Phase I metabolic profiling and unexpected reactive metabolites in human liver microsome incubations of X-376 using LC-MS/MS: bioactivation pathway elucidation and in silico toxicity studies of its metabolites

X-376 is a novel new generation anaplastic lymphoma kinase (ALK) inhibitor that can cross the blood brain barrier, so it can be used in patients with both ALK-positive NSCLC with CNS metastases. In this study, X-376 was incubated with human liver microsomes, and in vitro phase-I metabolic reactions were performed to generate X-376 reactive intermediates, which cannot be detected directly as they are unstable. Utilizing LC-MS/MS, we characterized X-376 metabolites and checked for the reactive electrophile generation using nucleophilic trapping agents, namely, potassium cyanide and GSH, which form stable adducts for characterization by mass spectrometry. Four X-376 phase-I metabolites and three reactive intermediates (one quinone methide and two iminium ions) were characterized and the bioactivation pathways were proposed. X-376 bioactivation occurred through unexpected novel pathways. Although the pyridazine ring was found to be bioactivated, no bioactivation was detected in the expected N-methyl piperazine ring that is often found for similar structures. The dichloro-phenyl group was bioactivated by a novel mechanism that was verified by LC-MS/MS. We propose that the X-376 reported side effects may be due to the formation of reactive metabolites. The in silico toxicity assessment of X-376 metabolites was carried out using DEREK software and structural modification were proposed to reduce their side effects and to validate the bioactivation pathway theory using StarDrop DEREK module. Further drug discovery studies can be done according to this concept, thus allowing the development of new drugs with more safety profile that was confirmed by using StarDrop software. To our knowledge, this is the first study of in vitro metabolic profiling or structural characterization and toxicological properties of the generated metabolites for X-376.


Introduction
Cancer is associated with high mortality, causing approximately one fourth of incident deaths worldwide. 1 Lately, the treatment strategies based on molecular targeting have been utilized in cases of disseminated cancer, depending on the identied tumor suppressor oncogenes and genes that participate in the patient's cancer progression. 2 Among all cancer types, lung cancer is considered to have the highest mortality rate worldwide. In 2012, lung cancer caused 1.59 million deaths, which represents 20% of all cancer deaths. 3 Non-small cell lung cancers (NSCLCs) represent 90% of all lung cancers and involve numerous subtypes driven by several activated oncogenes. 4,5 The treatment of NSCLC patients greatly improved when the targeted personalized therapies were used. [6][7][8] Anaplastic lymphoma kinase (ALK) belongs to the insulin receptor tyrosine kinase family (RTK). 9 ALK inhibitors are antineoplastic agents that act on tumors with rearrangements in ALK genes such as the EML4-ALK gene fusion. 10 EML4-ALK translocations have been found in 4-7% of NSCLC cases. 11 The rst ALK inhibitors generation (e.g., Crizotinib) has become less effective due to resistant mutations in NSCLCs that have become treatable by the development of a new ALK inhibitor generation (e.g., brigatinib, X-376 and X-396 (ensartinib)). [12][13][14] Compared with crizotinib, X-376 and X-396 inhibit ALK with greater potency in biochemical assays (10-folds) and cell-based assays (3-to 10-folds) with additional activity as MET inhibitors. Both X-376 and X-396 are being developed by Xcovery. 15 X-376 is considered to be a novel, more potent, and specic new generation ALK TKI with potential therapeutic relevance. It displays synergistic antitumor activity when combined with the mTOR inhibitor, rapamycin. It can also cross the blood brain barrier, so it can be used in patients with both ALK-positive NSCLC and CNS metastases. 12 Detoxication of drugs through metabolic reactions that increase the hydrophilicity of endogenous compounds and xenobiotics occurs to aid excretion outside the human body. The formed metabolites oen exhibit low toxicity in comparison to the precursor molecules. Though, in special cases, the generation of reactive intermediates occurs that exhibit more toxicity. [16][17][18] Reactive intermediates are electron decient, thus, they can modify proteins and DNA through covalent bonds, which is the initial step in drug-mediated organ toxicities. 19,20 The detection of reactive metabolite formation is an important task in understanding drug-induced toxicity. Reactive intermediates are usually generated by phase-I metabolic reactions and cannot be identied directly as they are unstable in nature. So, a trapping nucleophile can be utilized to interact with the reactive intermediates and capture the newly formed adducts, which are stable and suitable for characterization by LC-MS/ MS. 21,22 The designation of X-376 according to the International Union of Pure and Applied Chemistry is 6-amino-5- Fig. 1). The chemical structure of X-376 involves cyclic tertiary amine rings (piperazine) and a dichloro-uorophenyl ring. The piperazine ring can undergo bioactivation by iminium ion formation that can be captured using potassium cyanide. [23][24][25][26] The dichloro-uorophenyl ring can perform bioactivation by forming a quinone derivative that can be trapped by glutathione (GSH). 27 The adducts formed through nucleophilic-electrophilic reactions are considered stable, and can be isolated and identied by LC-MS/MS. [21][22][23]28,29 We hypothesize that these reactive electrophiles potentially participate in the reported side effects of X-376.
In this study, we utilized LC-MS/MS for in vitro X-376 metabolites' characterization that were formed on incubation with HLMs and then tested the generation of reactive electrophiles using trapping agents. Importantly, we found that X-376 bioactivation occurred through unexpected novel pathways, which was conrmed using in vitro laboratory work and in silico soware. This could be a new strategy for reducing the side effects of newly developed drugs without affecting their pharmacological activity, which was proposed using StarDrop soware. Knowing the bioactive center and structural alerts in the X-376 structure helped in making the targeted modications to improve its safety and retain its efficacy, which were conrmed using DEREK soware. Four in vitro phase I metabolites, two cyano adducts, and one GSH conjugate of X-376 were characterized. The proposed pathways involved oxidation, reduction, N-demethylation, and hydroxylation. The in silico toxicity assessment of X-376 metabolites was carried out using DEREK soware and structural modications were proposed to reduce their side effects and to validate the bioactivation pathway theory using StarDrop soware.

