Characterization of stress degradation products of blonanserin by UPLC-QTOF-tandem mass spectrometry

Abstract

Stress studies of drugs are very important in the drug development process. As per regulatory guidelines forced degradation studies and characterization of resulting degradation products is mandatory to establish inherent stability of the drug. Blonanserin is an important drug used for the treatment of schizophrenia. As there are no reports in the literature on the degradation study of the drug, the present work has been undertaken. Blonanserin was subjected to forced degradation studies under the conditions of hydrolysis (acidic, basic and neutral), oxidation, photolysis and thermal stress conditions. A selective separation was achieved on a Waters BEH C18 analytical column (50 mm × 2.1 mm, 1.7 μm). The structural characterization of the degradation products was performed using UPLC/QTOF/MS/MS. The drug was found to degrade in oxidative and photolytic conditions, whereas it was stable under hydrolytic, photolytic and thermal stress conditions. A total of seven hitherto unknown degradation products were characterized and probable mechanisms have been proposed for the formation of the degradation products. Moreover, in silico toxicity of all degradation products was also evaluated.

---

Introduction

Blonanserin (BLN) is a second-generation antipsychotic drug used in the treatment of schizophrenia with a superior efficacy to the first generation antipsychotics.^1–3 Schizophrenia is a diverse devastating psychiatric disorder that presents in late adolescence or early adulthood. It is associated with an increased risk of mortality and social or work-related dysfunction. BLN is the highly selective dopamine (D₂ and D₃) and serotonin (5-HT_2A) receptor antagonist and it has a low propensity to cause adverse effects such as orthostatic hypotension, oversedation, weight gain, metabolic abnormalities, and peripheral anticholinergic side effects.^4–6 It was observed that BLN has long-term efficacy and safety in patients with first-episode schizophrenia.⁷ It also has an effect on cognitive and social function in acute phase Japanese schizophrenia compared with risperidone.⁸ BLN is presently under clinical investigation in a Phase III trial in the People's Republic of China.¹

An extensive literature search revealed that only two analytical methods were reported for the determination of BLN in tablets using HPLC,^9,10 stability study by LC¹¹ and a few bioanalytical methods have been developed alone as well as in combination with other antipsychotic drugs in the human plasma using LC/MS/MS experiments.^12–17 To the best of our knowledge, characterization of stress degradation products of this drug has not been reported so far. Stress study is the foremost tool that is used to develop a selective stability indicating assay method (SIAM) that can selectively analyze the drugs and its degradation products (DPs) or impurities.¹⁸ Moreover, as per regulatory guidelines stress degradation studies and characterization of its degradation impurities at or above 0.05% should be carried out on the drug substance.^19,20 Structural characterization of DPs and mechanisms of their formation are useful in developing stable formulations and in the explanation of side effects of drugs.²¹ Moreover, as per regulatory guideline, it is essential to do biological evaluation of DPs which are coming above 0.15%.¹⁹ Hence, nowadays in silico technologies have been widely used by the pharmaceutical industry as a guard tool for early assessment of the toxic potential of candidate molecules or degradation products. In silico toxicology methods are practical and high throughput, with high accuracy.²² The National Research Council (NRC) recently published report stated that advancements in toxicogenomics, bioinformatics, systems biology, epigenetics, and computational toxicology could transform and change toxicity testing from a whole-animal system-based model to in silico methods. These days, hyphenated techniques such as LC/ESI/MS/MS combined with accurate mass measurements have been well established for identification and characterization of DPs.^23–27

Due to the significance of drug stability studies and the lack of information on BLN stability in the literature, we decided to investigate the degradation behavior of the drug under different stress conditions using ultra-high pressure liquid chromatography (UPLC) method and all degradation products were characterized by UPLC/ESI/MS/MS and UPLC/APCI/MS experiments. Finally, in silico toxicity including genotoxicity, an additional safety concern leading to significant risk for carcinogenicity of the characterized DPs using TOPKAT (toxicity prediction by computer-assisted technology) and DEREK (deductive estimate of risk from existing knowledge) softwares were evaluated.

Experimental

Chemicals and reagents

BLN (purity > 98% w/w) was purchased from Sigma-Aldrich, Bangalore, India. LC-MS CHROMASOLV® grade methanol (MeOH) and acetonitrile (ACN) were procured from Sigma-Aldrich (Bangalore, India). Analytical reagent (AR) grade ammonium acetate, ammonium formate, formic acid, hydrochloric acid (HCl), sodium hydroxide (NaOH) and acetic acid were purchased from S.D. Fine Chemicals (Mumbai, India). AR grade H₂O₂ was purchased from Merck (Mumbai, India). HPLC grade water was obtained by using Milli-Q Gradient system (Millipore, Bedford, MA, USA) and was used to prepare all solutions.

Instrumentation

Ultra high performance liquid chromatography (UPLC) and its conditions. Stress degradation study samples were analyzed using the Acquity UPLC-H class system from M/s Waters with a flow-through needle design integral sample manager (SM-FTN) and bio quaternary gradient pump and an on-line degasser. The column compartment having a temperature control and a photodiode array (PDA) detector was employed throughout the analysis. Chromatographic data was acquired using Empower-3 software. The Acquity BEH C18 column (50 × 2.1 mm, 1.7 μm) and acetonitrile was used for optimization of the method. To attain an acceptable separation of the BLN and its DPs, buffer solutions of different pH (2.5, 3, 4 and 5) were investigated at a flow rate of 0.5 ml min⁻¹ with the column temperature of 30 °C. It was found that at pH 4 and 5; O1, O2 and O3 were eluted before the drug while at pH 3 the drug might be ionized and eluted before O3. The desirable separation and symmetrical peak shapes were obtained with ammonium formate buffer (A) (pH 3.0; 10 mM, adjusted by formic acid) and ACN (B) in a gradient mode (T_min/% B): 0/30, 2/70, 2.5/70, 4/30. All the stress samples were analysed using a PDA detector at 245 nm. The same method was transferred to Agilent UHPLC-Q-TOF/MS/MS studies by optimizing all the MS parameters as discussed in the mass spectrometry section.

Mass spectrometry. Mass spectrometric studies were carried out on the Agilent 1200 series LC instrument (Agilent Technologies, USA) coupled with a quadrupole time-of-flight (Q-TOF) mass spectrometer (Q-TOF LC/MS 6540 series, Agilent Technologies, USA). It was equipped with either electrospray ionization (ESI) or an atmospheric pressure chemical ionization (APCI) source and operated in positive ionization mode. The controlling software was Mass Hunter Workstation. The typical operating source conditions were: the capillary, 3000 V; the fragmentor voltage, 165 V; the skimmer, 60 V and collision energy, 10–30 eV. Nitrogen was used as the drying (320 °C, 10 l min⁻¹) and the nebulizing (45 psi) gas. Ultrahigh pure nitrogen gas was used as collision gas. In APCI, the vaporizer temperature was set at 400 °C and the source current was set at 4 μA. All the other source parameters were the same as those used in the ESI.

Sonicator, pH meter, hot air oven, water bath and photo stability chamber. The ultra-sonicator from Power Sonic-405 (Hwashin Technology Co. Seoul, South Korea) and the pH meter from pH tutor (Eutech Instruments, Singapore) were employed to dissolve the sample and measure the pH of the mobile phase, respectively.

The hydrolytic and thermal stress degradation studies were carried out using a high precision water bath and hot air oven, respectively (Osworld Scientific Pvt. Ltd India). The stress photo degradation was carried out in a photostability chamber (Osworld OPSH-G-16-GMP series, Osworld scientific Pvt Ltd India) capable of controlling the temperature and humidity within a range of ±2 °C and ±5% RH, respectively. The chamber was equipped with an illumination bank made of light source as described in the ICH guideline Q1B.²⁰

Stress degradation study. Stress studies were performed using 500 μg ml⁻¹ solution of BLN, starting with a milder conditions followed by a stronger conditions so as to get sufficient degradation. In general, degradation of the drug up to 5–20% is considered as sufficient degradation. If degradation is performed to higher extent, it may lead to the formation of secondary degradation products. The secondary degradation products are not realistic. Hence, in the present study, 6.1% and 19.3% drug was degraded in photolytic and oxidative stress conditions, respectively. Stress degradation studies were carried out on the bulk drug as per ICH guidelines Q1A (R2). Acidic, basic and neutral hydrolytic degradations were carried out by refluxing the drug in 3 N HCl, 3 N NaOH and water at 80 °C for 48 h, 48 h and 72 h, respectively. For oxidative stress degradation, BLN was subjected to 15% H₂O₂ at room temperature for 15 h. As per the stress degradation study, concentration of the stressor is increased until drug gets degrade around 20%. Initial, lower concentration of HCl/NaOH and H₂O₂ was tried for degradation of the drug. But the drug was not degraded. Hence, the stressor concentration was increased to 3 N HCl/NaOH and 15% H₂O₂. Thermal degradation study was also carried out in the solid state by exposing pure BLN in a Petri plate with a very thin layer to dry heat at 105 °C for 5 days. A photolytic stress study was carried out in solution and solid sample at 40 °C by exposing to a total dose of 200 W h m⁻² of UV-illumination and 1.2 × 10⁶ lux h of fluorescent light. A parallel set of the drug solutions was stored in dark at the same temperature to serve as control.

Sample preparation. Due to poor solubility of BLN in water, all stress samples were prepared in methanol and water in the ratio of 50 : 50 v/v. All stress samples were diluted to 5 times with mobile phase and filtered through 0.22 μm filter paper prior to UPLC and UPLC-MS analysis.

Results and discussion

Degradation behavior of BLN

The present stress degradation study samples were analyzed using the Acquity UPLC-H class system from M/s Waters and the same method was transferred to Agilent LC-QTOF-MS/MS for structural characterization. The degradation behavior of BLN under various stress conditions was investigated using the optimized UPLC method. The overlay of UPLC chromatograms of all stress degradation samples is given in Fig. 1. The proposed structures of DPs and their elemental compositions are given in Fig. 2 and Table 1, respectively.

Fig. 1 The overlay of UPLC chromatogram of (A) acidic, (B) basic, (C) neutral, (D) photo solid, (E) photo liquid, (F) thermal and (G) oxidative stress conditions.

Fig. 2 Proposed structures of protonated degradation products of blonanserin produced under oxidative and photolytic stress conditions.

Table 1 LC-QTOF/MS data of the drug and its degradation products along with their possible molecular formulae and their fragment ions

Degradation product	Retention time (min)	Molecular formula [M + H]^+	Calculated m/z	Observed m/z	Error (ppm)	MS/MS fragment ions
BLN	1.98	C23H31FN3^+	368.2497	368.2486	2.99	323, 297, 283, 271, 241, 98
O1	1.27	C23H31FN3O2^+	400.2395	400.2378	4.25	382, 383, 364, 335, 313, 309, 297, 283, 271, 84, 56
O2	1.38	C23H31FN3O2^+	400.2395	400.2387	2.00	382, 383, 364, 335, 309, 297, 283, 271, 84, 56
O3	2.18	C23H31FN3O^+	384.2446	384.2439	1.82	366, 356, 340, 323, 297, 283, 270, 271, 242, 241, 228
P1	0.37	C6H15N2^+	115.1230	115.1231	0.87	87, 85, 72, 70, 58, 56
P2	0.90	C21H29FN3^+	342.2340	342.2350	−2.92	297, 271, 229
P3	1.71	C23H29FN3O2^+	398.2238	398.2245	−1.76	370, 313, 299, 271, 245
P4	1.84	C21H27FN3^+	340.2184	340.2189	−1.47	323, 297, 271, 229, 70

---

UPLC analysis of all the hydrolytic (acidic, basic and neutral) solutions showed that no degradation product was formed in 3 N HCl, 3 N NaOH and water at 80 °C after 48 h, 48 h and 72 h, respectively (Fig. 1A–C). This shows that drug is stable in all hydrolytic conditions. On photolytic stress degradation of solid sample, no degradation product was observed (Fig. 1D), while photolytic stress liquid sample showed four minor DPs (P1–P4) (Fig. 1E). No degradation product was observed under the thermal stress degradation study (Fig. 1F). The drug showed an extensive degradation in 15% H₂O₂ after 15 h in the dark room at an ambient temperature. A total of three DPs, O1, O2 and O3, were formed during oxidative stress condition (Fig. 1G). In oxidative (H₂O₂) stress conditions, intense peak corresponding to H₂O₂ was observed at near to void time (0.25 min). This peak is appeared in overlapping chromatograms (Fig. 1G).

Mass fragmentation pathway of the drug

The ESI-MS/MS spectrum of protonated BLN (R_t = 1.98 min; [m/z 368]) displays only one major peak at m/z 297 (loss of C₄H₉N form m/z 368) and five low abundance ions at m/z 323 (loss of C₂H₇N from m/z 368), m/z 283 (loss of C₅H₁₁N from m/z 368), m/z 271 (loss of C₆H₁₁N from m/z 368), m/z 241 (loss of C₄H₈ from m/z 297) and m/z 98 (loss of C₁₇H₁₉FN₂ from m/z 368) (Fig. 5a) (Scheme 1a). It can be noted that all the product ions are formed mostly from the breakdown of piperazine ring. The formation of m/z 297 can be explained by the loss of 1-ethylaziridine through scission of C–C bonds of piperazine moiety. The presence of an intense peak at m/z 297 is characteristic for 4-(4-fluorophenyl)-N-methyl-N-methylene-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-2-aminium. The generation of m/z 323 can be explained by ‘H’ migrations from the methylene groups adjacent to N-2 group to N-1 followed by the loss of ethyl amine. The elemental compositions of all these product ions have been confirmed by accurate mass measurements (Table 2).

Scheme 1 (a) Proposed fragmentation pathway of protonated BLN and O3. (b) Proposed fragmentation pathway of protonated O1 and O2.

Table 2 Elemental composition for fragment ions of the drug and its degradation products

BLN and DPs	Molecular formula [M + H]^+	Calculated m/z	Observed m/z	Error (ppm)
BLN	C23H31FN3^+	368.2497	368.2486	2.99
	C21H24FN2^+	323.1918	323.1917	0.31
	C19H22FN2^+	297.1762	297.1762	0.00
	C18H20FN2^+	283.1605	283.1600	1.77
	C17H20FN2^+	271.1605	271.1601	1.48
	C15H14FN2^+	241.1136	241.1126	4.15
	C6H12N^+	98.0964	98.0966	−2.04
O1	C23H31FN3O2^+	400.2395	400.2378	4.25
	C23H30FN3O^+˙	383.2367	383.2370	−0.78
	C23H29FN3O^+	382.2289	382.2282	1.83
	C23H27FN3^+	364.2184	364.2196	−3.29
	C21H22FN3^+˙	335.1792	335.1780	3.58
	C19H22FN2O^+	313.1711	313.1709	0.64
	C20H22FN2^+	309.1762	309.1757	1.62
	C19H22FN2^+	297.1762	297.1756	2.02
	C18H20FN2^+	283.1605	283.1607	−0.71
	C17H20FN2^+	271.1605	271.1599	2.21
	C5H10N^+	84.0808	84.0812	−4.76
	C3H8N^+	56.0495	56.0493	3.57
O2	C23H31FN3O2^+	400.2395	400.2387	2.00
	C23H30FN3O^+˙	383.2367	383.2371	−1.04
	C23H29FN3O^+	382.2289	382.2297	−2.09
	C23H27FN3^+	364.2184	364.2179	1.37
	C21H22FN3˙^+	335.1792	335.1783	2.69
	C20H22FN2^+	309.1762	309.1757	1.62
	C19H22FN2^+	297.1762	297.1761	0.34
	C18H20FN2^+	283.1605	283.1607	−0.71
	C17H20FN2^+	271.1605	271.1592	4.79
	C5H10N^+	84.0808	84.0807	1.19
	C3H6N^+	56.0495	56.0497	−3.57
O3	C23H31FN3O^+	384.2446	384.2439	1.82
	C23H29FN3^+	366.2340	366.2341	−0.27
	C21H27FN3O^+	356.2133	356.2137	−1.12
	C21H27FN3^+	340.2184	340.2183	0.29
	C21H24FN2˙^+	323.1918	323.1922	−1.24
	C19H22FN2^+	297.1762	297.1758	1.35
	C18H20FN2^+	283.1605	283.1601	1.41
	C17H20FN2^+	271.1605	271.1592	4.79
	C17H19FN2˙^+	270.1527	270.1524	1.11
	C17H19FN2˙^+	242.1214	242.1216	−0.83
	C15H14FN2^+	241.1136	241.1126	4.15
	C14H13FN2˙^+	228.1057	228.1046	4.82
P1	C6H15N2^+	115.1230	115.1231	−0.87
	C4H11N2^+	87.0917	87.0919	−2.30
	C4H9N2^+	85.0760	85.0763	−3.53
	C4H10N^+	72.0808	72.0811	−4.16
	C4H8N^+	70.0651	70.0653	−2.85
	C3H8N^+	58.0651	58.0649	3.44
	C3H6N^+	56.0495	56.0496	−1.78
P2	C21H29FN3^+	342.2340	342.2350	−2.92
	C19H22FN2^+	297.1762	297.1771	−3.03
	C17H20FN2^+	271.1605	271.1601	1.48
	C14H14FN2^+	229.1136	229.1129	3.06
P3	C23H29FN3O2^+	398.2238	398.2245	−1.76
	C22H29FN3O^+	370.2289	370.2302	−3.51
	C19H22FN2O^+	313.1711	313.1725	−4.47
	C18H20FN2O^+	299.1554	299.1541	4.35
	C17H20FN2^+	271.1605	271.1609	−1.48
	C14H14FN2O^+	245.1085	245.1080	2.04
P4	C21H27FN3^+	340.2184	340.2189	−1.47
	C21H24FN2˙^+	323.1918	323.1898	6.19
	C19H22FN2^+	297.1762	297.1767	−1.68
	C17H20FN2^+	271.1605	271.1599	2.21
	C14H14FN2^+	229.1136	229.1125	4.80
	C4H8N^+	70.0651	70.0653	−2.85

---

Structure elucidation of DPs using UPLC-QTOF-MS/MS experiments

Online UPLC-QTOF-/MS/MS experiments were performed to characterize all the DPs (O1–O3, P1–P4) formed under oxidative and photolytic solution stress conditions (Fig. 2). Initially, mass spectral data acquisition was tried in both positive and negative mode. But no peak was observed in negative ion mode (Fig. S1, ESI†). This may be due to the presence of nitrogen in the structure that facilitates ionization in positive mode only. The TIC (Total Ion Chromatogram) of degraded stress samples (photolytic and oxidative) and ESI-MS positive ion spectra of each degradation product are given in Fig. 3 and 4, respectively. Most plausible structures have been proposed for all the DPs based on the m/z values of their [M + H]⁺ ions and the MS/MS data in combination with elemental compositions derived from accurate mass measurements.

Fig. 3 The TIC (total ion chromatogram) of pure BLN, photolytic and oxidative stress samples.

Fig. 4 ESI/MS spectra of BLN and its degradation products (O1–O3, P1–P4).

Oxidative stress DPs, O1 and O2 were eluted at 1.27 min and 1.38 min, respectively. Accurate mass measurements of protonated O1 and O2 at m/z 400 yield the same elemental composition, C₂₃H₃₁FN₃O₂ [M + H]⁺. A mass difference of 32 Da between DPs (O1 and O2) and BLN indicates that DPs are formed by attachment of two oxygen atoms in the BLN moiety. The ESI/MS/MS spectra of both the protonated DPs show almost similar product ions (except that the product ion at m/z 313 was observed only in O1) at m/z 383 (loss of OH), m/z 382 (loss of H₂O), m/z 364 (loss of two H₂O molecules), m/z 309 (loss of C₃H₅N from m/z 364), m/z 297 (loss of C₄H₅N form m/z 364), m/z 283 (loss of ethyne from m/z 309), m/z 271 (loss of C₆H₇N from m/z 364), m/z 84 (loss of C₁₈H₁₇FN₂ from m/z 364) and m/z 56 (loss of C₂H₄ from m/z 84) (Scheme 1b), suggesting isomeric structures for O1 and O2 (Fig. 5b and c). Both the spectra display product ions at m/z 383 and 382, corresponding to the loss of OH radical and H₂O, respectively, suggesting the possibility of O1 and O2 being the N-oxides of BLN. The % intensity of m/z 383 was found more than 50% of m/z 382 which indicates that m/z 383 is not an isotope of m/z 382. Generally, oxidation of any heterocyclic compound may form hydroxyl or N-oxide degradation products. It is difficult to differentiate hydroxyl degradation product from N-oxides by ESI-MS/MS spectra alone because they have the same product ion spectrum and elemental composition particularly when an oxygen atom resides in the same product ion. N-oxides are thermally labile groups that undergo decomposition and deoxygenation in the atmospheric chemical ionization (APCI) source and that occurs due to thermal energy activation in the vaporizer of the APCI source.^28,29 Hence, the stress sample was analysed in full scan LC/APCI-MS mode to confirm the presence of N-oxide. As shown in Fig. 6a and b, the full scan APCI mass spectra of O1 and O2 displayed peaks at m/z 384 and m/z 368, resulting from stepwise thermally induced deoxygenation in the vaporizer of the APCI source. This proves that N,N-dioxide is formed from oxidation of two nitrogens in BLN. It is interesting to note that other major m/z 372, m/z 356 and m/z 272 were also observed in the APCI mass spectrum of O2. The peak at m/z 356 of O1 and O2 appears to have resulted from a thermally induced Meisenheimer rearrangement. It is a pyrolytic reaction in which alkyl or benzyl group on the N-oxide nitrogen undergoes a N-R to O-R rearrangement above 200 °C followed by elimination of an aldehyde (or a ketone) (Scheme 2). This may involve the migration of the N-ethyl group from the nitrogen atom to the oxygen in the piperazine ring to form N-ethoxylpiperazine, followed by elimination of acetaldehyde to yield the ion at m/z 356 (Fig. 6a and b).²⁹ The peak at m/z 272 can be formed by the migration of the pyridine nucleus from N to O in N-aryl piperazine, followed by the elimination of 1-ethoxy-1,2,3,6-tetrahydropyrazine. The peak at m/z 372 was observed only in O2 which was characteristic for O2 to differentiate from O1. The peak at m/z 372 can be explained by the migration of both N-ethyl group and 4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridine group from N to O followed by the loss of ethene. These product ions indicated that oxidation occurred on the N-2 and N-3 of the piperazine ring where 4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridine and ethyl groups are attached, respectively. The absence of abundant ions at m/z 372 and m/z 272 in the spectrum of O1 indicated that oxidation occurred on the N-1 of the pyridine ring and N-3 of the piperazine ring, where ethyl group is attached. Generally, Meisenheimer rearrangement is not observed in heterocyclic N-oxides of pyridine ring. These ions (m/z 372, m/z 356 and m/z 272) were absent in ESI product ions mass spectrum. The product ions observed in ESI-MS/MS and full scan APCI of O1 and O2 are consistent with the structure, 1-ethyl-4-(4-(4-fluorophenyl)-1-oxido-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-2-yl) piperazine 1-oxide and 1-ethyl-4-(4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-2-yl)piperazine 1,4-dioxide, respectively. A probable mechanism for the formation of O1 and O2 may involve nucleophilic addition of hydroperoxide anions to the tertiary nitrogens in BLN followed by hydroxide elimination and abstraction of hydrogens by hydroxide anions as shown in Scheme 3. The elemental compositions of product ions have been confirmed by accurate mass measurements (Table 2).

Fig. 5 ESI/MS/MS spectrum of [M + H]⁺ ions of (a) BLN (m/z 368) at 30 eV, (b) O1 (m/z 400) at 20 eV, (c) O2 (m/z 400) at 20 eV and (d) O3 (m/z 384) at 20 eV.

Fig. 6 Full scan LC/APCI-MS spectrum of (a) O1 (m/z 400) (b) O2 (m/z 400) and (c) O3 (m/z 384).

Scheme 2 Explanation of Meisenheimer rearrangement.

Scheme 3 Probable mechanism of formation of O1, O2 and O3 under oxidation condition.

The major oxidative degradation product, O3 ([M + H]⁺, m/z 384) was eluted at 2.18 min, suggesting the addition of an oxygen to BLN (Fig. 5d; Table 1). The ESI/MS/MS spectrum shows diagnostic ions at m/z 356 and m/z 340 corresponding to the loss of ethene and acetaldehyde (Meisenheimer rearrangement), respectively. These products ions were more prominent in full scan APCI/MS (Fig. 6c) confirming that O3 is an oxidative DP. It is likely that the tandem mass spectrum of N-oxide shows neutral loss of water resulting in formation of product ion at m/z 366.^30,31 The MS/MS spectrum of O3 also shows the product ions at m/z 323, m/z 297, m/z 283 and m/z 271, which were also observed in MS/MS spectrum of protonated BLN. The elemental compositions of O3 and its product ions have been confirmed by accurate mass measurements (Table 2). Based on these discussions, most probable structure of O3 is proposed as 1-ethyl-4-(4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-2-yl)piperazine 1-oxide. The mechanism of formation of O3 is similar to those of O1 and O2 (Scheme 3).

The ESI/MS/MS spectrum of [M + H]⁺ ion (m/z 115) of P1 (R_t = 0.37 min) (Fig. 7a; Table 1) and its elemental composition indicate that it is formed by the loss of 4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridine. The MS/MS spectrum displays product ions at m/z 87 (loss of C₂H₄ from m/z 115), m/z 85 (loss of C₂H₆), m/z 72 (loss of NH₃ from m/z 72), m/z 70 (loss of C₂H₇N from m/z 115), m/z 58 (loss of CNH₃ from m/z 87) and m/z 56 (loss of CNH₃ from m/z 85). Proposed fragmentation pathway for P1 is given in Scheme 4, which has been supported by accurate mass measurements of precursor ion and its product ions. Based on these results, P1 was identified as 1-ethylpiperazine. A probable mechanism for the formation of P1 under photolytic condition is shown in Scheme 5.

Fig. 7 ESI/MS/MS spectrum of [M + H]⁺ ions of (a) P1 (m/z 115) at 25 eV, (b) P2 (m/z 342) at 20 eV, (c) P3 (m/z 398) at 20 eV and (d) P4 (m/z 340) at 20 eV.

Scheme 4 (a) Proposed fragmentation pathway of protonated degradation products P1, P2 and P4. (b) Proposed fragmentation pathway of protonated degradation product P3.

Scheme 5 Probable mechanism of formation of P1, P2, P3 and P4 under photo degradation condition.

The photolytic degradation product P2 ([M + H]⁺, m/z 342) was eluted at 0.90 min. A mass difference of 26.0136 Da between P2 and BLN indicates that DP is produced by the loss of ethyne group. Formation of the diagnostic product ion at m/z 297 from the protonated molecule by the loss of ethyl amine confirms that 4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridine nucleus was intact. This suggests that P2 was formed by the loss of ethyne group from the piperazine moiety.³² The elemental compositions of P2 and its product ions have been confirmed by accurate mass measurements (Fig. 7b; Table 1). All these data strongly suggest that the ethyne group is eliminated from the piperazine ring and the proposed structure is 1-ethyl-4-(4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-2-yl)ethane-1,2-diamine. The formation of P2 can be explained by the homolysis of C–N bond of piperazine ring followed by an addition of hydrogen radical as shown in Scheme 5.

The photolytic degradation product, P3 at m/z 398 was formed under photo stress degradation condition and was eluted at 1.71 min. The molecular formula of protonated P3 (C₂₃H₂₉FN₃O₂⁺) indicates the inclusion of two oxygen atoms in BLN with the loss of two hydrogens. ESI/MS/MS spectrum of [M + H]⁺ of P3 does not show any product ions at m/z 380 ([M − 18]⁺) (due to H₂O loss) and m/z 354 ([M − 44]⁺) (due to CO₂ loss) ruling out the hydroxylation and carboxylation of BLN. The presence of diagnostic ions at m/z 370 (loss of CO from m/z 398) and m/z 271 (loss of CO from m/z 299) involving loss of two –CO groups, suggests the formation of two aldehyde moieties in the drug (Fig. 7c). Further support comes from the appearance of other product ions at m/z 313 (loss of C₃H₇N from m/z 370), m/z 299 (loss of C_4SH₉N from m/z 370) and m/z 245 (loss of C₄H₆ from m/z 299) suggesting that P3 is a dialdehyde product of BLN. This is in line with a study by Condorelli et al. who reported that breakdown of piperazine ring of rufloxacin results in the formation of aldehyde under photolytic stress condition.³³ Based on all these results P3 has been identified as N-ethyl-N-(2-(N-(4-(4-fluorophenyl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridin-yl) formamido)ethyl) formamide (Table 1). BLN in the solution may generate a diradical by cleavage of C–C bond in the piperazine ring on exposure to UV light.³³ This diradical on reaction with O2 may form peroxide which on decomposition via homolysis of O–O bond gives an oxygen radical which readily loses hydrogen to give P3 (Scheme 5).

The photolytic degradation product, P4 at m/z 340 ([M + H]⁺) with an elemental composition of C₂₁H₂₇FN₃ was eluted at 1.84 min, suggesting the loss of the ethene from BLN. The ESI/MS/MS spectrum of protonated P4 shows the diagnostic product ion at m/z 297 from the protonated P4 by the loss of aziridine confirming the elimination of ethene from 1-ethylpiperazine of BLN under photolytic stress condition (Fig. 7d). The appearance of other product ions at m/z 323 (loss of NH₃ from m/z 340), m/z 271 and m/z 229 support the structure of P4. The elemental compositions of P4 and its product ions have been confirmed by accurate mass measurements and P4 is identified as 4-(4-fluorophenyl)-2-(piperazin-1-yl)-5,6,7,8,9,10-hexahydrocycloocta[b]pyridine. A most probable mechanism for the formation of P4 under photolytic condition is shown in Scheme 5.

The recovery of the drug was calculated in the presence of oxidative and photolytic stress degradation samples. The recoveries of the added drug were obtained from the difference between peak areas of fortified and unfortified degraded samples. The percentage recovery range and % RSD values were found to be 98.39–101.34 and <2%, respectively (Table S1, see ESI†).

In silico toxicity prediction

In order to assess the potential toxicity of BLN and its degradation products, TOPKAT (Discovery Studio 2.5, Accelrys, Inc., San Diego, CA, USA) and DEREK (Nexus v2.0, Lhasa Ltd, Leeds, UK) software were used.³⁴ By employing statistically robust, cross-validated, and rigorously developed Quantitative Structure Toxicity Relationship (QSTR) models, TOPKAT predicts specific toxicological effects exclusively from molecule's chemical structure. TOPKAT uses descriptors that quantify the properties related to the transport (e.g., molecular mass, shape, symmetry) and the chemistry (electronic attributes of molecular structure). Probability values from 0.0 to 0.30 are likely to produce a negative response in an experimental assay and considered low probabilities for any toxicological effect. However, probability values greater than 0.70 are expected to produce a positive response in an experimental assay and considered as high probabilities. Probabilities values greater than 0.30 but less than 0.70 are considered as indeterminate.³⁵ All these probability values are only for toxicological end points. Other values which are greater than 1 are related to the prediction of drug dose or concentration (e.g. Rat Oral LD50 (v3.1), Chronic LOAEL etc.) not for any toxicological end point. It is just additional information obtained from this software. The results of predicted toxicity using TOPKAT is shown in Table 3. The toxicity of DPs were assessed and compared with the drug in different animal models. Broadly, the probability values of the drug and its DPs show similarity but with some major exceptions. P1 showed greater deviation in toxicity profile than any other DP e.g. P1 shows lesser toxicity for NTP Carcinogenicity Call (Male Rat) (v3.2) and FDA Carcinogenicity Female Rat Non vs. Carc (v3.1) model while the probability values for FDA Carcinogenicity Male Mouse Non vs. Carc (v3.1) and Skin Irritation (v6.1) model is pretty higher than other DPs and parent compound. However, other DPs also demonstrated different toxicity values for different models. For example, P3 showed higher values for Skin Sensitization MLD/MOD v SEV (v6.1) model than parent drug which can be explained by the generation of two aldehyde groups due to the hydrolysis of piperazine ring.

Table 3 Probability values of different toxicity models of the drug and its degradation products (O1–O3, P1–P4) by TOPKAT analyses

Model	BLN	O1	O2	O3	P1	P2	P3	P4
NTP Carcinogenicity Call (Male Rat) (v3.2)	0.992	1.000	1.000	0.998	0.006	0.996	1.000	0.997
Ames Mutagenicity (v3.1)	0.000	0.000	0.000	0.000	0.075	0.000	0.000	0.000
NTP Carcinogenicity Call (Female Rat) (v3.2)	0.000	0.000	0.050	0.000	0.000	0.000	0.000	0.000
NTP Carcinogenicity Call (Male Mouse) (v3.2)	0.209	1.000	1.000	1.000	0.009	1.000	0.000	0.352
NTP Carcinogenicity Call (Female Mouse) (v3.2)	1.000	0.996	0.999	1.000	0.679	1.000	1.000	1.000
FDA Carcinogenicity Male Rat Non vs. Carc (v3.1)	0.000	0.000	0.000	0.000	0.009	0.000	0.000	0.000
FDA Carcinogenicity Male Rat Single vs. Mult (v3.1)	0.949	0.000	0.000	0.000	1.000	0.999	0.000	0.000
FDA Carcinogenicity Female Rat Non vs. Carc (v3.1)	1.000	0.991	0.601	1.000	0.015	0.965	0.995	0.420
FDA Carcinogenicity Female Rat Single vs. Mult (v3.1)	1.000	1.000	1.000	1.000	0.739	1.000	0.991	1.000
FDA Carcinogenicity Male Mouse Non vs. Carc (v3.1)	0.004	0.000	0.000	0.000	0.999	0.002	0.000	0.652
FDA Carcinogenicity Male Mouse Single vs. Mult (v3.1)	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
FDA Carcinogenicity Female Mouse Non vs. Carc (v3.1)	0.000	0.000	0.000	0.000	0.671	1.000	0.000	0.000
FDA Carcinogenicity Female Mouse Single vs. Mult (v3.1)	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Weight of Evidence Carcinogenicity Call (v5.1)	1.000	0.552	0.953	0.535	1.000	0.999	0.644	0.944
Developmental Toxicity Potential (DTP) (v3.1)	0.000	0.000	0.062	0.000	0.566	0.209	0.000	0.000
Rat Oral LD50 (v3.1) (g kg^−1)	1.000	0.348	0.398	0.391	2.000	1.500	1.500	1.800
Rat Maximum Tolerated Dose – Feed/Water (v6.1) (μg kg^−1)	0.098	0.478	0.130	0.027	153 000.00	0.147	1.20	0.118
Rat Maximum Tolerated Dose – Gavage (v6.1) (μg kg^−1)	0.272	1.30	0.359	0.076	422 800.00	0.407	3.30	0.326
Rat Inhalational LC50 (v6.1) (g m^−3 h^−1)	0.121	10.000	10.000	10.000	10.000	10.000	10.00	2.100
Chronic LOAEL (v3.1) (mg kg^−1)	339.700	27.800	275.700	320.700	0.943	310.200	272.20	334.300
Skin Irritation (v6.1)	0.000	1.000	0.000	0.061	1.000	0.074	0.001	1.000
Skin Sensitization NEG v SENS (v6.1)	0.000	0.000	0.000	0.000	0.280	0.000	0.000	0.000
Skin Sensitization MLD/MOD v SEV (v6.1)	0.000	0.000	0.000	0.000	0.000	0.997	1.000	0.729
Ocular Irritancy SEV/MOD vs. MLD/NON (v5.1)	1.000	1.000	1.000	1.000	0.976	1.000	1.000	1.000
Ocular Irritancy SEV vs. MOD (v5.1)	1.000	1.000	1.000	1.000	0.557	1.000	0.995	1.000
Ocular Irritancy MLD vs. NON (v5.1)	0.684	1.000	1.000	1.000	1.000	0.845	1.000	0.106
Aerobic Biodegradability (v6.1)	0.000	0.000	0.000	0.000	0.489	0.401	0.000	0.000
Fathead Minnow LC50 (v3.2) (μg l^−1)	2.00	11.70	11.80	17.40	3 700 000.00	2.80	9.70	7.60
Daphnia EC50 (v3.1) (mg l^−1)	439.40	247.60	219.20	297.00	169.30	276.60	246.20	1.60

---

DEREK, knowledge based toxicity prediction tool, uses a set of rules derived from the collective expertise of toxicologists worldwide. Unlike TOPKAT, DEREK does not provide a quantitative estimate of the prediction. DEREK provides one of the nine levels of confidence for any toxicity prediction i.e. certain, probable, plausible, equivocal, doubted, improbably, impossible, open, and contradicted. It has been developed on several rule bases, consisting of descriptions of molecular substructures (structural alerts) that have been associated with toxic end points (e.g. genotoxicity, carcinogenicity, skin irritation etc.) for a number of species, including dog, hamster, mouse, guinea pig, mammal, monkey, primate, rabbit, rat, rodent and human. When a molecular structure is submitted, it first checks whether any alerts in the knowledge base match toxicophores in the query structure and then reasoning engine assesses the likelihood of a structure being toxic.³⁶ Table 4 gives the qualitative results for the drug molecule and its DPs on skin sensitization, hERG channel inhibition and phospholipidosis end points obtained using DEREK. P1 showed skin sensitization among all other molecules because of the presence of the structural alert diamine. Additionally, small lipophilic molecules are more readily absorbed into the skin and are therefore more likely to cause sensitisation. The hERG toxicity of parent compound, P2 and P4 can be explained by the presence of structural alert hERG pharmacophore II. This explains that oxidation or hydrolysis of parent compound leads to lesser hERG toxicity. Further, P2 also contains amine group which is a structural alert for phospholipidosis (details given in Table 4).

Table 4 Qualitative toxicity prediction of the drug and its degradation products (O1–O3, P1–P4) by DEREK analysis

Drug and DPs	Skin sensitisation	hERG channel inhibition	Phospholipidosis
Structural alert	• R1 = H, C (with no additional hetero atoms attached)	• R1 = A (aromatic) or A (non-aromatic)-A (aromatic)	• R1, R2 = H, CH3, CH2CH3, CH(CH3)2, C(CH3)3
	• [C] = carbon chain with no additional hetero atoms attached n = 2 to 6	• R2 = C (sp3)	• R3 = C, Q (no ring allowed)
	• N atoms cannot be in a ring of size 3	• R3, R4 = H or C (sp3, chain of maximum 3 atoms)	• Carbons * must be methylene
		• Only one of R3 or R4 may be H	• Structure must contain an aromatic moiety
		• A = any atom
		• n = 1–4 (A–A bonds may be of any type where valency allows)
		• If n = 2 or 3 then R1 may be attached to the beta-A-atom
		• If n = 4 then R1 may be attached to either the beta- or gamma-A-atom
		• Bonds - - - can be of any type
		• No CO2H group allowed anywhere
Comments	The presence of a skin sensitisation structural alert within a molecule indicates the molecule has the potential to cause skin sensitisation. Whether or not the molecule will be a skin sensitiser will also depend upon its percutaneous absorption. Generally, small lipophilic molecules are more readily absorbed into the skin and are therefore more likely to cause sensitisation	This alert describes a structure-based pharmacophore developed primarily from compounds that have been reported to be moderate or strong inhibitors of the hERG (human ether-a-go-go-related gene) potassium channel. In general, the substructure most common to hERG channel inhibitors is an aromatic ring at 3–6 bonds distance from an amine that is capable of forming a positive charge	This alert describes the phospholipidosis inducing potential of amines. Phospholipidosis is the build-up of excess intracellular phospholipids. Cationic amphiphilic drugs (CADs) consisting of a hydrophilic cationic amine, and a hydrophobic ring region have been found to induce phospholipidosis. Their amphiphilic nature mimics that of phospholipids, and allows a degree of solubility in both aqueous and lipid media
BLN	NA	✓	NA
O1	NA	NA	NA
O2	NA	NA	NA
O3	NA	NA	NA
P1	✓	NA	NA
P2	NA	✓	✓
P3	NA	NA	NA
P4	NA	✓	NA

---

Conclusion

The degradation behavior of BLN under various stressed conditions was investigated as per ICH guidelines. The drug showed four minor DPs in photolytic condition, whereas three degradation products were observed under oxidative stress conditions. A total of 7 unknown degradation products were identified and characterized using on-line UPLC/MS/MS experiments combining with accurate mass measurements. Additionally, a full scan APCI-MS was successfully applied to characterize N-oxide degradation products of the drug. The product ion resulting from Meisenheimer rearrangement of tert-N-oxides in the APCI mass spectrum can be used to distinguish N-oxides from isomeric oxidative degradation products. An extensive fragmentation pathway of the protonated drug and its degradation products was established. The proposed structures of the degradation products have been rationalized by appropriate mechanisms. Further, in silico toxicities were predicted for all DPs using TOPKAT and DEREK softwares. DEREK software shows structure alert for P1 that is likely to cause skin sensitization. Moreover, the presence of diamine in P2 shows high probability of phospholipidosis that is characterized by the excess accumulation of phospholipids in tissues. The drug, P2 and P4 show a high probability of hERG toxicity due to presence of structural alert, hERG pharmacophore II.

Acknowledgements

The authors thank Dr Ahmed Kamal, project director, NIPER, Hyderabad for facilities and their cooperation. P. K. and P. P. are thankful to Department of Pharmaceuticals, Ministry of Chemicals and Fertilizers, New Delhi for the award of a Senior Research Fellowship.

References

1. T. Kishi, Y. Matsuda, H. Nakamura and N. Iwata, J. Psychiatr. Res., 2013, 47, 149–154.
2. E. D. Deeks and G. M. Keating, CNS Drugs, 2010, 24, 65–84.
3. T. Kishi, Y. Matsuda and N. Iwata, Asia Pac. Psychiatr., 2014, 6, 462.
4. T. Kishi, Y. Matsuda, K. Fujita and N. Iwata, Neuropsychiatr. Dis. Treat., 2014, 10, 1861–1866.
5. E. Garcia, M. Robert, F. Peris, H. Nakamura, N. Sato and Y. Terazawa, CNS Drugs, 2009, 23, 615–625.
6. T. Kishi, Y. Matsuda and N. Iwata, PLoS One, 2014, 9, e88049.
7. Y. Ninomiya, S. Miyamoto, T. Tenjin, S. Ogino, N. Miyake, Y. Kaneda, T. Sumiyoshi and N. Yamaguchi, Psychiatry Clin. Neurosci., 2014, 68, 841–849.
8. H. Hori, K. Yamada, D. Kamada, Y. Shibata, A. Katsuki, R. Yoshimura and J. Nakamura, Neuropsychiatr. Dis. Treat., 2014, 10, 527–533.
9. Z. Caiyan and S. Yingbing, China Pharm., 2010, 10, 045.
10. H. Ping, J. North Pharmacy, 2013, 3, 003.
11. P. Mondal, S. Shobharani and R. Ramakrishna, Indo Am. J. Pharm. Res., 2014, 4, 1968–1975.
12. K. Iwakura, T. Ogawa, H. Hattori, M. Iwai, O. Suzuki and T. Nakagawa, Forensic Toxicol., 2014, 32, 30–38.
13. T. Ogawa, H. Hattori, R. Kaneko, K. Ito, M. Iwai, Y. Mizutani, T. Arinobu, A. Ishii, O. Suzuki and H. Seno, Anal. Sci., 2009, 26, 1099–1102.
14. Y.-G. Wen, X.-J. Ni, M. Zhang, X. Liu and D.-W. Shang, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2012, 903, 46–52.
15. Y. Zhou, M. Liu, J. Jiang, H. Wang and P. Hu, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2013, 939, 59–66.
16. J. Saruwatari, N. Yasui-Furukori, Y. Inoue and S. Kaneko, Eur. J. Clin. Pharmacol., 2010, 66, 899–902.
17. Y. Zheng, X. Hu, J. Liu, G. Wu, H. Zhou, M. Zhu, Y. Zhai, L. Wu and J. ShenTu, J. Anal. Methods Chem., 2014, 2014, 1–7.
18. S. W. Baertschi and P. J. Jansen, in Drugs and the pharmaceutical science, ed. S. W. Baertschi, Taylor & Francis Group, New York, 2nd edn, 2005, vol. 153, pp. 13–15.
19. ICH guideline, Q1A (R2) Stability Testing of New Drug Substances and Products, International Conference on Harmonisation, IFPMA, Geneva, Switzerland, 2000.
20. ICH guideline, Q1B Photostability Testing on New Drug Substances and Products, International Conference on Harmonization, IFPMA, Geneva, Switzerland, 2000.
21. S. Singh, T. Handa, M. Narayanam, A. Sahu, M. Junwal and R. P. Shah, J. Pharm. Biomed. Anal., 2012, 69, 148–173.
22. L. G. Valerio, Toxicol. Appl. Pharmacol., 2009, 241, 356–370.
23. M. V. N. K. Talluri, K. K. Keshari, P. D. Kalariya and R. Srinivas, J. Sep. Sci., 2015, 38, 1625–1632.
24. M. V. N. K. Talluri, N. R. Kandimallaa, R. Bandu, D. Chundi, R. Marupaka and R. Srinivas, J. Pharm. Anal., 2014, 4, 107–116.
25. P. D. Kalariya, M. V. N. K. Talluri, P. N. Patel and R. Srinivas, J. Pharm. Biomed. Anal., 2015, 102, 353–365.
26. M. V. N. K. Talluri, S. Dharavath, P. D. Kalariya, B. Prasanth and R. Srinivas, J. Pharm. Biomed. Anal., 2015, 105, 1–9.
27. M. V. N. K. Talluri, P. D. Kalariya, P. N. Patel and S. Ragampeta, New J. Chem., 2015, 39, 6303–6314.
28. S. Ma, Y. Xu and M. Shou, Rapid Commun. Mass Spectrom., 2009, 23, 1446–1450.
29. S. Ma, S. K. Chowdhury and K. B. Alton, Anal. Chem., 2005, 77, 3676–3682.
30. L. Sridhar, R. Karthikraj, M. Murty, N. P. Raju, M. Vairamani and S. Prabhakar, Int. J. Mass Spectrom., 2013, 333, 15–20.
31. M. Holcapek, R. Jirasko and M. Lisa, J. Chromatogr. A, 2010, 1217, 3908–3921.
32. P. Calza, C. Medana, F. Carbone, V. Giancotti and C. Baiocchi, Rapid Commun. Mass Spectrom., 2008, 22, 1533–1552.
33. G. Condorelli, G. D. Guidi, S. Giuffrida, S. Sortino, R. Chillemi and S. Sciuto, Photochem. Photobiol., 1999, 70, 280–286.
34. J. C. Dearden, J. Comput.–Aided Mol. Des., 2003, 17, 119–127.
35. K. Enslein, V. K. Gombar and B. W. Blake, Mutat. