Cytochrome P450-Mediated Metabolism and CYP Inhibition for the Synthetic Peroxide Antimalarial OZ439

Abstract

OZ439 is a potent synthetic ozonide evaluated for the treatment of uncomplicated malaria. The metabolite profile of OZ439 was characterized in vitro using human liver microsomes combined with LC/MS-MS, chemical derivatization, and metabolite synthesis. The primary biotransformations were monohydroxylation at the three distal carbon atoms of the spiroadamantane substructure, with minor contributions from N-oxidation of the morpholine nitrogen and deethylation cleavage of the morpholine ring. Secondary transformations resulted in the formation of dihydroxylation metabolites and metabolites containing both monohydroxylation and morpholine N-oxidation. With the exception of two minor metabolites, none of the other metabolites had appreciable antimalarial activity. Reaction phenotyping indicated that CYP3A4 is the enzyme responsible for the metabolism of OZ439, and it was found to inhibit CYP3A via both direct and mechanism-based inhibition. Elucidation of the metabolic pathways and kinetics will assist with efforts to predict potential metabolic drug–drug interactions and support physiologically based pharmacokinetic (PBPK) modeling.

Graphical Abstract

Synthetic 1,2,4-trioxolanes (ozonides) have been shown to be highly effective and orally active antimalarials.^1,2 A first generation ozonide, arterolane (1 (OZ277), Figure 1),^3-5 in combination with piperaquine (Synriam) is approved in India and is undergoing further clinical development in Africa⁶ for the treatment of uncomplicated malaria. Arterolane exhibits similar potency compared to the highly effective artemisinin derivatives but has a short in vivo half-life (2–4 h)³ that is only marginally longer than that for dihydroartemisinin (~2 h).⁷ Artefenomel (2 (OZ439), Figure 1) is a second generation synthetic ozonide that retains the potent biological activity and rapid onset of action seen with the artemisinins and arterolane, but with a significantly longer in vivo half-life. In preclinical animal models, the longer exposure profile following oral administration of 2 provided a 100% cure rate with a single dose as low as 20 mg/kg, a feature that could not be obtained with any of the clinically used comparator drugs when tested in the same model.¹ In Phase I⁸ and II⁹ clinical trials and volunteer infection studies,^10,11 2 exhibited good safety properties up to a dose of 1600 mg and a long terminal half-life of >60 h. Phase II results also showed that 2 rapidly cleared both Plasmodium falciparum and Plasmodium vivax blood stage infections with median parasite clearance half-lives of 4.1–5.6 h and 2.3–3.2 h, respectively.⁹ More recent work has indicated that 2 is also active against P. falciparum parasites harboring mutations in the Kelch protein, K13, which is associated with artemisinin resistance.^12-14

Figure 1.

Structures of the free base forms of OZ277 (1, arterolane) and OZ439 (2, artefenomel). 1 was isolated as a hydrogen maleate salt and 2 as a mesylate salt.

Preclinical pharmacokinetic studies with 2 suggested that clearance occurs primarily by hepatic metabolism with minimal recovery of intact compound in urine and a low contribution of blood-mediated degradation processes.¹ Our previous studies¹⁵ with the structurally similar 1 identified two major metabolites resulting from hydroxylation of the distal bridgehead carbons¹⁶ of the spiroadamantane substructure, both of which had very weak antimalarial activity in vitro.

A major focus of current target product and candidate profiles for new antimalarials¹⁷ is improved patient compliance through the delivery of combination therapies that can be utilized in a single dose regimen for treatment and post-treatment protection against reinfection. This ambitious target places high demands on the metabolism and pharmacokinetic properties of candidate molecules to ensure that they can deliver the required extent and duration of systemic exposure to clear parasitemia. To assist with this goal, modeling approaches are being utilized for the prediction of human pharmacokinetics and drug–drug interactions.¹⁸ Elucidation of the cytochrome P450 (CYP) metabolic pathways was therefore an important part of the development program for 2 and contributed to our understanding of its pharmacokinetic and pharmacodynamic properties. This information can be incorporated into a physiologically based pharmacokinetic (PBPK) model that will allow prediction of the risk of drug–drug interactions of 2 with combination partners or coadministered drugs.

RESULTS AND DISCUSSION

Metabolite identification was conducted via a combination of LC/MS-MS, derivatization methods and chemical synthesis. Incubation of 2 with NADPH-supplemented human liver microsomes (HLMs) produced 11 metabolite peaks by LC/MS (Figure 2 and Table 1). Four peaks (b–e, Figure 2, panels B and C) corresponded to an increase in molecular mass of 16 Da indicating oxygen incorporation, six peaks (f–k, Figure 2, panels D and E) had an increase of 32 Da indicating incorporation of two oxygen atoms, and one peak (l, Figure 2, panel F) had a decrease of 26 Da indicating a loss of an ethyl group.

Figure 2.

Mass chromatograms of (A) ozonide 2 (peak a), (B) +16 metabolites (peaks b–d), (C) +16 metabolite (peak e), (D) +32 metabolites (peaks f–h), (E) +32 metabolites (peaks i–k), and (F) −26 metabolite (peak l). Numbers above the peaks refer to retention time in minutes.

Table 1.

Metabolites of 2 Detected Following Incubation with Human Liver Microsomes and Their In Vitro Activity against Chloroquine-Resistant (K1) and Chloroquine-Sensitive (NF54) Strains of P. falciparum

peak (Figure 2)	metabolite	MH+ (ΔDa)	major fragment (m/z)	structural assignment (Figure 3)	IC50 (ng/mL) K1/NF54a
a	Parent	470	304	2	1.6/1.9b
be	monohydroxylation	486 (+16)	304	5 (OZ566)	43/84
de	monohydroxylation	486 (+16)	304	6 (OZ567)	19/37
ce	monohydroxylation	486 (+16)	304	7 (OZ579)	16/28
ed	monohydroxylation	486 (+16)	320	8 (OZ577)	1.4/2.3
L	deethylation	444 (−26)	278	9 (OZ564)	1.6/1.1
fe	dihydroxylation	502 (+32)	304	10 (OZ580)	70/77
id	dihydroxylation	502 (+32)	320	11 (OZ578)	>100/>100
ge	dihydroxylation	502 (+32)	304	12	c
he	dihydroxylation	502 (+32)	304	13	c
jd	dihydroxylation	502 (+32)	320	14	c
kd	dihydroxylation	502 (+32)	320	15	c

^aData represent the mean of two biological replicates from the 72 h [³H]-hypoxanthine incorporation assay.

^bData from ref 1.

^cAuthentic standards were not synthesized for these metabolites.

^dConsumed upon reduction with sodium metabisulfite.

^eResistant to reduction by sodium metabisulfite.

When the 2 microsomal incubation mixture was subjected to reduction with sodium metabisulphite, peaks e and i–k (Figure 2) were each consumed while the remaining peaks (b–d, f–h) were unaffected. This further suggested that metabolites corresponding to peaks e and i–k contained an N-oxide since sodium metabisulfite readily reduces N-oxides to the corresponding amine¹⁹ and that the remaining metabolites contained only hydroxylations on the adamantane substructure.

These metabolite peaks were further characterized by MS/MS, and collision-induced dissociation (CID) spectra are shown in Figure S1. The characteristic MS/MS fragmentation pattern of the spiroadamantane ozonides provided an initial indication of the position of oxygen incorporation upon incubation with HLMs. Peroxide bond scission with subsequent rearrangement resulted in loss of the adamantane lactone and a corresponding mass loss of 166 Da¹⁵ (Figure S1A). Among the four +16 metabolites, one (peak e, Figure S1C) exhibited a fragment ion at m/z 320 corresponding to loss of 166 Da while the remaining three (peaks b–d, Figure S1B) exhibited fragment ions at m/z 304 corresponding to loss of 182 Da. This pattern indicates the presence of one site of oxygen incorporation at the cyclohexane side (peak e) and three sites of oxygen incorporation at the adamantane (peaks b–d). Based on the longer retention time compared to the other three +16 metabolites (Figure 2B and C), peak e was tentatively assigned as the morpholine N-oxide (later confirmed by the synthesis of 8).