Chemicals
Chemicals and reagents, including their sources, are mentioned in Table 1.

Chromatography conditions
An Agilent 6410 triple quadrupole tted to an electrospray ionizer (ESI) with an Agilent 1200 rapid resolution liquid chromatographer (RRLC) was used. The optimized chromatographic and mass parameters were selected for X-376 and its related metabolites. Fragment ions for X-376 and its metabolites were formed inside the second hexapole (collision cell) by collision induced dissociation technique using high purity nitrogen. The selected chromatographic parameters for the separation of X-376 metabolites are shown in Table 2. The LC-MS/MS analytical parameters were adjusted to achieve the optimum resolution of X-376 metabolites with good sensitivity (Table 1). Slow gradient system with long run time was developed to separate X-376, phase I metabolites, and reactive metabolite adducts. A triple quadrupole (QqQ) mass analyzer, operated in the positive charge mode with an ESI, was utilized for estimation. Nitrogen gas (11 L min À1 ) was used for drying of spray in the ESI source and for collision (55 psi) for fragmentation studies in the collision cell. Flow injection analysis was utilized to optimize all the mass spectrometric parameters to achieve the highest ion intensity. The values of ESI temperature (T) and capillary voltage (V) were set at 350 C and 4000 V, respectively. Data acquisition was managed using the Mass Hunter soware. The  identication of X-376 metabolites was done using product ion (PI) scan mode for the mass transitions (parent to daughter ions). The fragmentor voltage (FV) was adjusted at 145 V with collision energy (CE) of 30 eV for X-376 metabolites. PI mode was used for structural identication to discard any interference from the HLM matrix constituents and to elevate the LC-MS/MS method sensitivity.

2.3.
In silico prediction of X-376 metabolites using P450 metabolism module of StarDrop soware Our objective was to identify the vulnerability of key sites of metabolism, as revealed by decreasing site lability that was indicated by composite site lability (CSL) and also the predicted regioselectivity of metabolism by the major isoforms supposed to be responsible for metabolism. The results from the WhichP450™ model are shown by the pie chart used for    Paper RSC Advances indication of the most likely cyp450 isoform that has a major role in X-376 metabolism.

2.4.
In silico prediction of the toxicity of X-376 metabolites and reactivity using DEREK soware The screening for the predicted toxicity of X-376 metabolites was performed using DEREK soware, which was also utilized to screen the structural alerts for X-376 so as to propose structural modications at the supposed bioactive centres that stop the toxicity sequence.

HLM incubation
The detection of X-376 generated metabolites was done by incubation of 30 mM X-376 with 1.0 mg mL À1 HLMs in 50 mM This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 5412-5427 | 5417

Identication of X-376 reactive metabolites
Aer a full MS scan over a specic range, the extracted ion chromatograms of selected molecular ion peaks (MIP) were used to locate the expected metabolites in the total ion chromatograms (TIC). Peak fractions containing the product ions with predicted identity were subjected to fragmentation using the product ion mode to generate daughter ions (DIs), which were needed to reconstruct the chemical structures of the original molecular ion. Thus, the fragmentation pattern of X-376 was utilized as a guide for the interpretation and conrmation of the supposed chemical structures of the reactive and in vitro metabolites formed during X-376 metabolism. X-376 HLM incubation was repeated using similar conditions by adding GSH or potassium cyanide for trapping the bioactive intermediates. To verify the results, all the experiments were repeated three times using controls (without HLMs or NADPH).

Results of the in silico X-376 metabolites' prediction
The metabolic landscape for X-376 indicates the lability of each site with respect to metabolism by CYP3A4 in absolute terms so as to guide the prediction of X-376 metabolites and also the optimization of chemical structures for improving the metabolic stability. This indicates that N34, C37, C33, and C35 on the N-methyl piperazine are predicted to be the most labile sites of metabolism, which matched with the experimental results. The CSL is shown at the top-right of the metabolic landscape. The results from the WhichP450™ model are shown in the pie chart used for the indication of the most likely cyp450 isoform that has a major role in X-376 metabolism (Fig. 2). Cyp3A4 was found to have a major role in X-376 metabolism.

Results of in silico X-376 structural alert sites and toxicity prediction
Based on the in silico predictions and the knowledge from literature, a list of probable metabolites and reactive intermediates was set. The expected potential atomic sites for the bioactivation and cyanide attack in the X-376 chemical structure are on two a carbon atoms adjacent to the nitrogen atoms of the piperazine ring (Fig. 3). The in silico toxicity assessment of X-376 metabolites was carried out using DEREK soware; structural modications were proposed to reduce their side effects and to validate the bioactivation pathway theory using StarDrop soware (Fig. 3). X-376 and its metabolites show structural alerts, as seen in Fig. 3, which gave rise to the proposed sites, as listed in Table 3. M3 and M4 did not exhibit phospholipidosis due to the presence of the oxidized piperazine ring. Groups that provide steric hinderance such as methyl (ensartinib) or oxygen (M3 and M4) stop the toxicity, as proposed by DEREK soware. X-376 and its metabolites show nephrotoxicity, carcinogenicity, and skin sensitization due to halogenated benzene, polyhalogenated aromatic, and activated benzene moieties. Table 3 shows the complete list of in vitro X-376 metabolites with DEREK results for the proposed toxicity prole. Small structural modication at the a carbon to N34 stopped the phospholipidosis toxicity (Fig. 3). This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 5412-5427 | 5419

Identication of X-376 related metabolites
Four X-376 metabolites were generated by four phase-I metabolic pathways (oxidation, N-demethylation, reduction, and hydroxylation). Two cyano adducts and one GSH adduct were identied (Table 4).
3.3.1. X-376 fragmentation pattern. The X-376 molecular ion peak (MIP) was eluted at 38.9 min in the total ion chromatogram (TIC) (Fig. 4A). Collision induced dissociation (CID) of the molecular ion at m/z 547 generated four characteristic DIs at m/z 447, 357, 257, and 138, indicating reactive bonds inside the X-376 chemical structure (Fig. 4B and Scheme 1). The DI at 3.3.2. M1 fragmentation pattern. The M1 MIP was eluted at 38.1 min in the TIC (Fig. 5A). The CID of the molecular ion at m/  formed four DIs at m/z 449, 359, 259, and 139 (Fig. 6). When compared to X-376, the fragments of M2 generated DIs with an increase of 2 m/z. The proposed metabolic reaction was the reduction of amide group attached to the pyridazine ring that matched with the DI at m/z 139 (Scheme 3).
3.3.4. M3 fragmentation pattern. The M3 MIP was eluted at 44.1 min in the TIC (Fig. 7A). The CID of the molecular ion at m/ z 561 formed three DIs at m/z 371, 257, and 138 (Fig. 7B). If compared to the X-376 fragments, the DI at m/z 371 was indicative of oxidation of the N-methyl piperazine group, which was conrmed by the other two DIs at m/z 257 and 138 (Scheme 4).
3.3.5. M4 fragmentation pattern. The M4 MIP was eluted at 42.0 min in the TIC (Fig. 8A). The CID of the molecular ion at m/ z 563 formed four DIs at m/z 447, 373, 257, and 138 (Fig. 8B). When compared to the X-376 fragments, the DI at m/z 373 was indicative of a hydroxylation metabolic pathway at the N-methyl piperazine group, which was veried by the two DIs at m/z 447 and 257 (Scheme 5).

Reactive metabolites
Two cyano and one GSH adduct were identied aer incubation of X-376 with HLMs and capturing agents (KCN or GSH).
3.4.1. M5 cyano adduct fragmentation pattern. The M5 MIP was eluted at 39.2 min in the TIC (Fig. 9A). The CID of the molecular ion at m/z 572 formed four DIs at m/z 472, 282, 120, and 104 (Fig. 9B). When compared to the X-376 fragments, the DIs at m/z 472 and 282 exhibited an increase of 25 m/z units, which revealed the addition of a cyano group at the bioactivated carbon of the pyridazine ring (Scheme 6).
3.4.2. M6 cyano adduct fragmentation pattern. The M6 MIP was eluted at 42.9 min in the TIC (Fig. 10A). The CID of the molecular ion at m/z 588 formed four DIs at m/z 472, 282, 120, and 104 (Fig. 10B). When compared to the X-376 fragments, the DIs at m/z 472 and 282 exhibited an increase of 25 m/z units, which revealed the addition of a cyano group. The metabolic pathways for M6 were hydroxylation at the piperazine ring and cyano addition at the bioactivated carbon of the pyridazine ring (Scheme 7).
3.4.3. M7 GSH conjugate fragmentation pattern. The M7 MIP was eluted at 28.9 min in the TIC (Fig. 11A). The CID of the molecular ion at m/z 852 formed four DIs at m/z 496, 257, 357, and 176 (Fig. 11B). When compared to the ESB fragments, the DIs at m/z 357 and 257 conrmed that the bioactivation and conjugation with GSH occurred at the dichlorophenyl group aer deuorination and hydroxylation (Scheme 8). This mechanism of bioactivation has been proposed for similar drugs (e.g., diclofenac). 27 Scheme 9 Supposed pathways for reactive iminium intermediate generation in X-376 metabolism and the possible capturing strategy.

Scheme 10
The proposed mechanism of GSH conjugation with quinone methide intermediate.

Mechanism of X-376 bioactivation
The generation of M5 and M6 cyanide adducts revealed the production of reactive iminium intermediates in the in vitro metabolism of X-376. The hydroxylation metabolic pathway at the bioactivated pyridazine ring in X-376, followed by dehydration, created unstable iminium intermediates but they could be trapped as stable adducts with the cyanide nucleophile (Scheme 9). The supposed mechanism of the iminium intermediate generation and X-376 bioactivation pathway has been previously reported on similar cyclic tertiary aminecontaining drugs. [32][33][34][35] The formation of M7 revealed the production of a quinone methide reactive intermediate in X-376 metabolism, which was formed by deuorination, hydroxylation, and oxidation, and then trapped by GSH (Scheme 10). 27  This journal is © The Royal Society of Chemistry 2020

Conclusions
Four metabolic reactions types, viz., oxidation, N-demethylation, reduction, and hydroxylation, generated four in vitro phase-I X-376 metabolites. Bioactivation initiated the formation of two cyano adducts and one GSH conjugate (Fig. 12). The piperazine ring was not bioactivated as originally predicted. However, bioactivation occurred in the pyridazine ring because of the formation of an electron decient system that could potentially bind to DNA or other important proteins inside the human body. The dichlorophenyl group was bioactivated by a novel pathway, which was conrmed using LC-MS/MS. These ndings provide guidance for further work on X-376 toxicity. Further drug discovery studies can be done depending on this concept, thus allowing the development of new series of drugs with increased safety prole without affecting their pharmacological action. Blocking or adding isosteric substituents to the available bioactive pyridazine ring moiety would likely interrupt or block metabolic hydroxylation, which will stop the bioactivation sequence. The in silico toxicological studies for X-376 and its metabolites were performed using DEREK soware, which revealed structural alerts and proposed side effects of X-376. Phospholipidosis was stopped as proposed by StarDrop DEREK module using small structural modications, as in the case of M3, M4, and ensartinib (Fig. 13).

Ethics
The study's design (in vitro assays using commercially available liver microsomes) exempts it from the approval by Ethics Committees.