Reports continue to appear describing drug interactions involving oral contraceptive (OC) formulations containing 17α-ethinyl estradiol (EE). For example, Hilli et al. (2008) reported recently that melatonin (MEL) area under the plasma concentration versus time curve (AUC) is increased (∼5-fold), and the 6-hydroxy melatonin (MEL 6-OH)/MEL AUC ratio is decreased (88%). CYP1A2 is known to play a major role in the metabolism of MEL in human liver microsomes (HLM), so the authors deduced that the enzyme was the locus of the interaction (Facciolá et al., 2001; Härtter et al., 2001; Ma et al., 2005). EE-containing OCs exert a similar effect on the pharmacokinetics (PK) of tizanidine (Granfors et al., 2005). Both tizanidine and MEL are low oral bioavailability (<25%) CYP1A2 substrates because of first-pass metabolism (Härtter et al., 2001; Granfors et al., 2005). Although caffeine and theophylline also serve as CYP1A2 substrates, they undergo minimal first pass, are highly bioavailable, and the impact of EE is less marked (∼40% decrease in clearance) (Roberts et al., 1983; Balogh et al., 1995). Despite these clinical data, it is only very recently that assessment of CYP1A2 inhibition in vitro has been described previously (Karjalainen et al., 2008). In fact, Karjalainen et al. (2008) reported EE as a relatively weak inhibitor of phenacetin O-deethylation (POD) activity in HLM (low K_m component; IC₅₀ = 24 μM).

Drug interactions with EE-containing OC formulations have also included a number of CYP2C19 substrates, such as omeprazole, mephenytoin, and proguanil (Hägg et al., 2001; Rodrigues and Lu, 2004, references therein; Shelepova et al., 2005). Likewise, drug interactions with imipramine and selegiline have been described previously (Abernethy et al., 1984; Laine et al., 1999). The latter is a CYP2B6 and CYP2C19 substrate, also with a low oral bioavailability (<10%), and greater than 10-fold increases in AUC have been reported with OCs (Laine et al., 1999; Benetton et al., 2007). More importantly, the contribution of CYP2C19 after oral dosing of selegiline is thought to be minimal (Laine et al., 2001). Although CYP2B6 is implicated, there are no reports describing the impact of CYP2B6 genotype (or phenotype) on the oral PK of selegiline, and its contribution in vivo is not known. Moreover, the clinical drug interaction between EE and a CYP2B6 probe (bupropion) is not significant despite evidence for mechanism-based inhibition in vitro (Kent et al., 2002; Palovaara et al., 2003). The same can also be said for the observed mechanism-based inhibition of CYP3A4 and CYP3A5 in vitro because the effect of EE on the PK of CYP3A substrates (e.g., midazolam and nifedipine) is minimal (Balogh et al., 1998; Palovaara et al., 2000; Belle et al., 2002; Lin et al., 2002; Atkinson et al., 2005; Shelepova et al., 2005; Lin and Hollenberg, 2007).

Overall, available clinical data suggest that EE exerts a differential inhibitory effect across the various human cytochromes P450 (P450s). Such a hypothesis is supported by the data of Shelepova et al. (2005), who conducted an EE interaction study using the Cooperstown (“5 + 1”) drug mixture. In this instance, EE did not affect the PK of the CYP2C9, CYP2D6, and CYP3A4 probes, whereas the effect on the putative CYP1A2 and CYP2C19 trait measures was statistically significant. The reports of Hägg et al. (2001) and Hatorp et al. (2003) further support that EE-containing OCs do not affect certain P450s like CYP2D6 and CYP2C8.

More importantly, the total concentration of EE in enterocytes (∼6.0 nM) and the hepatic portal vein (∼0.4 nM) has been estimated using accepted methods and found to be low at clinically relevant doses (Zhang et al., 2007). For reference, the free fraction of EE in human serum is 1.7% (Karjalainen et al., 2008 and references therein). So why does EE elicit a significant interaction (>2-fold increase in AUC) with MEL and tizanidine? Why the observed interactions with selegiline and various CYP2C19 substrates? Based on current in vitro extrapolation methods, one would not anticipate clinically relevant reversible or mechanism-based inhibition of any P450s (Rodrigues and Lu, 2004; Zhang et al., 2007).