For the six +32 metabolites, peaks f–h showed fragment ions at m/z 304 corresponding to loss of 198 Da (Figure S1D) consistent with incorporation of two oxygen atoms on the adamantane. In comparison, peaks i–k showed fragment ions at m/z 320 corresponding to the loss of 182 Da (Figure S1E) consistent with incorporation of one oxygen atom on either side of the ozonide bridge. The −26 metabolite (peak l) exhibited a loss of 166 Da to give a fragment ion at m/z 278 (Figure S1F), indicating the site of metabolism was on the cyclohexyl side chain (later confirmed to be metabolite 9). No other metabolic products arising from modifications to the morpholinoethoxyphenyl substructure were detected.

Synthesis of Authentic Metabolite Standards.

Our previous metabolite identification work with 1¹⁵ suggested that the two distal bridgehead carbons (C5 and C7) were likely to be two of the three sites of adamantane hydroxylation. X-ray crystallography studies (see the Supporting Information) of 3 and 4, the ester precursors of 5 and 6, respectively, confirmed the assignment of 5 as the trans, cis isomer and 6 as the cis, cis isomer (Figure S2); these were confirmed by the synthesis of 5 and 6 to be the monohydroxyl metabolites, respectively (Figure 3).

Figure 3.

Proposed CYP-mediated pathways for the metabolism of 2 based on in vitro studies with human liver microsomes. Major pathways are shown with bold arrows. The structures of metabolites 5–11 were confirmed by the synthesis of authentic standards, whereas the structures of the remaining metabolites (12–15) were inferred on the basis of fragmentation patterns, chemical derivatization and incubations of the primary metabolites. Refer to Figure 2 for chromatograms.

The location of the third hydroxylation site was deduced by considering the symmetry of the 2-spiroadamantane moiety in relation to the C5 and C7 positions. Located along the C2–C6 axis of symmetry of the adamantane substructure, C6 would be the only other site (in addition to C5 and C7) that affords the formation of only three dihydroxylated adamantane metabolites: a 1,3-diol and two vicinal diols. The synthesis of metabolite 7 (Figure 3) confirmed the third site of monohydroxylation on the adamantane substructure.

Microsomal incubation of metabolites 5 and 6 produced two different dihydroxylated metabolites for each compound (Figure S3A and B). In addition to the common product (peak f), additional peaks (g and h) were generated from 5 and 6, respectively. When each of the incubation mixtures from 5 and 6 were treated with sodium periodate, peak f was resistant to periodate cleavage (Figure S3C and D), suggesting that it was the 1,3-diol, whereas peaks g and h were consumed and two new peaks (m and n) were detected (Figure S3E and F) with masses corresponding to the ketocarboxylic acid products 16 and 17. Dihydroxylation metabolite 10 was synthesized and confirmed the structure of peak f as the 1,3-diol metabolite. The structures of metabolites corresponding to peaks g and h were therefore proposed to be 12 and 13, respectively (Figure 3). The proposed structures for peaks j (14) and k (15) were inferred based on their observed formation during incubations of preformed 7 and 6, respectively (Figure 3).

Authentic metabolite standards were assessed for their in vitro activity against P. falciparum (Table 1). The only metabolites with activity comparable to 2 were 8 and 9, both of which were metabolized on the morpholine, but not the adamantane, substructure. All other metabolites were much less active. Consistent with the general structure–activity relationship (SAR) trends for this class of antimalarial peroxides^20,21 and what we had previously observed for 1,¹⁵ the relatively polar metabolites of 2 with oxidized adamantane substructures were at least an order of magnitude less active than the parent compound.

The biotransformation of 2 when incubated with HLMs was dependent on the presence of NADPH with no degradation detected in the absence of cofactor. Reaction products resulting from Hock fragmentation²² of the ozonide heterocycle of 2 were not detected under the incubation conditions utilized. Following a 2 h incubation of 2 with NADPH-activated microsomes, there was approximately 92% consumption of the parent compound and 80–90% of this could be accounted for through the formation of the proposed metabolites (Table S1). Monohydroxyadamantane metabolites 5–7 and the secondary dihydroxyadamantane metabolites 10, 12, and 13 were the predominant species accounting for approximately 75% of the parent substrate consumed. The remaining metabolites were relatively minor with each representing less than 5% of the substrate consumed. No additional peaks were observed upon incubation of 2 with human cryopreserved hepatocytes.

