The Metabolism of Separase Inhibitor Sepin-1 in Human, Mouse, and Rat Liver Microsomes

Separase, a known oncogene, is widely overexpressed in numerous human tumors of breast, bone, brain, blood, and prostate. Separase is an emerging target for cancer therapy, and separase enzymatic inhibitors such as sepin-1 are currently being developed to treat separase-overexpressed tumors. Drug metabolism plays a critical role in the efficacy and safety of drug development, as well as possible drug–drug interactions. In this study, we investigated the in vitro metabolism of sepin-1 in human, mouse, and rat liver microsomes (RLM) using metabolomic approaches. In human liver microsomes (HLM), we identified seven metabolites including one cysteine–sepin-1 adduct and one glutathione–sepin-1 adduct. All the sepin-1 metabolites in HLM were also found in both mouse and RLM. Using recombinant CYP450 isoenzymes, we demonstrated that multiple enzymes contributed to the metabolism of sepin-1, including CYP2D6 and CYP3A4 as the major metabolizing enzymes. Inhibitory effects of sepin-1 on seven major CYP450s were also evaluated using the corresponding substrates recommended by the US Food and Drug Administration. Our studies indicated that sepin-1 moderately inhibits CYP1A2, CYP2C19, and CYP3A4 with IC50 < 10 μM but weakly inhibits CYP2B6, CYP2C8/9, and CYP2D6 with IC50 > 10 μM. This information can be used to optimize the structures of sepin-1 for more suitable pharmacological properties and to predict the possible sepin-1 interactions with other chemotherapeutic drugs.


INTRODUCTION
Separase is an enzyme that resolves chromosomal cohesion and centriole engagement during mitosis. It is a cysteine protease in the CD clan (Uhlmann et al., 2000), with a catalytic domain structure similar to that of caspase (Viadiu et al., 2005;Winter et al., 2015;Lin et al., 2016). Although the N-terminus of separase varies, its C-terminus containing the proteolytic-active site is conserved from yeast to humans (Zhang and Pati, 2017). The canonical role of separase in the cell cycle is to cleave cohesin Rad21 at the onset of anaphase to separate sister chromatids. Separase is also required for centrosome duplication by cleaving centrosomal Rad21 (Nakamura et al., 2009;Tsou et al., 2009;Schockel et al., 2011) and pericentrin/kendrin (Lee and Rhee, 2012;Matsuo et al., 2012) to disengage centrioles. In addition, separase is involved in DNA damage repair (Nagao et al., 2004;McAleenan et al., 2013) and vesicle trafficking (Bembenek et al., 2007;Richie et al., 2011;Moschou et al., 2013Moschou et al., , 2016Bai and Bembenek, 2017). Due to its importance in numerous cell processes, the activity of separate is tightly regulated (Ciosk et al., 1998;Zou et al., 1999;Stemmann et al., 2001;Gorr et al., 2005;Holland and Taylor, 2006). Notably, separase is overexpressed in many human cancers of breast, bone, brain, blood, and prostate (Pati, 2008;Meyer et al., 2009;Mukherjee et al., 2014a,b;Zhang and Pati, 2017). Overexpression of separase induces aneuploidy and tumorigenesis in mouse models (Zhang et al., 2008;Mukherjee et al., 2014b). Using a highthroughput screen, we have identified a novel small molecular inhibitor of separase, named sepin-1, which inhibits separase activity in a non-competitive way (Zhang et al., 2014). Sepin-1 selectively inhibits the growth of cancer cell lines including breast cancer, leukemia, and neuroblastoma. It also inhibits the growth of breast cancer xenografts in mice. Sepin-1 induces apoptosis and its effect on the inhibition of cell growth is positively correlated to the level of separase in the cancer cells and tumors (Zhang et al., 2014). It suggests that sepin-1 possesses a great potential to be used for cancer treatment, particularly to treat separase-overexpressed tumors. Additionally, the use of inhibitors could resolve the non-canonical functions of separase as well.
In drug development, a drug's metabolism plays a critical role in its efficacy and safety. As a preclinical evaluation of sepin-1, here we have profiled the Phase I metabolism and bioactivation of sepin-1 in HLM (Uhlmann et al., 2000), MLM, and RLM, using metabolomic approaches, which have been shown to be powerful tools for studying drug metabolism (Li et al., 2014;Liu et al., 2015Liu et al., , 2016. The metabolic enzymes contributing to the metabolism of sepin-1 were identified using recombinant CYP450s, and the possible reactive metabolites were investigated in HLM using glutathione (GSH) as the trapping agent. The inhibitory effects of sepin-1 on seven common CYP450s were also evaluated. These results can be used further to optimize the structures of sepin-1, resulting in more suitable pharmacological properties and reduced possible toxicity, as well as to predict the metabolism-mediated possible interactions of sepin-1 with other chemotherapeutic drugs.
Metabolism of Sepin-1 in HLM, MLM, RLM, and Recombinant CYP450s Incubations were conducted in 1× phosphate-buffered saline (1× PBS, pH 7.4), containing 30 µM sepin-1, 1.0 mg HLM, MLM, RLM, or 2 pmol of each cDNA-expressed P450 enzyme (control, CYP1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, and CYP3A4) in a final volume of 190 µl. After 5 min of pre-incubation at 37 • C, the reaction was initiated by adding 10 µl of 20 mM NADPH (final concentration 1.0 mM) and continued for 30 min, with shaking at 37 • C. Incubations lacking NADPH served as controls. Reactions were terminated with 200 µl of ice-cold methanol, followed by vortexing for 30 s and centrifuging at 15,000 × g for 15 min. Each supernatant was transferred to an auto sampler vial, and 5.0 µl was injected on to UHPLC coupled with a QTOFMS system for metabolite analysis. Incubations were conducted in quadruplicate for HLM, in triplicate for MLM and RLM, and in duplicate for cDNA-expressed P450 enzymes.

