In vitro evaluation of fenfluramine and norfenfluramine as victims of drug interactions

Abstract Fenfluramine (FFA) has potent antiseizure activity in severe, pharmacoresistant childhood‐onset developmental and epileptic encephalopathies (e.g., Dravet syndrome). To assess risk of drug interaction affecting pharmacokinetics of FFA and its major metabolite, norfenfluramine (nFFA), we conducted in vitro metabolite characterization, reaction phenotyping, and drug transporter−mediated cellular uptake studies. FFA showed low in vitro clearance in human liver S9 fractions and in intestinal S9 fractions in all three species tested (t1/2 > 120 min). Two metabolites (nFFA and an N‐oxide or a hydroxylamine) were detected in human liver microsomes versus six in dog and seven in rat liver microsomes; no metabolite was unique to humans. Selective CYP inhibitor studies showed FFA metabolism partially inhibited by quinidine (CYP2D6, 48%), phencyclidine (CYP2B6, 42%), and furafylline (CYP1A2, 32%) and, to a lesser extent (<15%), by tienilic acid (CYP2C9), esomeprazole (CYP2C19), and troleandomycin (CYP3A4/5). Incubation of nFFA with rCYP1A2, rCYP2B6, rCYP2C19, and rCYP2D6 resulted in 10%−20% metabolism and no clear inhibition of nFFA metabolism by any CYP‐selective inhibitor. Reaction phenotyping showed metabolism of FFA by recombinant human cytochrome P450 (rCYP) enzymes rCYP2B6 (10%–21% disappearance for 1 and 10 µM FFA, respectively), rCYP1A2 (22%−23%), rCYP2C19 (49%−50%), and rCYP2D6 (59%−97%). Neither FFA nor nFFA was a drug transporter substrate. Results show FFA metabolism to nFFA occurs through multiple pathways of elimination. FFA dose adjustments may be needed when administered with strong inhibitors or inducers of multiple enzymes involved in FFA metabolism (e.g., stiripentol).


| INTRODUC TI ON
The severity and pharmacoresistance of these conditions often require multi−ASM regimens. Resultant drug-drug interactions (DDIs) may lead to loss of efficacy or to toxicity by reducing or elevating plasma drug levels to subtherapeutic or toxic concentrations, respectively. 6 Characterizing the metabolic stability, metabolites, and substrate potential for drug transporters of antiseizure medications (ASM) and the enzymes that catalyze their biotransformation and elimination can help predict pharmacokinetic DDIs. Early literature dating to the 1960s and 1970s-when FFA was marketed as an anorectic agentdescribes FFA and nFFA metabolism in multiple species, 7-12 but the clinical pharmacology of FFA and nFFA has not been systematically described using updated protocols to predict potential clinical DDIs according to EMA and FDA guidelines. 13,14 Seven cytochrome P450 (CYP450) enzymes mediate phase 1 biotransformation and elimination reactions for most orally administered xenobiotics in the liver and, to a lesser extent, in the intestine. 15 Reaction phenotyping studies in liver microsomes can reveal the key CYP450 enzymes catalyzing metabolic transformation, allowing prediction of pharmacokinetic DDIs.
In the context of ASM polypharmacy, ASMs with fraction metabolized (f m ) >25% by a single enzyme or clearance pathway have high victim potential. 14,16 ASMs that are substrates of drug transporters are potential DDI victims. Earlier reports document FFA and nFFA metabolism, 9,12,17,18 but FFA metabolism has not been fully characterized in the context of predicting DDIs when used as an add-on to existing ASM regimens in patients with DS or LGS. In vitro studies conducted according to DDI guidance have not yet been reported. 13,14 Developed by experts in drug metabolism, the FDA and EMA guidance documents were compiled based on the most current state of the science for predicting DDI. The guidelines constitute "a systematic, risk-based approach to assessing DDI potential of investigational drugs and making recommendations to mitigate DDIs." 14 Polypharmacy is common in treating developmental and epileptic encephalopathies; FFA is likely to be used in combination ASM regimens. The potential for DDIs must be fully evaluated, and the rigor-  of the cell) from pooled human liver (n = 50) and intestine (n = 10), pooled canine liver (n = 8) and intestine (n = 7), and pooled rat liver (n = 454) and intestine (n = 199). All rCYP450 enzymes except rCYP1A2 and rCYP2D6 were expressed with cytochrome b5.