As a first step to address such questions, EE was systematically evaluated as a reversible and time-dependent inhibitor of numerous drug-metabolizing P450s in pooled HLM (CYP1A2, CYP3A4, CYP2C9, CYP2C8, CYP2C19, CYP2D6, and CYP2B6). CYP2A6 and CYP2E1 were not part of the study; Draper et al. (1997) have reported that EE is a weak inhibitor of the former (coumarin 7-hydroxylase) in HLM. Inhibition of recombinant (r)CYP1A1, rCYP1A2, rCYP1B1, rCYP3A4, rCYP3A5, rCYP2B6, and rCYP2J2 also was evaluated. Some of these latter P450s (CYP1A1, CYP1B1, and CYP2J2) are expressed in the gut and are less prominent in the liver (Paine et al., 2006). Additional inhibition studies (MEL 6-OH formation) were conducted with pooled human intestine microsomes (HIM) and HLM. As much as possible, the different P450s were assessed under similar incubation conditions so that the generated IC₅₀ values could be ranked.

Materials and Methods

Materials. Furafylline, α-naphthoflavone (ANF), ticlopidine, troleandomycin, phencyclidine, paroxetine, phenelzine, phenacetin, acetaminophen, MEL 6-OH, 6-chloromelatonin, MEL, dextromethorphan, dextrorphan, diclofenac, 4′-hydroxydiclofenac, paclitaxel, midazolam, 1′-hydroxymidazolam, terfenadine, (S)-mephenytoin, 4′-hydroxymephenytoin, propranolol, trazodone, chlorzoxazone, 6-hydroxychlorzoxazone, saccharic acid 1,4-lactone, potassium phosphate dibasic, potassium phosphate monobasic, EDTA, and NADPH were purchased from Sigma-Aldrich (St. Louis, MO). 10-Deacetyltaxol-C was purchased from LKT Labs (St. Paul, MN). Tienilic acid was purchased from Toronto Research Chemicals Inc. (North York, ON, Canada). Acetaminophen-D₄, 6β-hydroxytestosterone-D₃, and α-hydroxymidazolam-D₄ were purchased from the Cerilliant Corporation (Round Rock, TX). 4′-Hydroxydiclofenac-D₄ was purchased from SynFine Research (Richmond Hill, ON, Canada). Testosterone, 6β-hydroxytestosterone, EE 3-O-sulfate, and EE 3-O-glucuronide were obtained from Steraloids (Newport, RI). Pooled HLM (n = 27 different organ donors), baculovirus-infected insect cell microsomes (Supersomes) containing various (individual) cDNA-expressed human P450s (coexpressed with P450 oxidoreductase), pooled HIM (n = 15 different organ donors), and 6α-hydroxypaclitaxel were purchased from BD Gentest (Woburn, MA). 2-Hydroxy EE and 2-methoxy EE were provided by Drs G. Newport and W. Slikker (National Center for Toxicology, National Center for Toxicological Research, Jefferson, AR). All the other reagents and chemicals were of analytical grade and of the highest quality available commercially.

Instrumentation. All the P450 assays were liquid chromatography/tandem mass spectrometry-based (Tables 1 and 2) using an API4000 QTrap triple quadrupole mass spectrometer equipped with a turbo-V ionization source (Applied Biosystems, Concord, ON, Canada), two LC-10ADvp pumps with an SCL-10ADvp system controller and DGU-14 solvent degasser (Shimadzu, Columbia, MD), and a LEAP CTC HTS PAL autosampler (LEAP Technologies, Carrboro, NC). Incubations were carried out in 96-well Thermowell Gold plates (Corning Inc., Corning, NY) using an 8-tip Genesis 150 liquid handler (Tecan Group Ltd., Research Triangle Park, NC) equipped with a temperature-controlled shaker.

