3. Mechanism of CYP2D6 Inhibition by Fluoxetine

S-Fluoxetine and S-norfluoxetine are both substrates and exhibit a high binding affinity to CYP2D6 (Ki = 68 nM and 35 nM, respectively; Ki is reflective of the binding affinity of a potential inhibitor; the lower the Ki, the higher the binding affinity is and smaller the amount of the drug needed in order to inhibit the activity of that enzyme) [39]. Thus, it is clear that competitive inhibition with sensitive CYP2D6 substrates is expected, and that fluoxetine and norfluoxetine will likely act as perpetrator drugs for several other CYP2D6 substrates. Accordingly, fluoxetine administration for 8 days is associated with a significant decrease in dextromethorphan clearance (a sensitive CYP2D6 substrate) and metabolism to dextrorphan: a 9- to 18-fold increase in the dextromethorphan/dextrorphan ratio [26,40]. Remarkedly, however, as is discussed further below (see Section 5), it took two to three weeks after discontinuation of fluoxetine for CYP2D6 activity to return back to baseline [26,27]. Understanding the exact mechanism of this prolonged inhibition of CYP2D6 activity poses certain challenges.

The first possibility is that prolonged CYP2D6 inhibition by fluoxetine might be due to noncompetitive inhibition. As a CYP2D6 substrate, fluoxetine must bind to the active site to be metabolized into norfluoxetine. By definition, pure noncompetitive inhibitors are not metabolized by the enzyme they inhibit [28]. Additionally, the 3D-conformational structure of noncompetitive inhibitors generally does not resemble that of enzyme substrates. Since fluoxetine is a CYP2D6 substrate, it appears not to meet some of the criteria of a pure noncompetitive inhibitor.

Second, fluoxetine could be a mixed CYP2D6 inhibitor, i.e., exhibiting both competitive and noncompetitive inhibition. This mechanism of inhibition gives rise to very potent enzyme inhibitors, often acting as perpetrator drugs with other sensitive substrates, while their pharmacokinetics are not affected by other inhibitors of the same enzyme [41]. In vitro studies showed that quinidine (a potent CYP2D6 inhibitor) was capable of inhibiting the N-demethylation of fluoxetine with a magnitude similar to that observed with anti-CYP2D6 antibodies [37,42]. These data suggest that the probability of noncompetitive CYP2D6 inhibition by fluoxetine is rather unlikely but leave open the possibility for competitive inhibition.

Third, since fluoxetine is a CYP2D6 substrate, studies have been conducted to determine whether fluoxetine produces the time-dependent, mechanism-based inactivation of CYP2D6 [43]. Bertelsen et al. determined the time-dependent inhibitory potency of fluoxetine in human liver microsomes [43]. The IC₅₀ for CYP2D6 inhibition by fluoxetine did not decrease with various times of pre-incubation periods to allow a potential reactive metabolite to be formed. These observations led to the conclusion that fluoxetine does not exhibit a time-dependent increase in its inhibitory potency and thus is not a mechanism-based inhibitor. Further evidence for the lack of noncompetitive and mechanism-based inhibition of CYP2D6 by fluoxetine is also obtained looking at results from pharmacogenetic studies. It is well known that poor metabolizers of CYP2D6 (carriers of two nonfunctional CYP2D6 alleles) exhibit decreased clearance characteristics of CYP2D6 probe substrates compared to normal CYP2D6 metabolizers (carriers of two functional CYP2D6 alleles) [44]. If noncompetitive or mechanism-based inhibition were to occur following administration of fluoxetine, the initial differences in fluoxetine pharmacokinetics observed between normal and poor metabolizers would disappear with time, as normal metabolizers would be phenoconverted to poor metabolizers due to the time-dependent decrease in CYP2D6 activity (phenoconversion being a phenomenon by which an genotype-predicted phenotype is transformed into another phenotype by factors such as drug interactions). In that sense, Fjordside et al. investigated the stereoselectivity of fluoxetine demethylation into norfluoxetine in patients on days 7, 14, and 23 following fluoxetine administration (20 mg) [38]. The plasma concentrations of fluoxetine and norfluoxetine enantiomers were measured in poor and normal metabolizers of CYP2D6. First, it was confirmed that CYP2D6-mediated demethylation of fluoxetine is stereoselective toward the S-enantiomer. Second, their results demonstrated a persistent difference between normal and poor metabolizers in the concentrations of S-fluoxetine and S-norfluoxetine even after 3 weeks of treatment [38]. Higher concentrations of S-fluoxetine and lower concentrations of S-norfluoxetine were observed in poor metabolizers compared to normal metabolizers. As indicated before, the persistent difference observed in fluoxetine pharmacokinetics between poor and normal metabolizers suggests a lack of CYP2D6 auto-inhibition via noncompetitive or mechanism-based inhibition. Scordo et al. also reported on the influence of CYP2D6 polymorphisms on fluoxetine steady-state pharmacokinetics: the median S-norfluoxetine/S-fluoxetine ratios were higher in homozygous compared to the heterozygous normal metabolizers for CYP2D6 [45]. Finally, a correlation between the fluoxetine/norfluoxetine ratio and CYP2D6 genotype (number of active CYP2D6 genes) was also reported in other clinical studies [36,46,47,48]. Overall, these findings support that (1) fluoxetine is a substrate of CYP2D6, (2) fluoxetine is not a noncompetitive inhibitor of CYP2D6, (3) CYP2D6 regulates the concentration of the S-enantiomer and the formation of S-norfluoxetine even after multiple doses, and (4) CYP2D6 inhibition by fluoxetine is not time-dependent, and therefore incompatible with mechanism-based inhibition. Thus, by eliminating the alternatives, it appears that fluoxetine is a strong affinity substrate of CYP2D6 and that the reduced CYP2D6 activity is likely mediated through competitive inhibition at the active site. However, the prolonged time-course of CYP2D6 inhibition which subsides even after stopping fluoxetine administration requires further investigation, since it is not commonly encountered when dealing with competitive inhibitors.

