Exemestane potency is unchanged by common nonsynonymous polymorphisms in CYP19A1: results of a novel anti‐aromatase activity assay examining exemestane and its derivatives

Abstract Exemestane (EXE) treats estrogen receptor positive (ER+) breast cancer in postmenopausal women by inhibiting the estrogen‐synthesizing cytochrome P450 CYP19A1. Variability in the severity and incidence of side effects as well as overall drug efficacy may be partially explained by genetic factors, including nonsynonymous variation in CYP19A1, also known as aromatase. The present study identified phase I EXE metabolites in human liver microsomes (HLM) and investigated mechanisms that may alter the extent of systemic estrogen deprivation in EXE‐treated women with breast cancer, including whether functional polymorphisms in aromatase cause differential inhibition by EXE and whether EXE metabolites possess anti‐aromatase activity. The potency of EXE and ten of its derivatives was measured with HEK293‐overexpressed wild type aromatase (CYP19A1*1) using a rapid novel UPLC tandem mass spectrometry method. Of the ten compounds assayed, five were poor inhibitors (IC 50 ˃ 50 μmol/L) of wild type aromatase while five others, including the major metabolite, 17β‐dihydroexemestane (17β‐DHE), exhibited moderate potency, with IC 50 values ranging between 1.2 and 7.1 μmol/L. The anti‐aromatase activity of EXE was also tested with two common allozymes, aromataseThr201Met (CYP19A1*3) and aromataseArg264Cys (CYP19A1*4). Differential inhibition of variant aromatase is unlikely to account for variable clinical outcomes as EXE‐mediated inhibition of aromataseThr201Met (IC 50 = 0.86 ± 0.12 μmol/L) and aromataseArg264Cys (IC 50 = 1.7 ± 0.65 μmol/L) did not significantly differ from wild type (IC 50 = 0.92 ± 0.17 μmol/L). Although less potent than the parent drug, these results suggest that active metabolites may contribute to the therapeutic mechanism of EXE.