Incubation and Evaluation of EE as a P450 Inhibitor. The incubation mixture (final volume of 0.2 ml) consisted of the following: 0.1 M potassium phosphate buffer (100 mM, pH 7.4), protein (HLM, HIM, or rP450), EDTA (1 mM), and EE (final concentration of 10 nM to 45 μM) dissolved in dimethyl sulfoxide (0.2%, v/v, final concentration). For each assay, the incubation conditions are described in Tables 1 and 2. When studying reversible inhibition, substrate was added to the reaction mixture and warmed to 37°C. Reactions were then initiated with NADPH (1 mM). After incubation, the reaction mixtures were transferred to a filter plate preloaded with cold methanol (POD and testosterone 6β-hydroxylase assays) or acetonitrile (all the other assays) containing the internal standard. Assessment of time-dependent inhibition involved preincubation of EE (same concentration range) with NADPH-fortified (1 mM) enzyme (30 min), followed by substrate addition without dilution of incubate. After the requisite incubation period, reactions were terminated by transfer of the assay contents to a preloaded filter plate as described above. For all the assays, the filter plate was stacked with a 2-ml injection plate, centrifuged, and the filtrates were subjected to liquid chromatography/mass spectrometry analysis (Tables 1 and 2).

Where possible, the maximal rate of inactivation (k_inact)/concentration of inhibitor that supports half the rate of maximal inactivation (K_I) ratio was estimated from the IC₅₀ at 30 min (IC_50(t)) and compared with positive controls (Maurer et al., 2000; Berry and Zhao, 2008). Only estimates of k_inact/K_I ratio are reported herein, and no attempt was made to obtain the actual values of k_inact and K_I. Inhibition studies with rCYP1A1 and rCYP1A2 were also extended to include four metabolites of EE (2-hydroxy EE, 2-methoxy EE, EE 3-O-sulfate, and EE 3-O-glucuronide). All the incubations were as described above, except that saccharic acid 1,4-lactone (0.1 mM) was included in the incubations with EE 3-O-glucuronide.

Determination of Free Fraction in the Incubation. Equilibrium dialysis was used to determine the binding of EE (2.0 μM) to microsomal (0.05, 0.1, and 0.25 mg/ml) and rP450 (5 pmol/ml) protein. Protein was prepared in the assay buffer (100 mM potassium phosphate, pH 7.4, 1 mM EDTA) and added to a 96-well equilibrium dialysis apparatus (HTDialysis, Gales Ferry, CT). After loading protein spiked with EE (compartment A) and buffer (compartment B) to the dialysis membrane (HTDialysis; molecular weight cutoff, 12,000–14,000 Da), the apparatus was incubated in a 37°C oven with a reciprocating shaker for 6 h. Protein samples removed from the dialysis wells were diluted (10-fold) with buffer, and buffer samples were diluted (2-fold) with protein. Two volumes of methanol (containing 0.1% formic acid and internal standard, 6β-hydroxytestosterone) were then added to each sample. A portion of the supernatant (10 μl) was subjected to liquid chromatography/tandem mass spectrometry using a QTrap triple quadrupole mass spectrometer (as described above). Chromatographic separation was accomplished using a Phenomenex (Torrance, CA) Luna Phenyl-hexyl column (2 × 150 mm, 5 μm). For the mobile phase, a gradient system was used at a flow rate of 0.3 ml/min. Initially, the mobile phase comprised methanol/water (40:60, v/v) containing 0.1% formic acid. The mobile phase was then programmed so that the amount of methanol was 100% (containing 0.1% formic acid) in 2 min. At which time, the mobile phase was held constant for another 2 min. Detection of EE was achieved in the electrospray positive ion multiple reaction monitoring mode by monitoring m/z transitions of 279 → 133 (EE) and 305 → 269 (internal standard). The declustering potential was set at 51 V (EE) and 61 V (internal standard); the collision energy was set at 25 V (EE) and 23 V (internal standard); and the turbo-V source temperature was set at 400°C.