| LC-MS/MS data analysis and processing
The line of best-fit for calibration standards was calculated by weighted (1/x) linear regression based on peak-area ratios of analyte to the internal standard using Analyst Instrument Control and Data Processing software (version 1.6.1; SCIEX). Mean analyte peakarea ratio at each time point was normalized to controls at t = 0 min (100%).

| Reaction phenotyping: rCYP450s
Methods used to evaluate the role of CYP450s in the metabolism of FFA have been described. [22][23][24]

| Chemical inhibition of FFA and nFFA metabolism
To determine the metabolic role of individual CYP450 enzymes
Rat and dog liver S9 fractions showed NADPH-dependent substrate loss of FFA ( Figure 1A). The overall percent of FFA substrate loss in liver S9 fractions after 10 µM FFA was approximately half of substrate loss after 1 µM FFA for all species (compare Figure 1A and 1C). At 10 μM, substrate loss was also dependent on NADPH in rat and dog but not human liver S9 fractions ( Figure 1C). In both rat and dog liver S9 fractions at 10 µM FFA, there was approximately twice as much NADPH-independent FFA substrate loss compared to NADPH-dependent substrate loss ( Figure 1C). In human S9 liver S9 intestinal fractions compared with rat and dog. In vitro CL int was lower in human liver S9 fractions than rat or dog S9 fractions, and FFA clearance was uniformly low in rat, dog, and human intestinal S9 fractions.

| Metabolite identity
FFA, its de-alkylated metabolite, nFFA, and six other FFA-related metabolites (C1−C6) were detected (Table 1). FFA and nFFA were present in extracted ion chromatograms from rat, dog, and human liver and in intestinal S9 fractions at retention times of 9.91 and 9.12 min, respectively (Table 1; Figure 2). FFA was present in incubations with and without NADPH (Figure 2A)  (Table 1), and nFFA was the only metabolite detected in human intestinal microsomes ( Table 1). The FFA metabolites detected in human liver and intestine were also observed in both rat and dog. Evidence of hydroxylation, dehydrogenation, and glucuronidation was observed in rat and dog but not human test systems. Glucuronide conjugation was observed in rat and dog but not human liver S9 fractions in the fragmentation patterns of C3 and C6 (Figure 4, Figure 5). No human-specific metabolites were detected (Table 1). TA B L E 1 FFA metabolites identified in rat, dog, or human liver or intestinal S9 fractions    Table 2). Taken together, the recombinant enzyme and F I G U R E 3 Identification of C2 in 120-min incubations of Fenfluramine (FFA) (10 µM) with rat, dog, or human liver S9 fractions (2 mg protein/mL). (A) Extracted ion chromatogram of C2 from rat, dog, or human liver S9 fractions in the presence or absence of NADPHgenerating system. (B) CID MS/MS spectrum of C2 (m/z 220; t R = 10.8 min). CID, collision-induced dissociation; FFA, fenfluramine; MS/MS, tandem mass spectrometry; m/z, mass-to-charge ratio; NADPH, nicotinamide adenine dinucleotide phosphate (hydrogen); t R , retention time.
In the absence of NADPH, no C2 peak was observed in any species.

| DISCUSS ION
Determining the contribution of CYP450 enzymes to the biotransformation and elimination of FFA and nFFA is an important component of characterizing potential DDIs for FFA in the context of the multi-ASM regimens used to treat DS. 34 This study is the first to characterize the victim potential of FFA by in vitro metabolite identification, CYP450 reaction phenotyping, and drug transporter substrate determination according to recent FDA guidance for industry. 24 Our results suggest that FFA was metabolized to nFFA by CYP2D6, CYP2B6, and CYP1A2, with potential contributions by CYP2C9, CYP2C19, and CYP3A4/5.
In a companion study to the current analysis, neither FFA nor nFFA significantly inhibited or induced CYP450 enzymes, suggesting minimal perpetrator potential at intended clinical doses (0.2−0.7 mg/kg/day, maximum 26 mg/day) (Prescribing information: https://www.finte pla.com/). The absolute oral bioavailability of FFA is reported to be 68%−83%, 35,36 with extensive metabolism by N-dealkylation to nFFA, 7 as confirmed by the current study. Most orally administered, radiolabeled doses of FFA are recovered in the urine as FFA, nFFA, and other metabolites (e.g., glucuronide conjugates of the diol metabolite). 12 All metabolites recovered in urine are also present in plasma, with marked species differences noted in earlier reports. 12,37 For example, deamination was not a major metabolic route in the rat, whereas both N-dealkylation and deamination were important metabolic pathways in the dog, and deamination was observed in the mouse. 12 Little radioactivity was reported to be excreted in feces in these studies. Taken together with the current study, these results suggest that FFA has multiple mechanisms of elimination. Thus, FFA metabolism is unlikely to be critically affected by inhibition of a single pathway of metabolism.
In a recent clinical DDI study, FFA was co-administered with a DS ASM regimen of stiripentol, clobazam, and valproate in healthy subjects. 38 This combination inhibited all of the potential metabolizing enzymes described above and had a significant effect on the pharmacokinetics of FFA, increasing C max , AUC 0-t , and AUC 0 -∞ , while the C max and AUC 0-t of nFFA were reduced. This study suggested a dose adjustment for FFA as a victim drug but suggested there was little propensity for FFA to act as a perpetrator, even in this combination of ASMs with overlapping metabolic pathways. 38 Stiripentol is approved for the treatment of Dravet syndrome in   The appearance of NADPH-independent substrate loss in rat S9 fractions at the 10-µM but not the 1-µM dose suggests that doses used are resulting in CYP450-independent metabolism in rat and other species evaluated, as was observed in another study. 16 A proposed metabolic pathway for FFA is shown in Figure 8, adapted from Brownsill 1991. 9 In accordance with prior reports, in our study, only FFA, nFFA, and the N-oxygenation product of nFFA were found in human liver S9 fractions. All metabolism of FFA appears to lead to nFFA as the first step. None of the observed metabolites could be formed without removal of the N-ethyl group to form nFFA. Evidence of dehydrogenation, hydroxylation, and glucuronidation was observed in rat and dog but not human S9 fractions.
These findings are consistent with prior reports showing that the primary pathway of FFA metabolism is N-dealkylation to nFFA. 9,12 Similar to our observations, species differences have been noted in the catalytic activity of enzymes in the CYP1A, CYP2C, CYP2D, and CYP3A families. 43  Rat and dog did not metabolize FFA in the same way as humans.
FFA appears to be more extensively and rapidly metabolized in some species relative to humans. 12,44 Enzymes from all of these families, especially CYP2D6, were observed to contribute to the metabolism of FFA in our studies. Our reaction phenotyping approach provides data to support and extend prior investigations of FFA metabolism by CYP450s, which used indirect methods and isolated investigations of specific CYP450s. Prior reports supported a general role for CYP450s in FFA metabolism, 45 identified cases where adding FFA to existing regimens resulted in DDIs, 46 and supported a role for CYP2D6 and CYP1A2 in dexfenfluramine and/or l-fenfluramine metabolism. 17,47,48 Reports were contradictory or unclear on CYP2C, CYP3A4, and to our knowledge, CYP2B6 was not specifically investigated in prior reports. Thus, using reaction phenotyping and studies recommended in the US Food and Drug Administration's most recent Guidance for Industry, 24  Overexpression of transporters in the brain and excretory organs has led to some subtherapeutic ASM plasma levels, especially in refractory epilepsies. 49 Lamotrigine, an ASM sometimes prescribed in LGS and/or DS, is a P-gp and a BCRP substrate, but the ASMs phenytoin, phenobarbital, carbamazepine, valproate, topiramate, and levetiracetam did not show substrate capacity for P-gp and BCRP. 50 For FFA and nFFA, net efflux or uptake ratios calculated in the current study were outside the threshold range for substantial DDI potential (net efflux or uptake ratio <2 It should be noted that some drug metabolism studies incorporate relative activity factor analyses to account for system variations and relative enzyme abundance when using a recombinant CYP approach. For the FFA reaction phenotyping we followed current FDA recommendations that two methods are to be used: enzyme inhibition in HLM/hepatocytes and the recombinant CYPs. We agree that application of the relative activity factor improves extrapolation of in vitro results to in vivo. The results of the two methods we report converge on three major CYPs with indication of possible minor contributions from additional enzymes. The three major enzymes comprise inducible and non-inducible CYPs, as well as polymorphically and non-polymorphically expressed CYPs. We consider this information sufficient for addressing a concern of potential susceptibility of FFA to DDI due to a single or a limited number of clearance pathways. Were the results different, application of the relative activity factor would have been considered.
The apparent half-lives of the parent drug loss were significantly different between 1 and 10 µM and were variable in the species tested, suggesting potentially different apparent Km values among species. To derive the intrinsic clearance from apparent half-life obtained at a single concentration, it is important to ensure that the concentration is well-below the apparent Km; alternatively, both 1 and 10 µM should be considered. In human liver S9 incubations, at 10 μM F I G U R E 8 Metabolic pathway of fenfluramine. Adapted from Brownsill 1991 9 FFA with cofactors, the percent of the drug remaining was 24.4%, 8.8%, and 9.9% at 120, 60, and 30 min, respectively. At 1 μM FFA, the percent of the drug remaining was 20.9%, 11.1%, and 1.9% at 120, 60, and 30 min, respectively. In the absence of the cofactors the percent of the drug remaining was 22.9% and 21.6% in incubations of 10 and 1 μM FFA, respectively. These data are fairly close for the two drug concentrations examined. Aside from that observation, we proceeded on the side of caution and used the data for the lower concentration of the drug, 1 μM, for the evaluation of the half-life of FFA. Our data support differences in FFA half-life values obtained in the rat and dog were larger than in the human at two concentrations of the drug.
In the metabolic stability and metabolite identification experiments, UDPGA was included as the cofactor for UGT, but pore forming agents (e.g., alamethicin) were not included in the incubations. The effect of alamethicin on the activity of UGT enzymes is attributed to alamethicin-mediated formation of pores in the microsomal membrane and an increased access of the UGT substrate to the microsomal lumen where the enzyme binding site is exposed. The addition of alamethicin could increase enzymatic activity of the UGT enzymes. In our studies, FFA was found not to be a substrate for the uptake transporters recommended in the current FDA guidance; therefore, providing an additional route of access of the drug to the enzyme could overestimate its UGTmediated clearance.
In conclusion, the in vitro studies described in the current report suggest that FFA may have victim potential, which can manifest when the drug is co-administered with other ASM regimens containing drugs with substantial effects on combinations of CYP450 family enzymes. The CYP450s with victim potential are CYP2D6, CYP2B6, CYP1A2, and (to a lesser degree) CYP2C9, CYP2C19, and CYPA4/5.

ACK N OWLED G EM ENTS
The authors thank scientists from Enzyme Incubations and from Analytical Sciences and Data Processing groups at Sekisui XenoTech for their technical expertise in execution of this study. This study was funded by Zogenix, Inc. Medical writing and editorial assis-

DATA AVA I L A B I L I T Y S TAT E M E N T
Zogenix is in the process of establishing a data sharing policy. Written requests for data by legitimate investigators/researchers/clinicians may be submitted to Zogenix, Inc. These requests will be considered on a case-by-case basis and reviewed for appropriateness.