Introduction
Exemestane is a synthetic androgen prescribed to postmenopausal women with ER+ breast cancer (Coombes et al. 2007;Paridaens et al. 2008). As an adjuvant endocrine therapy, EXE irreversibly inhibits the aromatasemediated production of estrogens from androgen precursors, a process known as aromatization (Hong et al. 2007). A previous pharmacokinetics study found that the maximum plasma concentration of EXE in postmenopausal women with a prior history of breast cancer ranged from 3.0 to 15.6 ng/mL following 2 weeks of oral dosing (25 mg/day) while the maximum amount of its 17b-DHE metabolite varied 7-fold with reported values of 0.22-1.58 ng/mL (Traina et al. 2008).
Prescriptive information states that EXE is extensively metabolized, in part by aldo-keto reductases (AKRs) (Pfizer, 2016). A key phase I metabolic pathway of EXE is C17 reduction to form a hydroxyl moiety vulnerable to phase II conjugation and excretion (Sun et al. 2010). Recent studies independently confirmed that five purified hepatic cytosolic reductases, AKRs 1C1-4 and carbonyl reductase 1 (CBR1), reduce EXE to the active metabolite 17b-DHE (Platt et al. 2016). Formation of 17a-dihydroexemestane (17a-DHE), a novel metabolite with unknown anti-aromatase activity (AAA), was catalyzed by AKR1C4 and CBR1 (Platt et al. 2016). A second metabolic pathway in human liver preparations is C6 exomethylene oxidation by CYP3A4 to form multiple secondary metabolites (Pfizer, 2016). The chemical structures of the C6-oxidized metabolites, as well as detailed information regarding their capacity to inhibit aromatase are omitted from the product leaflet dispensed with EXE tablets (Pfizer, 2016).
Several studies imply that EXE hepatic metabolism may be more complex than previously believed with possibly undiscovered metabolites and coaction by additional cytochrome P450s (CYP450s) (Kamdem et al. 2011;Platt et al. 2016). Comprehensively identifying phase I EXE metabolites is warranted, because EXE derivatives may contribute to systemic estrogen blockade through aromatase inhibition. The presence of 17b-DHE as a major metabolite in human plasma has been unequivocally confirmed in studies of postmenopausal women taking EXE (Evans et al. 1992;Traina et al. 2008). However, past attempts to identify less-studied metabolites have been speculative due to the lack of standard reference compounds. Using GC-MS, three peaks likely corresponding to C6-oxidized metabolites were detected in the urine of healthy male volunteers (Cavalcanti Gde et al. 2011). Another study found six metabolites, including 17b-DHE, in human urine following administration of radiolabeled EXE (Cocchiari et al. 1994). However, both studies of urinary EXE metabolites were hampered by a lack of comparison of physiochemical properties between the suspected metabolites and known standards. Six possible metabolite peaks were observed in human liver microsomes presented with EXE substrate (Kamdem et al. 2011). One peak was confirmed to be 17b-DHE and another was tentatively designated as 6-hydroxymethylandrosta-1,4,6-triene-3,17-dione (6-HME) (Kamdem et al. 2011). The identities of the remaining four peaks could not be established (Kamdem et al. 2011).
The current study addresses methodological issues that have historically undermined phase I EXE metabolite identification. First, a reference library of C6 and C17modified EXE analogs was synthesized to confirm the identity of suspected metabolites observed in incubations of EXE with human liver microsomes. Secondly, a newly developed UPLC/MS/MS method eliminates the need for organic extraction to remove residual substrate prior to analysis unlike previous scintillation-based studies of AAA (Thompson and Siiteri 1974). Instead, low levels of estrone formation are quantitated directly rather than extrapolated from tritiated water release during the aromatization of radiolabeled androstenedione. Interestingly, aromatase from human placental microsomes is used in traditional AAA screenings (Thompson and Siiteri 1974). CYP1A1 is well-expressed in human placenta and extensively metabolized EXE in an in vitro assay using recombinant baculosome-expressed CYP450s (Kamdem et al. 2011;Uhl en et al. 2015). Therefore, background phase I metabolism in human placental microsomes may complicate the analysis of AAA assays. However, expression analysis has shown that HEK293 are CYP450 and UDPglucuronosyltransferase (UGT)-null (data not shown). To circumvent potential confounding from endogenous enzymes in placental preparations, aromatase-overexpressing HEK293 were created in the present study to evaluate the potency of EXE analogs in impeding estrogen biosynthesis.
While it is well-accepted that genetic differences may influence an individual's drug disposition for many pharmaceuticals, the extent to which polymorphisms in aromatase explain interindividual variation in EXE potency is unclear. Interestingly, aromatase has several common nonsynonymous variants which might contribute to variability in drug disposition by altering its affinity for EXE (Ma et al. 2005), potentially affecting EXE efficacy or toxicity risk. Consequently, we also compared the efficacy of EXE in inhibiting two allozymes, aromatase Thr201Met and aromatase Arg264Cys relative to the wild type enzyme.

Reference library synthesis
EXE and ten C6-oxidized or C17-reduced EXE analogs were resuspended in ethanol and stored at À80°C following synthesis at Washington State University (Spokane, WA). Previous studies provide detailed descriptions of the synthesis, purification, and NMRbased identity verification of each compound (Buzzetti et al. 1993;Marcos-Escribano et al. 2009;Platt et al. 2016;Vat ele 2007).

Creation of aromatase-overexpressing HEK293
Stable overexpression of wild type aromatase in HEK293 was driven by a pcDNA3.1/V5-His-TOPO mammalian expression vector as previously described (Sun et al. 2010). Constitutive overexpression vectors encoding common aromatase variants Thr201Met and Arg264Cys were produced via site-directed mutagenesis using the wildtype plasmid as template. Variant expression vectors were amplified in BL21 grown under ampicillin selection for 16 h at 37°C. Sanger sequencing was used to confirm successful mutagenesis. Lipofectamine 2000 was used to transfect HEK293 with variant overexpression plasmids. Transfected HEK293 were grown in high-glucose DMEM containing 700 lg/mL G418, 10% FBS, and penicillin/ streptomycin for at least 3 weeks. The cells were then harvested by resuspension in PBS, lysed via 4 freeze-thaw cycles, and centrifuged for 15 min at 13,200g at 4°C. Microsomes for each cell line were prepared from the supernatant through differential centrifugation (1 h, 34000g) in a chilled Beckman L7-65 ultracentrifuge (Brea, CA), resuspended in PBS, and stored at À80°C. The relative expression of aromatase was quantitated in triplicate by subjecting 20 lg of protein from each overexpressing cell line to SDS-PAGE in a 10% tris-glycine polyacrylamide gel. Following transfer to PVDF for 90 min at 30 V, the membrane was blocked overnight at 4°C in 5% nonfat dry milk, washed for 30 min in 0.1% Tween, and probed overnight with anti-aromatase primary antibody (1:2500). The next day, the membrane was again washed for 30 min, and probed with HRP-conjugated goat antirabbit antibody (1:7500) for 1 h at ambient temperature. Following another 30 min wash, the blot was incubated with SuperSignal West Femto Maximum Sensitivity Substrate per the manufacturer instructions and imaged on a ChemiDoc Imager (BioRad, Hercules, CA). Image J software (NIH, Bethesda, MD) was used to measure band density while Ponceau staining was used to validate even loading between lanes.

EXE Metabolite Identification
A 50-ll incubation containing 50 lg of HLM in PBS (pH 7.4), 400 lmol/L EXE, and an NADPH regeneration system was placed in a 37°C water bath for 4 h before termination with 50 lL of cold acetonitrile. After a 15-min refrigerated centrifugation at 13,200g, the supernatant was examined for phase I EXE metabolites. A 10-min UPLC method was used to separate and detect EXE and the ten other reference compounds through multiple reaction monitoring with positive mode electrospray ionization on a Waters ACQUITY UPLC/MS/MS system (Milford, MA). The 1.7 lm ACQUITY UPLC BEH C18 column (2.1 mm 9 50 mm, Ireland) used for these analyses was protected by a 0.2 lm in-line filter. The UPLC gradient conditions used have previously been described (Platt et al. 2016). The fragmentation characteristics and retention time of suspected metabolite peaks were compared to compounds from the reference library.

Impact of nonsynonymous polymorphisms on EXE potency
IC 50 values describing EXE-mediated aromatase inhibition did not significantly differ (P = 0.71) between wild type enzyme (0.92 AE 0.17 lmol/L), aromatase Thr201Met (0.86 AE 0.12 lmol/L), and aromatase Arg264Cys (0.97 AE 0.09 lmol/L) in AAA assays normalized for relative aromatase expression (Fig. 2). Many aromatase polymorphisms exist, but data regarding the functional significance of variant alleles on human health is inconsistent (Baxter et al. 2001;Ma et al. 2005;Miyoshi et al. 2003). The prevalence of the Thr201Met allele is estimated as 5% in Caucasians and African Americans while the frequency of the Arg264Cys allele is 2.5% and 22.5% in Caucasian and African Americans respectively (Ma et al. 2005). One study of variant aromatase found that enzyme activity strongly correlated with expression levels in transiently transfected COS-1 and further concluded that any differences from wild-type in the overall activity of the Thr201Met and Arg264Cys allozymes are likely mediated by differential expression (Ma et al. 2005).

EXE metabolite identification
17b-DHE, 6-HME, 6a/b-hydroxy-6a/b-hydroxy-methylandrosta-1,4-diene-3,17-dione, and 6a/b,17b-dihydroxy-6a/b-hydroxymethyl-androsta-1,4-diene-3-one were identified in incubations of EXE with pooled human liver microsomes through comparison to reference compounds (Fig. 3). Although we found four EXE metabolites, an in vitro study of EXE metabolism by Kamdem et al. (2011) detected six peaks corresponding to putative metabolites. Our assay was not designed to identify phase II metabolites suggesting that the two additional peaks observed in the previous study may correspond to conjugated metabolites, such as the 17b-DHE-glucuronide produced by UGT2B17 (Sun et al. 2010). Considering their low abundance and limited capacity to inhibit aromatase in our novel AAA assay, the three C6-oxidized metabolites detected are unlikely to contribute to the overall pharmacology of EXE in vivo. However, these results show that 17b-DHE is not only the predominant EXE metabolite formed in human liver microsomes, but also capable of inhibiting aromatase with moderate potency suggesting that it may make clinically relevant contributions to the overall response to EXE in women with ER+ breast cancer (Platt et al. 2016).

Author Contribution
Participated in research design: Peterson and Lazarus. Conducted experiments: Peterson. Contributed new reagents or analytic tools: Xia, Chen, and Peterson. Performed data analysis: Peterson. Wrote or contributed to the writing of the manuscript: Peterson, Chen, Xia, and Lazarus.