Results

Reversible (Non–Time-Dependent) Inhibition of P450s. EE was evaluated as an inhibitor of seven different P450s in HLMs (CYP3A4, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP1A2, and CYP2B6). Inhibition was assessed also with seven different rP450 proteins (CYP3A4, CYP3A5, CYP1A1, CYP1A2, CYP1B1, CYP2J2, and CYP2B6). IC₅₀s were generated and were then ranked (Table 3). In all the cases, EE was found not to be a submicromolar inhibitor of P450 activity catalyzed by rP450s and HLM preparations. Binding of EE to HLM was low [free fraction in the incubation (f_u,inc), 0.84 to 0.64], and based on the IC₅₀ not corrected for f_u,inc (IC_50(total)) and IC₅₀ corrected for f_u,inc (IC_50(free)) values obtained, CYP2C19 was ranked the lowest (IC_50(total) = 4.1 μM and IC_50(free) = 2.8 μM). The next highest IC₅₀ in HLM was CYP2C9, followed by CYP2C8, CYP2B6, CYP2D6, and CYP3A4. The weakest inhibition in HLM was observed with CYP1A2 (POD) activity (IC_50(total) > 45 μM and IC_50(free) > 39 μM). Under the same incubation conditions, ANF behaved as a potent inhibitor of POD activity (IC₅₀ ∼ 7 nM) (data not shown). Therefore, the POD assay was able to detect inhibition of CYP1A2.

Perhaps the most important finding in the present study was the relatively potent inhibition of rCYP1A1-catalyzed POD by EE (Table 3). Because assays were run at the K_m of the substrate, the K_i(free) was estimated to be ∼1000 nM (assuming competitive inhibition, IC_50(free)/2). The results of more formal kinetic studies revealed that EE was in fact a competitive inhibitor (K_i = 1.4 ± 0.12 μM) of rCYP1A1-catalyzed POD (Fig. 1). Binding of EE to the rCYP1A1 preparation was not detectable (f_u,inc ∼ 1.0); thus, the K_i generated is regarded as a “free K_i.” In comparison, EE was found to be a weaker inhibitor of rCYP3A4, rCYP3A5, rCYP1B1 (IC₅₀ ∼ 9.0 μM), rCYP1A2 (IC₅₀ = 14 μM), rCYP2J2 (IC₅₀ = 31 μM), and rCYP2B6 (IC₅₀ = 41 μM) (Table 3).

Time-Dependent Inhibition. After 30 min of preincubation, IC₅₀ shifts (≥2-fold) were only observed for CYP2C8 and CYP3A4 activity in HLM and rCYP2J2, rCYP3A4, and rCYP3A5 (Table 3). Therefore, no time-dependent inhibition of either CYP1A2 or CYP1A1 was observed. Although the moderate (∼2-fold) IC₅₀ shift for rCYP2C8 and rCYP2J2 precluded estimates of a k_inact/K_I ratio, the expected greater than 4-fold shift enabled estimates of the k_inact/K_I ratio for HLM-(0.005 min^–1μM^–1) and rCYP3A4-catalyzed (0.031 min^–1μM^–1) midazolam 1′-hydroxylase activity. There was some evidence for the time-dependent inhibition of rCYP3A5, but the IC₅₀ shift was lower versus rCYP3A4 (6-fold versus ∼2-fold). Such an observation is consistent with reports indicating that the k_inact/K_I ratio is lower for rCYP3A5 versus rCYP3A4 (Lin et al., 2002; Lin and Hollenberg, 2007). Despite the report of Kent et al. (2002), no time-dependent inhibition of CYP2B6 activity was observed with HLM or recombinant protein (Table 3; see Discussion).

EE Metabolites as CYP1A Inhibitors. EE is known to be extensively metabolized and undergoes first-pass extraction in the gut and liver. Sulfation at the 3-hydroxy position is major in the gut, although the products of 2-hydroxylation (followed by methylation) and 3-O-glucuronidation are also present in the excreta following a radiolabeled dose (Zhang et al., 2007 and references therein). Therefore, four metabolites (EE 3-O-sulfate, EE 3-O-glucuronide, 2-hydroxy EE, and 3-methoxy EE) were evaluated as inhibitors of rCYP1A1 and rCYP1A2 (Fig. 2). For both the sulfate and glucuronide conjugates, relatively minimal inhibition of POD activity (IC₅₀ > 45 μM) was observed (<30% inhibition at the highest concentration tested), and no time-dependent inhibition was evident (data not shown). The 2-hydroxy and 2-methoxy EE derivatives were no more potent than parent EE as rCYP1A1 and rCYP1A2 inhibitors (Fig. 2). Thus, despite extensive EE sulfation during first pass, the sulfate conjugate is unlikely to contribute to the observed drug interactions with CYP1A substrates.