The relative distribution of metabolites seen in this in vitro study was largely consistent with the results from the Phase I study with 2.⁸ Even though the volumes of distribution of the metabolites are not known, the primary metabolite 7 (OZ579) and the two secondary metabolites 10 (OZ580) and 13 (designated AA3 in⁸) appeared to be the predominant species based on the relative area under the plasma concentration versus time curves (AUC). The remaining metabolites, 6 (OZ567), 8 (OZ577), and 9 (OZ564) were detected but had comparatively lower AUC values compared to 7, 10, and 13 (note that a standard for 5 was not available at the time the Phase I study was conducted). Since 8 and 9, the only metabolites with notable antimalarial activity, were minor metabolites both in vitro and in humans, it can be concluded that the antimalarial activity resides almost exclusively with the parent molecule 2. This result is also consistent with that reported previously as part of the Phase I study⁸ where analysis of selected human serum samples obtained following dosing with 2 revealed similar concentrations using either a P. falciparum bioassay or LC/MS, again showing that biological activity is associated with the parent molecule.

Figure 4A shows the concentration dependency of the apparent first order rate constant for 2 depletion following incubation with HLMs. Rate constants reduced substantially as concentrations increased above approximately 0.2 μM with the data being consistent with a saturable metabolic process. Since subsequent studies (see below) indicated both reversible and irreversible CYP inhibition, it was not possible to deconvolute saturation and autoinhibition to determine the apparent Michaelis constant.

Figure 4.

(A) Apparent first order depletion rate constant (k_dep) as a function of 2 concentration. (B) Intrinsic clearance (CL_int) data for 2 following incubation with HLMs in the absence (blue) and presence (green) of specific CYP chemical inhibitors. Error bars represent the standard error of the estimated intrinsic clearance determined with duplicate measurements. There was no statistically significant difference (α = 0.05) between the control and inhibited data for any enzyme with the exception of CYP3A4/5 (*). (C) Effect of 2 preincubation conditions (±0.1 or 0.5 μM ketoconazole (KTZ), 4 mM reduced glutathione (GSH), and dialysis) in the absence and presence of NADPH on the activity of CYP3A4/5 using midazolam as a probe substrate. (D) CYP3A4/5 inactivation kinetics by 2 using MDZ or TST as a probe substrate. Symbols represent the experimental data for the observed inactivation rate constant at each concentration of 2 (±standard error of the estimate of the slope), and the lines represent the fits of the data to eq S1.

To identify the enzyme(s) responsible for metabolism, chemical inhibition studies were conducted and indicated that the metabolism of 2 is mediated exclusively by CYP3A with the intrinsic clearance reduced by more than 95% in the presence of the pan-CYP3A inhibitor, ketoconazole (Figure 4B). For all other isoforms tested, there was no statistically significant difference between the control and chemically inhibited intrinsic clearance values. In CYP3A5-genotyped microsomes (Table S2), there was a significant reduction in the rate of degradation using both *1/*1 and *3/*3 microsomes in the presence of ketoconazole, with >94% of the metabolism due to CYP3A. In the *1/*1 donor (containing functional CYP3A5), the extent of inhibition in the presence of the specific CYP3A4 inhibitor, CYP3cide,²³ was comparable to that observed in the presence of ketoconazole. Based on these results, CYP3A5 does not appear to be a major contributor to the CYP3A-mediated metabolism of 2. The predominant role of CYP3A4 raises the potential for in vivo drug–drug interactions between 2 and partner compounds that are also CYP3A4 substrates.

The potential for 2 to inhibit CYP enzymes was first assessed using a direct CYP inhibition assay, where 2 did not inhibit metabolite formation for pathways mediated by CYP1A2,²⁴ 2B6, 2C8, 2C9,²⁴ 2C19,²⁴ or 2D6²⁴ (IC₅₀ > 20 μM for each isoform with no measurable inhibition). For CYP3A, 2 inhibited the metabolism of both midazolam and testosterone with IC₅₀ values of 9.8 and 3.9 μM, respectively (Table 2).

Table 2.

Direct and Time-Dependent Inhibition of Cytochrome P450 Enzymes Following Incubation of 2 with HLMs^a

parameter	midazolam	testosterone
direct inhibition
CYP3A IC50 (μM)	9.8	3.9
time-dependent inhibition
CYP3A IC50 fold-shift	>25	20
kinetic parameters for time-dependent inhibition of CYP3A4/5
kinact (min−1)	0.074 (0.063–0.086)	0.10 (0.094–0.11)
KI (μM)	9.6 (5.0–17)	13 (9.4–19)
kinact/KI (mL/min/μmol)	7.7	7.7

^aKinetic inactivation parameters for time-dependent inhibition of CYP3A by 2 obtained by varying the preincubation 2 concentration and time prior to assessment of enzyme activity. Values for k_inact and K_I represent the estimated parameters with the 95% confidence interval in parentheses based on nonlinear curve fitting of the apparent inactivation rate constant versus inhibitor concentration to eq S1.

Further studies were conducted to assess the potential for time-dependent or mechanism-based inhibition. These studies used a preincubation and dilution procedure, where the potential for reversible competitive inhibition by 2 and formed metabolites was minimized, and the assumption was made that any time-dependent inhibition was due to 2 and not to any of the metabolites. Utilizing the “IC₅₀ shift” method, no time-dependent inhibition was detected for CYP1A2, 2B6, 2C8, 2C9, 2C19, or 2D6 with IC₅₀ values being >20 μM when the preincubation with HLMs was conducted in the absence or presence of the cofactor, NADPH. However, for CYP3A, the IC₅₀ shifted more than 20-fold to lower values when 2 was preincubated with HLMs in the presence of NADPH compared to the absence (Table 2), indicating the presence of time-dependent inhibition.

Additional experiments were conducted to confirm whether 2 is a mechanism-based inhibitor according to established criteria^25,26 (Figure 4C). First, there was no change in CYP3A activity when 2 was preincubated with HLMs in the absence of NADPH over 30 min confirming that the time-dependent inhibition is indeed associated with NADPH-dependent 2 metabolism. For preincubations containing only 2, CYP3A activity was reduced by approximately 40% in the absence of NADPH reflecting direct (i.e., reversible) inhibition and 80% when NADPH was included in the incubation. Second, inclusion of a competitive CYP3A inhibitor, ketoconazole, protected against the NADPH-dependent inactivation of the enzyme by 2 in a concentration dependent manner with >90% enzyme activity recovered when 0.5 μM ketoconazole was included in the preincubation mixture. Third, inclusion of an electrophilic scavenging agent, reduced glutathione (GSH), had no impact on the extent of inactivation suggesting that inactivation occurs prior to release of the reactive species from the active site of the enzyme. And finally, dialysis of the preincubation mixtures containing 2 and HLMs (±NADPH) was conducted prior to dilution and measurement of enzyme activity. In the absence of NADPH, dialysis led to recovery of enzyme activity to approximately 80% of the control value; however, there was no attenuation of the NADPH-dependent inactivation of CYP3A with enzyme activity inhibited by approximately 80–90%. This indicated that the inhibiting specie(s) formed during the preincubation of 2 with NADPH-activated HLMs was not freely diffusible, and by inference, was covalently bound to the enzyme. Collectively, these results strongly suggested that 2 is a mechanism-based inhibitor of CYP3A.

The kinetic properties of CYP3A inactivation were assessed by measuring the percent activity remaining as a function of 2 concentration and preincubation time. The apparent first-order inactivation rate constants were obtained at each 2 concentration (Figure S4) and then plotted against the 2 concentration (Figure 4D) to yield values for K_I (9.6–13 μM), the concentration giving the half maximal rate of inactivation, and k_inact (0.074–0.10 min⁻¹), the maximal inactivation rate constant as shown in Table 2. As described previously,^25,27 the ratio of k_inact/K_I gives a measure of the inactivation efficiency (Table 2). While additional data (e.g., unbound maximum hepatic inlet concentration) would be needed to accurately predict in vivo drug–drug interactions, comparison of the k_inact/K_I to that for known inhibitors²⁷ suggests that 2 is a moderate inhibitor with potency against CYP3A similar to that of diltiazem or erythromycin and is considerably less potent than the strong inhibitor, ritonavir. This risk as a perpetrator will be further explored in parallel studies using PBPK simulations of 2 (at efficacious doses) with typical CYP3A substrates such as midazolam or antimalarial CYP3A substrates.

These investigations describe the identification of the CYP-mediated metabolites of the antimalarial 2. Further, the results confirmed that metabolism occurs exclusively by CYP3A4 with minimal, if any, contribution from CYP3A5 or other CYP enzymes. The work also confirmed that in vivo antimalarial activity of 2 is almost exclusively associated with the parent molecule, with the more polar metabolites having minimal antimalarial activity consistent with previous SAR trends.^20,21 The data demonstrated that 2 inhibits CYP3A via both reversible and mechanism-based inhibition, with k_inact and K_I values that indicate it is a moderate inhibitor relative to other known CYP3A mechanism-based inhibitors. Both saturable metabolism and autoinhibition of CYP3A likely contribute to concentration-dependent degradation kinetics of 2 in HLMs as well as to the dose-dependent kinetics seen in a volunteer infection study where there was a 50% reduction in apparent clearance with an increase in dose from 100 to 500 mg.¹⁰

Given that current recommendations for treatment of blood-stage infections are for double, or possibly even triple, combination products to reduce the likelihood of developing resistance, information regarding the enzyme responsible for metabolism and potential for CYP inhibition is critical for the conduct of physiologically based pharmacokinetic (PBPK) modeling and the prediction of in vivo drug–drug interactions between combination partners or with coadministered drugs.

METHODS

Materials.

2 (lot# OZ-439/PC-276/02) was synthesized as the mesylate salt by Unimark Remedies Ltd. (Kerala, India). Pooled HLMs from anonymous male and female donors were sourced from either BD Gentest (Discovery Labware Inc., Woburn, MA) or Sekisui XenoTech LLC (Kansas City, KS). All other chemicals, reagents, and solvents used were of analytical or HPLC grade.

In Vitro Metabolism of 2.

For metabolite detection and identification, 2 (2 μM) was incubated with HLMs (1 mg/mL) for 2 h according to previously published methods.²⁴ The concentration dependence of 2 depletion was assessed by incubating with HLMs (0.4 mg/mL) at nominal 2 concentrations ranging from 0.02 to 25 μM, and reactions were quenched at seven time points over 60 min and analyzed by LC/MS.

Chromatographic and Mass Spectrometric Characterization of Metabolites.

LC/MS analysis was conducted using a Waters Micromass Quattro Ultima Pt triple quadrupole mass spectrometer coupled to a Waters Acquity UPLC system (Waters Corporation, Milford, MA) (see the Supporting Information for conditions).

Chemical Derivatization Studies.

2 was incubated with HLMs as described, the reaction was quenched by the addition of two volumes of methanol, and samples were vortexed and centrifuged. For N-oxide reduction, an aliquot of the supernatant was treated with excess sodium metabisulfite (10 mg/mL in water) for 30 min at room temperature and analyzed by LC/MS.

Authentic standards for metabolites 5 and 6 were incubated with HLMs as described above, and the reactions were quenched by the addition of 2 volumes of methanol before the sample was vortexed and centrifuged. For oxidative cleavage of the vicinal diols, an aliquot of the supernatant was treated with excess sodium periodate (50 mg/mL in water) at room temperature for 1 h followed by centrifugation and analysis by LC/MS.

Synthesis of Authentic Metabolites.

Monohydroxylated metabolites 5 and 6 were synthesized as follows (Figure S5A). A 4:1 mixture of ozonide esters 3 and 4 was obtained in 51% yield by Griesbaum coozonolysis²⁸ of oxime ether 18¹⁵ and keto acetate 19. The mixture of 3 and 4 was readily separated by two fractional crystallizations from 1:1 and 5:1 DCM/ ethanol. These crystallization conditions afforded suitable crystals for X-ray analysis of both 3 and 4 (Supporting Information), which established 3 as the trans, cis diastereomer and 4 as the cis, cis diastereomer (Figure S2). A one-pot acetate hydrolysis/alkylation (N-(2-chloroethyl) morpholine) of 3 and 4 followed by benzoate ester hydrolysis and salt formation with methanesulfonic acid (MSA) afforded 5 and 6, respectively, in high yields.

Monohydroxylated metabolite 7 was obtained starting with a three-step conversion of ethylene ketal carbinol 20²⁹ to oxime ether acetate 23 (Figure S5B). Griesbaum coozonolysis²⁸ of the latter and keto acetate 19 afforded ozonide diacetate 24 (48% yield) which was converted to 7 in high yield by successive acetate ester hydrolysis, alkylation (N-(2-chloroethyl)morpholine), and salt formation with MSA.

Dihydroxylated metabolite 10 was obtained starting with a two-step conversion of 5,7-dihydroxyadamantan-2-one 25³⁰ to oxime ether acetate 27 (Figure S5C). Griesbaum coozonolysis²⁸ of the latter in the presence of keto acetate 19 afforded ozonide triacetate 28 in 29% yield. Regioselective acetate hydrolysis of 28 with hydrazine afforded 29 which was converted to 10 in 75% yield by sequential one-pot alkylation (N-(2-chloroethyl) morpholine) and acetate ester hydrolysis. N-Oxide metabolites 8 and 11 were obtained in 93 and 96% yields by treatment of the free bases of 2 and 5 with a solution of dimethyldioxirane³¹ in acetone. The morpholine cleavage product 9 was obtained as previously described.³²

In Vitro Activity of Metabolites against P. falciparum.

Synthesized metabolites were screened against chloroquine-resistant (K1) and chloroquine-sensitive (NF54) strains of P. falciparum in vitro as previously described.²

CYP Reaction Phenotyping.

2 (1 μM) (±isoform specific inhibitors) was incubated in duplicate with HLMs (0.5 mg/mL) at 37 °C for 10 min prior to the addition of an NADPH-regenerating system (see the Supporting Information).

To delineate the contributions of CYP3A4 and CYP3A5, 2 (1 μM) was incubated at with CYP3A5-genotyped HLMs (0.5 mg/mL) with or without functional CYP3A5 activity (*1/*1 and *3*3 alleles, respectively). Incubations were performed for 60 min in the absence and presence of a pan-CYP3A inhibitor (ketoconazole) or a specific CYP3A4 inhibitor (CYP3cide²³). The relative loss of 2 was monitored by LC/MS. See the Supporting Information for further details.

Inhibition of CYP Enzymes by 2 in Human Liver Microsomes.

The direct CYP inhibition assay³³ was conducted using HLMs and a substrate-specific interaction as described previously (see the Supporting Information). To assess time-dependent inhibition, an “IC₅₀ shift” method was used²⁷ (see the Supporting Information).

To confirm the irreversible nature of the inhibition of CYP3A by 2, preincubations (30 min) were conducted with 2 (33.3 μM) ± NADPH with the addition of ketoconazole (0.1 or 0.5 μM) or reduced glutathione (4 mM). The remainder of the assay was conducted as described for the direct inhibition following a 10-fold dilution and using midazolam as the probe substrate to determine enzyme activity. For selected preincubation conditions (e.g., no inhibitor control and 2 (33.3 μM) ± NADPH), the preincubation mixture was dialyzed against 20% v/v human plasma (to reduce the impact of nonspecific adsorption of 2) using Slide-a-Lyzer dialysis units (10 000 MWCO; Thermo Fisher Scientific, Rockford, IL). Following the 27 h dialysis period, each of the reaction mixtures was diluted 10-fold and enzyme activity assessed as described above using midazolam as the probe substrate.

Inactivation kinetic parameters (K_I and k_inact) for 2 were obtained by varying the preincubation concentration and time in the presence of NADPH prior to dilution and addition of the probe substrate for assessment of enzyme activity (see the Supporting Information).

ABBREVIATIONS

CID

collision-induced dissociation

MSA

methanesulfonic acid

HLMs

human liver microsomes

CYP

cytochrome P450

MDZ

midazolam

TST

testosterone

CL_int

intrinsic clearance