Trapping Reactive Metabolites Using Glutathione
The reactive metabolites were trapped with GSH in our current study. The experiments were conducted in 1× PBS (pH 7.4), containing 30 µM sepin-1, 1.0 mg HLM, and GSH (2.5 mM) in a final volume of 190 µl. After 5 min of pre-incubation at 37 • C, the reactions were initiated by the addition of 10 µl of 20 mM NADPH (final concentration 1.0 mM) and continued for 30 min with gentle shaking. Incubations in the absence of NADPH and trapping agents were used as controls. The reactions were quenched by adding 200 µl of ice-cold methanol. The mixtures were vortexed for one 30 s and centrifuged at 15,000 × g for 15 min. The supernatants were transferred to sample vials for analysis. Incubations were performed in triplicate.

UHPLC-QTOFMS Analyses
The separation of sepin-1 and its metabolites was achieved using a 1260 Infinity Binary LC System (Agilent Technologies, Santa Clara, CA, United States) equipped with 100 mm × 2.1 mm (Agilent XDB C18) column. The column temperature was maintained at 40 • C. The flow rate was 0.3 ml/min, with a gradient ranging from 2% to 98% aqueous acetonitrile containing 0.1% formic acid in a 15-min run. QTOFMS was operated in a positive mode with electrospray ionization. Ultra-high pure nitrogen was applied as the drying gas (12 l/min) and the collision gas. The drying gas temperature was set at 325 • C, and the nebulizer pressure was kept at 35 psi. The capillary voltages were set at 3.5 kV. During MS, real-time mass correction and accurate mass were achieved by continuously measuring standard reference ions at m/z 121.0508 and 922.0098 in the positive mode. The MS/MS of sepin-1 metabolites was performed in a targeted mode with a default isolation width of m/z 4 and collision energy ramp ranging from 10 to 45 V.

UHPLC-QQQMS Analyses
The separations of each enzyme substrate and its specific metabolite were achieved using a 1260 Infinity Binary LC System (Agilent Technologies, Santa Clara, CA, United States) equipped with the 50 mm × 4.6 mm (Agilent XDB C18). The flow rate was 0.3 ml/min, and the mobile phases were water and acetonitrile with 0.1% formic acid. QQQMS was operated in a positive mode with electrospray ionization. Ultra-high pure nitrogen was applied as the drying gas (14 l/min) and the collision gas. The drying gas temperature was set at 280 • C and the nebulizer pressure was kept at 20 psi. The capillary voltages were set at 3.6 kV for positive mode and 3.0 kV for negative mode. The MRM transitions for the specific metabolites are listed in Table 1. The second transitions were used for confirmation purposes.