MEL as CYP1A1 and CYP1A2 Substrate. Ma et al. (2005) have reported MEL 6-OH formation in the presence of rCYP1A1, and the authors described K_m and V_max values similar to those of rCYP1A2. The data presented in Fig. 3 show that it is possible to normalize the turnover rates reported by Ma et al. (2005), the most complete recombinant panel described to date, and predict that CYP1A2 dominates (66%) in HLM, with contribution from CYP3A4 (13%), CYP2C9 (6%), and CYP2C19 (7%) (Fig. 3B). Because no reaction phenotype data exist for HIM, the rP450 data reported by the authors were also used to estimate the contribution of different P450s in HIM. In this instance, CYP1A1 (52%) and CYP3A4 (43%) were predicted to account for the majority of MEL 6-OH formation in HIM (Fig. 3C).

Although not shown in Fig. 3, the P450 reaction phenotype of MEL O-demethylation also was considered. Based on the data of Ma et al. (2005), it was estimated that CYP1A1, CYP2C19, CYP2C9, and CYP3A4 contribute to 24, 53, 9, and 11% of total O-demethylase activity in HIM, respectively. For HLM, CYP1A2, CYP2C19, CYP2C9, and CYP3A4 are estimated to contribute 50, 35, 6, and 8%, respectively. Overall, the data suggest that CYP1A1 and CYP2C19 (∼80%) may play a significant role in the gut.

EE as Inhibitor of MEL 6-OH Formation in HIM and HLM. As described above, MEL 6-OH formation is predicted to be catalyzed by CYP1A1 (52%) and CYP1A2 (66%) in HIM and HLM, respectively. To confirm the reaction phenotype in HIM and HLM, ANF was used as a CYP1A reaction phenotyping tool. The compound is a potent inhibitor of both CYP1A1 and CYP1A2 (∼10-fold lower IC₅₀ for CYP1A2) and does not inhibit other drug-metabolizing P450s at concentrations less than 1.0 μM (Tassaneeyakul et al., 1993; Bourrié et al., 1996; Shimada et al., 2007). In the presence of human microsomes, the formation of MEL 6-OH was characterized by a single K_m of ∼20 μM (data not shown), similar to that of rCYP1A1 and rCYP1A2 reported by Ma et al. (2005). However, as expected a differential pattern was observed for the inhibition of MEL 6-OH formation by ANF in HIM versus HLM (Fig. 4A). For example, ANF was not able to completely inhibit the reaction in HIM, and the residual (uninhibited) activity (44%) could be attributed to CYP3A4 (Fig. 3C). In comparison, ANF (1 μM) was able to inhibit MEL 6-OH formation in HLM by 66%, consistent with the furafylline data presented by Facciolá et al. (2001).

As expected for EE, the IC₅₀ value for the inhibition of MEL 6-OH formation in HIM was low compared with HLM (IC₅₀ = 0.91 ± 0.06 versus >40 μM) (Fig. 4B) and was similar to the IC₅₀ value obtained with rCYP1A1 (Table 3). For EE, the I_max (maximal percentage inhibition) of 61 ± 2.5% was similar to that of ANF (56 ± 3.8%) in HIM, consistent with the CYP1A1-catalyzed formation of MEL 6-OH therein.

Discussion

Despite eliciting interactions with a number of drugs, relatively little effort has been made to assess the inhibition profile of EE across numerous P450s in vitro (Rodrigues and Lu, 2004; Zhang et al., 2007). Some of the observed drug interactions are marked, with greater than 2-fold increases in AUC. For example, OCs containing EE increase the AUC of selegiline, MEL, and tizanidine >10-, ∼5-, and ∼4-fold, respectively (Laine et al., 1999; Granfors et al., 2005; Hilli et al., 2008). Such effects on AUC cannot be rationalized based on the known dose and exposure of EE (Rodrigues and Lu, 2004; Zhang et al., 2007). Therefore, an attempt was made to systematically evaluate EE as a reversible and time-dependent inhibitor of numerous human P450s in vitro.

Based on the results described herein, it is concluded that EE is not a submicromolar inhibitor of human P450s in vitro (Table 3). The lack of submicromolar inhibition of CYP2C19 (Laine et al., 2003; Rodrigues and Lu, 2004; Di Marco et al., 2007), CYP2C8 (Walsky et al., 2005), CYP2C9 (Laine et al., 2003), CYP2B6 (Walsky et al., 2006), and CYP1A2 (Karjalainen et al., 2008) activity in vitro is consistent with the literature. To our knowledge, however, this is the first report describing the assessment of EE as an in vitro inhibitor of CYP2J2, CYP2D6, CYP1B1, and CYP1A1. Pepper et al. (1991) have reported the inhibition of CYP2D6-catalyzed metoprolol oxidation in HLM (83%), although no IC₅₀ was reported, and only a single high EE concentration (100 μM) was evaluated. Also consistent with the literature, there was evidence for time-dependent inhibition of CYP3A4 and CYP3A5, although the inhibition is not considered to be clinically meaningful given the minimal effect on the PK of oral midazolam and nifedipine (Balogh et al., 1998; Palovaara et al., 2000; Belle et al., 2002; Shelepova et al., 2005; Lin and Hollenberg, 2007).

Unexpectedly, it was not possible to show time-dependent inhibition of rCYP2B6 with EE under conditions where phencyclidine elicited a marked IC₅₀ shift (Table 3, legend). Consequently, a number of additional incubation formats were attempted (e.g., change of substrate concentration, increased protein concentration during preincubation with subsequent dilution of incubate), and still it was not possible to observe an IC₅₀ shift with EE. Finally, additional studies were conducted with 7-ethoxy-fluorocoumarin as substrate, and there was no further evidence for time-dependent inhibition of rCYP2B6. In fact, the IC₅₀ generated (∼20 μM) was similar to that reported by Walsky et al. (2006) (data not shown). Therefore, the results are contrary to those of Kent et al. (2002), who were able to observe mechanism-based inhibition of 7-ethoxy-fluorocoumarin O-deethylation. However, it is worth noting that commercially available (insect cell-expressed) rCYP2B6 was used in the present study, whereas the studies of Kent et al. (2002) used rCYP2B6, expressed in Escherichia coli, purified to homogeneity, and incubated in a standard reconstitution system. It is possible that mechanism-based inhibition by EE may be sensitive to assay conditions (P. Hollenberg, personal communication). EE also did not behave as a time-dependent inhibitor of CYP2B6 activity in native (pooled) HLM, and it is assumed that such a preparation expresses the requisite ratio of P450/P450 reductase and cytochrome b₅. More importantly, under the same assay conditions, phencyclidine did exhibit time-dependent inhibition of bupropion hydroxylation in HLM (IC₅₀ > 120 μM; IC_50(t) = 5.5 ± 1.6 μM). Furthermore, the IC₅₀ for EE generated with either rCYP2B6 or HLM (Table 3), as well as subsequent estimate of concentration of inhibitor ([I])/K_i ratio, would correctly predict a minimal drug interaction with bupropion (Palovaara et al., 2003).

Perhaps the most important finding in the present study was the relatively potent inhibition of CYP1A1 by EE (Fig. 1). In reality, the interaction with rCYP1A1 was not completely unexpected, given that the enzyme has been shown to metabolize EE (Wang et al., 2004). When compared with other P450s, the IC_50(total) values for rCYP1A1 (2.7 μM) and HLM CYP2C19 (4.4 μM) were ranked the lowest, yielding a K_i of 1.4 μM (determined) and ∼2.2 μM (estimated), respectively. The latter estimate compliments the recent report of Foti and Wahlstrom (2008) using (S)-omeprazole, (R)-omeprazole, and (S)-fluoxetine as rCYP2C19 substrates. However, the same authors reported a lower K_i (289 nM) with (S)-mephenytoin. When adjusted for HLM binding, the CYP2C19 K_i(free) (IC_50(free) = 2.8 μM) in the present study is estimated to be similar to that of CYP1A1 (1.4 μM). Other authors have reported IC₅₀s for (S)-mephenytoin with HLM in the range of ∼3.5 to 20 μM, although no attempt was made to correct for microsomal binding (Rodrigues and Lu, 2004; Di Marco et al., 2007). To date, all the reported K_i and IC₅₀ values for CYP2C19 are still considerably higher than the calculated total concentration of EE in enterocytes (∼6.0 nM) and hepatic portal vein (∼0.4 nM). Karjalainen et al. (2008) have come to the same conclusion for CYP1A2 and have suggested that accumulation of EE in tissues is a possibility.

Because of the marked interaction with MEL and because most reports have focused on the metabolism of MEL by HLM and liver P450s (Facciolá et al., 2001; Härtter et al., 2001; Ma et al., 2005; Hilli et al., 2008), the present study also focused on the P450s involved in the metabolism of MEL. CYP1A2 is expressed in the liver, and one would expect the enzyme to play a major role in HLM. In agreement, reaction phenotyping of MEL with furafylline indicates that CYP1A2 plays a significant role (∼60%) therein, although some inhibition by CYP2C9-selective sulfaphenazole (20%) and CYP3A-selective ketoconazole (∼15%) can be observed (Facciolá et al., 2001). The data presented in Fig. 3 show that it is possible to normalize the turnover rates reported by Ma et al. (2005) and predict that CYP1A2 dominates (66%) in HLM, with contribution from CYP3A4 (13%), CYP2C9 (6%), and CYP2C19 (7%). For HIM, CYP1A1 (52%) and CYP3A4 (43%) are predicted to account for the majority of MEL 6-OH formation. MEL O-demethylation was predicted to be catalyzed by CYP1A2 (50%), CYP2C19 (35%), CYP2C9 (6%), and CYP3A4 (8%) in HLM (see Results). The contribution of CYP1A2 and CYP2C19 is consistent with the data of Facciolá et al. (2001), who reported up to 40% inhibition of HLM-catalyzed MEL O-demethylation with furafylline and up to 40% inhibition with omeprazole (used as a CYP2C19 inhibitor). On the other hand, CYP1A1 (24%) and CYP2C19 (53%) would be expected to contribute to the majority of MEL O-demethylase activity in HIM. This process implies that CYP1A1 (52%) and CYP1A2 (66%) contribute to MEL 6-OH formation in HIM and HLM, respectively. For MEL O-demethylation, CYP2C19 (∼40%) plays a role in HLM, whereas CYP2C19 and CYP1A1 (∼80%) dominate in HIM. Therefore, given the differences in the IC₅₀ for CYP1A1 and CYP1A2, one would expect EE to be a more potent inhibitor of MEL 6-OH in HIM (versus HLM) (Fig. 4B). Although not determined, inhibition of HIM-catalyzed MEL O-demethylation by EE is implicated.

The results of the present study raise some important questions. EE is a more potent inhibitor of CYP1A1 (versus CYP1A2), which is expressed in the gut (Paine et al., 2006). Because rCYP1A1 also metabolizes MEL (Fig. 3), is it possible that the interaction between EE and MEL involves the inhibition of gut CYP1A1 during first pass (Fig. 4B)? Caffeine and theophylline also serve as rCYP1A1 substrates (Ha et al., 1995, 1996), so could EE also affect their metabolism in the gut? Is it possible that the concentrations of EE exceed 1000 nM in the gut during first pass to the point that CYP2C19 (in addition to CYP1A1) is inhibited with a relatively minimal effect on other P450s? CYP2C19 is present in the intestine and probably contributes to the first-pass metabolism of its probe drugs (Galetin and Houston, 2006; Paine et al., 2006). Moreover, the interaction of EE-containing OCs and CYP2C19 substrates appears to be significant. For example, the increased (0.28 versus 0.11) ratio of (S)-mephenytoin to (R)-mephenytoin in urine (S/R ratio) with EE is similar to that observed with CYP2C19*1/*2 (heterozygous) subjects and is suggestive of up to 50% inhibition of the enzyme (Rodrigues and Lu, 2004 and references therein).

It is worth noting that the ranked IC₅₀s described in Table 3 are consistent with the results of Shelepova et al. (2005), who evaluated the impact of an EE-containing OC formulation on five P450s using a5 + 1 drug mixture. In the study, no significant inhibitory effect was seen with the CYP3A (midazolam), CYP2C9 [(S)-warfarin], and CYP2D6 (dextromethorphan) probes. On the other hand, the effect on the putative CYP2C19 (100% increase in the omeprazole/5-hydroxy omeprazole AUC ratio) and CYP1A2 (23% decrease in the caffeine demethylation ratio of [5-acetyl-amino-6-formylamino-3-methyluracil + 1-methylxanthine + 1-methylurate]/1,7-dimethylurate in urine) trait measures was statistically significant. Given the arguments above, is it possible that the effect on the caffeine demethylation ratio is more reflective of CYP1A1 inhibition (versus CYP1A2 inhibition)?

In light of the absence of submicromolar P450 inhibition in vitro, the magnitude of the effect on MEL, tizanidine, and selegiline PK, as well as the low dose of 30 μg, it is apparent that additional in vitro and in vivo studies with EE are necessary. In particular, for drugs like MEL that undergo considerable first pass and are metabolized by CYP1A1, the possibility that EE directly inhibits CYP1A1 and affects gut extraction should be investigated, and it cannot be assumed that interactions simply involve the direct inhibition of liver CYP1A2. More importantly, the majority of the P450s evaluated in the present study (e.g., CYP3A4, CYP2D6, CYP2C8, and CYP2C9) exhibited lower IC₅₀s than CYP1A2, and clinical drug interactions with the same P450s are minor. For MEL in particular, the inhibition CYP2C19-catalyzed O-demethylation may also contribute to the observed interaction with EE.

Although codosed with EE, the observed interactions with MEL cannot be ascribed to gestodene. Like in Karjalainen et al. (2008), gestodene was shown to be a very weak inhibitor (IC₅₀ and IC_50(t) > 100 μM) of POD activity in HLM (data not shown). Likewise, the interaction with MEL likely does not involve the conjugate metabolites of EE. Neither the EE 3-O-sulfate nor the 3-O-glucuronide behaved as potent inhibitors of CYP1A1 or CYP1A2 (Fig. 2). On the other hand, both the 2-hydroxy and 2-methoxy metabolites of EE were shown to inhibit CYP1A1 (similar IC₅₀ to parent EE). Therefore, the possible contribution of both should be considered in vivo. It is unfortunate that there are limited data on the exposure of both of these metabolites in humans (Zhang et al., 2007).

One also has to consider the possibility that chronic EE dosing may affect specific forms of P450 (e.g., CYP1A1) at the transcriptional level. Such effects may involve transcription factors, aryl hydrocarbon receptor (AhR) antagonism (or transrepression), and could be direct or involve the estrogen receptor (Beischlag and Perdew, 2005). For example, preliminary data show that EE may behave as an AhR antagonist in HepG2 cells stably transfected with the CYP1A1 promoter upstream of the luciferase reporter gene. In this instance, EE decreases the induction of luciferase by 2,3,7,8-tetrachlorodibenzo-p-dioxin (up to ∼40%) in a concentration-dependent manner (0.1 nM to 10 μM). In comparison, 3′,4′-dimethoxyflavone (5 μM) decreases the induction by 95% (H. Swanson and E.-Y. Choi, unpublished data). The chronic effects of AhR antagonism are not known.

From the viewpoint of P450 drug interactions, EE continues to be an enigmatic drug. Additional research is needed to enable a better understanding of such interactions at a mechanistic level. This is important because most, if not all, new chemical entities are evaluated as perpetrators and victims of OC interactions. The results of such studies appear in the product label, which, in some cases, can influence competitive marketing of OCs to women of child-bearing potential (Zhang et al., 2007).