2 Following 18 days of ziritaxestat dosing, a 1.4-fold decrease was noted for the EE 3-O-glucuronide area under the plasma concentration versus time curve (AUC)/parent EE AUC ratio, compared with a more significant 5.6-fold decrease in the same ratio (EE 3-O-sulfate AUC/parent EE AUC) following arylsulfatase hydrolysis (Table S1). With the extensive EE first pass sulfation, and the known high levels of EE 3-O-sulfate in circulation,8 Schwartz et al were able to directly measure EE 3-O-sulfate levels in plasma and show that they were decreased by approximately 40% (Table 1).5 Based on changes in the EE 3-O-sulfate AUC/parent EE AUC ratio, it is possible to estimate the percent SULT1E1 inhibition in vivo as 82% and 60% for ziritaxestat and etoricoxib, respectively. Such values do not distinguish inhibition in the gut versus liver (Table 1). For both etoricoxib and ziritaxestat, it was possible to generate Pfizer in-house in vitro SULT1E1, UGT1A1, and CYP3A4 IC50 data and obtain an inhibition signature for each compound (Figure 1; Table S2). It should be noted that EE has been described as a P-glycoprotein (Pgp) substrate in vitro, although its role in vivo has not been confirmed.9-11 Even so, etoricoxib was found to be a weak Pgp inhibitor (vs ziritaxestat), consistent with its minimal effect on the PK of digoxin (Table S3). Although ziritaxestat presented as a potent Pgp inhibitor in vitro, we are not aware of any published clinical data for ziritaxestat with Pgp probe drugs. As described in Figure 1, it was possible to predict the EE AUC ratio (AUCR = AUCperpetrator/AUCreference) for etoricoxib and ziritaxestat (ignoring Pgp inhibition). Although a physiologically based PK (PBPK) model for EE has been described,12 in this instance predictions were based on a simple static model that incorporated the estimated percentage inhibition of UGT1A1, SULT1E1, and CYP3A4 in the gut and liver (Table S2), the assigned EE fraction metabolized by each enzyme in both organs, as well as the fraction of the dose surviving gut first pass (Table S4). The model was validated using etoricoxib, as it is not a Pgp inhibitor in vivo. Interestingly, it was estimated that etoricoxib is a weak inhibitor of liver (2%) versus gut (48%) CYP3A4, consistent with its minimal effect on the intravenous [N-14C-methyl]-erythromycin breath test (ERBT).7 The relatively weak inhibition of liver (vs gut) SULT1E1 and UGT1A1 by etoricoxib might also explain its minimal effect on the plasma half-life (vs the maximal plasma concentration, Cmax) of EE.5 In the case of ziritaxestat, the EE AUCR was underestimated (with a value of 2.0) compared with the observed AUC geometric mean ratio (GMR) of 2.4 and 90% confidence interval (90%CI) of 2.2–2.6 (Table 1). Whether or not the underestimate is the result of not accounting for Pgp inhibition is unknown. Compared with etoricoxib, ziritaxestat was a more potent inhibitor of hepatic SULT1E1 (45% vs 17%), which may partly explain why Helmer et al documented an increase in both EE plasma Cmax and half-life.2 The model was also able to predict the AUCR of drospirenone following ziritaxestat administration (Table S4). With some assumptions, the model also predicted no effect on plasma bilirubin, consistent with clinical data.2 However, a greater inhibition of gut (vs liver) UGT1A1 is predicted for ziritaxestat (75% vs 2%; Figure 1), and Helmer et al did report a 1.4-fold decrease (predicted to represent 29% inhibition) in the plasma EE 3-O-glucuronide AUC/parent EE AUC ratio (Table 1). It is possible that bilirubin and EE present differently as UGT1A1 substrates, if in fact the fraction surviving gut first pass (fg) is lower for EE versus bilirubin (EE fg = 0.56 vs bilirubin fg = 1; Table S4). Of note, the omission of intestinal UGT1A1 inhibition in the static model led to an even lower estimate of the EE AUCR (1.8) versus the observed AUC GMR and 90%CI (Figure 1). This implies that there is some contribution from the inhibition of gut UGT1A1. It should also be noted that the plasma EE glucuronide/parent EE ratio may be impacted by SULT1E1 inhibition. Specifically, the inhibition of SULT1E1 increases the parent EE AUC, thus decreasing both the EE sulfate and glucuronide/parent EE ratios. In such a scenario the inhibition of UGT1A1 would be wrongly implicated. Therefore, an assessment of UGT1A1 inhibition necessitates the use of orally administered UGT1A1 probe drugs (Figure 2). To try and identify additional SULT1E1 inhibitors, the University of Washington DDI database (https://www.druginteractionsolutions.org) was leveraged using EE and oral midazolam (MDZ) DDI data as filters (Table S5). Of the drugs identified, rofecoxib and rucaparib were chosen because they presented as weak CYP3A4 inhibitors and rendered differentiated EE AUC GMR (90%CI) values of 1.13 (1.06 to 1.19) and 1.4 (1.2 to 1.8), respectively (Figure 1).13, 14 Neither compound greatly impacted the PK of digoxin, and EE AUCR predictions were successful in both cases (Tables S6–S9). It was noted that liver CYP3A4 was weakly inhibited by rofecoxib (≤2%), consistent with its minimal impact on the ERBT (Table S5). Rucaparib presents the inhibition of both liver (21%) and gut (26%) CYP3A4, which supported a successful prediction of MDZ AUCR (Table S9). Of note, a third compound (teriflunomide) presented EE and MDZ AUC GMR values of 1.5 and 1.2, respectively (Table S5). But, in this instance, only very weak inhibition was observed for the 3 enzymes in vitro (IC50 ≥ 100 μM, data not shown). For SULT1E1, the weak inhibition could be rationalized using computer-aided enzyme active site docking (Figure S1). Therefore, the mechanism(s) underpinning the teriflunomide DDI with EE cannot be explained based on the present data set. In conclusion, one can attempt an integrated approach to assess the role and importance of SULT1E1 (vs CYP3A4 and UGT1A1) in EE victim DDI (Figure 2). Such an approach involves the integration of clinical probe data (e.g., MDZ AUCR, drospirenone AUCR, and plasma bilirubin analysis), the profiling of metabolites in plasma from an EE DDI study (e.g., direct measurement of EE 3-O-sulfate or enzyme hydrolysis), and the generation of in vitro IC50 values (e.g., SULT1E1, CYP3A4, and UGT1A1) to support PK modeling exercises. For example, one may decide that only EE AUC GMR values above the prespecified boundaries for no DDI (AUC GMR = 0.8 to 1.25), with no impact on progestin or CYP3A4 probe PK, should trigger investigative studies, as described by Helmer et al and Schwartz et al.2, 5 Alternatively, an in vitro CYP3A4–UGT1A1–SULT1E1 inhibition signature can be proactively generated for a new chemical entity and the data used as input to support EE DDI modeling based on its projected or known PK. It is also concluded that although the development of both etoricoxib and ziritaxestat has been terminated, the data described herein are very informative and further reinforce the importance of assessing enzymes other than CYP3A4 when investigating EE DDI mechanisms.1, 2 At the same time, it is acknowledged that the PK and disposition profile of EE (as a CYP3A4, SULT1E1, and UGT1A1 substrate) and its conjugated metabolites (as transporter substrates) is complex, which warrants careful interpretation of parent EE AUCR and metabolite/parent ratios in plasma or excreta, and the further refinement of static and PBPK-based EE DDI models.1, 2, 8, 12, 15 The authors would like to thank Zheng Yang, PhD (Alnylam Pharmaceuticals), for the derivation of the static model described in Table S4. The authors declare that they have no competing interests associated with this work. No funding was received for this work. As well as the provided supplemental materials, the authors can provide additional data upon request. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.