Data Analysis
Mass chromatograms and mass spectra were acquired by MassHunter Workstation Data Acquisition Software (Agilent, Santa Clara, CA, United States) in centroid and profile formats from m/z 100 to 1000. The acquisition rate was set as 1.5 spectra per second. Centroid and integrated mass chromatographic data were processed by Mass Profinder and Mass Profiler Professional Software (Agilent, Santa Clara, CA, United States) to generate a multivariate data matrix. The corresponding data matrices were then exported into SIMCA13 (Umetrics, Kinnelon, NJ, United States) for multivariate data analysis. OPLS-DA was conducted on Pareto-scaled data. For the QQQMS data, QQQ Quantitative Analysis Software (Agilent, Santa Clara, CA, United States) was used for metabolite analysis. For chemometric analysis, matrix data were processed from m/z 50 to 600. The experimental data are presented as mean ± SEM. Statistical differences between two groups were determined by Student's t-test.

Profiling of Sepin-1 Metabolism in HLM Using a Metabolomic Approach
The results of metabolomic analysis on the ions generated from UHPLC-QTOFMS analysis of control and sepin-1 group are shown in Supplementary Figure S1. Metabolomic analysis unraveled two clusters (Supplementary Figure S1A) corresponding to the control and sepin-1 group in the score plots, which indicated the chemical components are different between control and sepin-1 groups. The S-plot generated from OPLS-DA displays ion contribution to group separation in HLM (Supplementary Figure S1B). The top ranking ions were identified as sepin-1 metabolites, which were marked with M1-6 in the S-plot (Supplementary Figure S1B). The majority of the sepin-1 metabolites in HLM were also found in MLM and RLM (Figure 1). The information associated with sepin-1 metabolites is summarized in Table 1. The relative abundance of metabolites in HLM, MLM, and RLM is presented in Figure 1. Overall, seven sepin-1 metabolites and adducts, including one cysteine-sepin-1 adduct (M6, Cys-sepin-1) and one GSH-sepin-1 adduct (M7, GSH-sepin-1), were identified in HLM.

Identification of Reduced Metabolites (M1 and M2) in Sepin-1 Metabolism
In the incubation of sepin-1 in HLM, two major reduced metabolites (M1 and M2) were observed and characterized (Figure 2). The formation of metabolites M1 and M2 is NADPH dependent (Figure 2A). M1 was eluted at 4.78 min (Figure 2A), at 6.50 min (Figure 3A), also had a protonated molecule [M+H] + at m/z = 385. The fragmental ions at m/z 339 suggested that denitration has occurred. The other fragmental ion at m/z 324 is interpreted in the inlaid structural diagram ( Figure 3D).

DISCUSSION
In this study, we employed a LC-MS-based metabolomic approach to profile sepin-1 (a potent separase inhibitor and an investigational chemotherapeutic drug candidate) metabolism and bioactivation in HLM, MLM, and RLM. Metabolomic approaches could readily screen out the sepin-1-related metabolites and reactive intermediates from the biomatrix (Li et al., 2011(Li et al., , 2012. We identified seven sepin-1 metabolites and adducts, including reduced metabolites M1 and M2, dimer metabolites M3-M5, and cysteine-sepin-1 adduct (Cys-sepin-1) M6 and GSH-sepin-1 adduct M7, using the metabolomic technology. All the metabolites were presented in HLM, MLM, and RLM. M4 was the most abundant dimer in RLM, but only a trace amount was observed in HLM. However, the Cys-sepin-1 adduct M6 was mainly formed in HLM and MLM ( Figure 1B). M1 and M2 are dominant metabolites in all the three species of HLM, MLM, and RLM (Figure 1). Among these metabolites, one Cys-sepin-1 adduct (M6) was detected and characterized. The formation of M6 was also NADPH dependent. In HLM, sepin-1 is metabolized rapidly to the dominant-reduced metabolites M1 and M2. Therefore, the most abundant metabolite M1 was synthesized (Do et al., 2016) and the inhibitory activity of M1 on separase will be assayed next.
The reactive metabolites play a critical role in the pathogenesis of idiosyncratic adverse drug reactions (Attia, 2010;   Park et al., 2011;Thompson et al., 2016). The formation of Cys-sepin-1 adduct (M6) indicated that sepin-1 could be bioactivated to generate the reactive metabolite in HLM. The mechanism of the formation of Cys-sepin-1 adduct in HLM is not known yet, but the adduct M6 is derived from the major metabolite M1 by conjugating with cysteine in HLM. Reduced GSH is an abundant physiological nucleophile and frequently used to capture soft electrophiles (e.g., epoxides, quinones, quinone imines, and quinone methides) in in vitro systems (Evans et al., 2004). Our study revealed one GSH-sepin-1 adduct (M7) in the incubations of sepin-1 in HLM with GSH. Although the mechanisms of formation of GSH-sepin-1 adducts in HLM are not clear, the formation of GSH-sepin-1 adduct further suggested that the sepin-1 could produce the reactive electrophiles in HLM. Generally, excess reactive electrophiles are capable of covalent binding to protein, DNA, and other biomolecules. It is thought that cellular function is compromised in certain cases, followed by organ toxicity (O'Brien et al., 2005). These reactive metabolites from sepin-1 may induce some adverse effects. In drug development, researchers attempt to reduce the risk factors of toxicity by minimizing the formation of reactive metabolites. This information will be helpful for medicinal chemists to further optimize the structure of sepin-1 to improve its safety. For example, medicinal chemists could decrease the formation of reactive metabolites by blocking their metabolic sites via the structural modifications.
The mechanisms of sepin-1 reduction (M1 and M2) and dimer formation (M3-M5) are not clear yet. The liver microsomes involved in the formation of reduced metabolite is known (Leskovac and Popovic, 1980;Placidi et al., 1993;Wang et al., 2005). Our study suggested that multiple P450s enzymes contribute to sepin-1 metabolism. CYP3A4 and CYP2D6 are the primary enzymes responsible for the generation of metabolites M2-M5 ( Table 2). For M1, all the test isoenzymes make similar contributions to the formation of M1. Because CYP3A4 is the most abundant drug-metabolizing enzyme in the liver, the formation of M1 should be mainly mediated by CYP3A4. The role of CYP3A4 in the formation of M1 can be investigated further using Cyp3a-null mice. As the GSH-sepin-1 adduct was formed by GSH conjugating with the major metabolite M1, the formation of GSH-sepin-1 could have been mediated by multiple P450 enzymes as well. Other enzymes (e.g., cytochrome b5 reductase in the microsomes) have been reported to be implicated in the reduction of N-oxide (Zheng et al., 2011). In addition, the reduction of N-oxide (e.g., quinoxaline-1,4-dioxide) could be mediated by aldehyde oxidase and xanthine oxidase in the cytosol as well (Zheng et al., 2011;Mu et al., 2014). The exact role of P450s and reductases in the sepin-1 reduction will be further investigated using CYPP450 inhibitor (1H-benzotriazol-1-amine and carbon monoxide) and the liver microsomes from CYP450 reductase-knock out mice. Non-enzymatic reduction of N-oxide by heme of P450 will be tested using the boiled microsomes as well in the future study (Takekawa et al., 2001).
In summary, the metabolism of sepin-1 in HLM, MLM, and RLM was extensively studied using LC-MS-based metabolomic approaches. This study identified a total of seven metabolites and adducts related to sepin-1 (Figure 6) including Cys-sepin-1 and GSH-sepin-1 adducts. The enzyme contributing to the formation of sepin-1 metabolism was identified using recombinant P450 isoenzymes. Additionally, the inhibitory effects of sepin-1 on the seven main CYP450 isoenzymes were evaluated. This study provides the foundation for optimizing the structures and predicting the possible drug-drug interactions from the metabolic prospective of sepin-1, our lead separase inhibitor for cancer therapy. Further studies are suggested to illustrate the role of other enzymes contributing to reduced metabolites of sepin-1, metabolism of sepin-1 in vivo, and the inductive effects of sepin-1 on CYP450 isoenzymes in vitro and in vivo.

AUTHOR CONTRIBUTIONS
DP and FL participated in the research design. NZ, SG, SRG, and FL conducted the experiments. FL performed the data analysis. DP, NZ, and FL wrote or contributed to the writing of the manuscript.