4. Time Course of CYP2D6 Inhibition

Conceptually, separating the time of administration between two competing substrates, such that their maximum concentrations in the liver or plasma do not overlap, is a meaningful way to mitigate DDIs caused by competitive inhibition. For most drugs, CYP2D6 inhibition is observed following drug absorption during the hepatic first pass phase (CYP2D6 is not expressed in the intestinal wall) since drug concentrations in the portal vein (which irrigates perilobular hepatocytes) are much higher than concentrations observed after peak plasma concentrations when drugs return to the liver via the hepatic artery. Generally, the ratio of drug concentrations observed in the portal vein during the absorption phase to that measured in the hepatic artery after the peak plasma concentration (when the absorption phase is almost completed) is very large (likely in the range of 100 to 1000). This is why separating the time of administration of two substrates that compete for the same enzyme circumvents competitive inhibition between these two substrates. However, as mentioned previously, in the case of fluoxetine, a prolonged CYP2D6 inhibition over two to three weeks was observed when dextromethorphan was used as a probe substrate [26,27]. Other reports have suggested that these prolonged effects may advantage fluoxetine over other SSRIs in avoiding withdrawal syndromes [49,50]. Significant DDIs have also been reported between fluoxetine and tricyclic antidepressants, several of them also being CYP2D6 substrates [51]. Desipramine (50 mg) mean plasma concentrations were increased 4.4-fold when co-administered with 20 mg fluoxetine for 20 days [52]. Twenty-one days post-discontinuation of fluoxetine, mean plasma concentrations of desipramine still remained 2.4-fold higher than baseline, demonstrating the long-term impact of fluoxetine administration on other CYP2D6 substrates [52]. It is established that both fluoxetine and norfluoxetine have very long half-lives of 4–6 days and 4–16 days, respectively [18]. Additionally, it has been reported that during chronic dosing, the terminal half-life of fluoxetine and norfluoxetine might increase to become slightly longer (8 and 19.3 days, respectively) [53]. Some particular characteristics may contribute to the prolonged effects of fluoxetine administration on CYP2D6. Persistent high concentrations of unbound fluoxetine and norfluoxetine in plasma are one of these factors. Hence, the degree of in vivo interaction between an enzyme and a substrate can be estimated using the [I]u/Ki ratio, where [I]u refers to the concentration of the unbound drug and Ki is the in vitro inhibition constant for this enzyme [39]. Using Ki values of 68 nM and 35 nM for S-fluoxetine and S-norfluoxetine, Sager et al. reported a [I]u/Ki ratio of 5.8 for S-fluoxetine and of 4.5 for S-norfluoxetine (calculated using the unbound concentration in the liver), suggesting that the unbound concentration of these drugs at steady-state is much higher than the concentration required to block 50% of the CYP2D6 active site [39]. This concept can also be appreciated through the consideration of in vivo metabolic inhibition potential using pharmacokinetic parameters. Since the CYP2D6-mediated demethylation of fluoxetine to norfluoxetine is stereoselective, a more precise estimation is obtained by looking at S-enantiomer concentrations. Sager et al. studied the plasma levels of fluoxetine enantiomers following the administration of fluoxetine 60 mg for 12 days to healthy volunteers; the average plasma concentration of S-fluoxetine and S-norfluoxetine was 770 ± 270 nM and 320 ± 110 nM (mean ± SD), respectively [39]. The plasma fraction unbound (fu) was estimated at 0.14 and 0.13 for S-fluoxetine and S-norfluoxetine, respectively. Based on these parameters, the predicted unbound plasma concentrations of S-fluoxetine (108 nM) and S-norfluoxetine (42 nM) are within the range of CYP2D6 Ki values reported in different studies for S-fluoxetine (68–220 nM) and for S-norfluoxetine (35–310 nM). In these estimations, the average steady-state plasma concentrations instead of the maximal concentration were used to assess the potential metabolic inhibition of CYP2D6. The use of [I]u plasma/Ki can underestimate the prediction of in vivo drug interactions compared to the use of [I]in,u liver/Ki ([I]in,u being the maximal unbound hepatic inlet concentration) [54]. In brief, the long-lasting CYP2D6 competitive inhibition by S-fluoxetine and S-norfluoxetine could be explained by their unbound concentrations remaining much higher than their Ki value throughout a dosing interval and days thereafter.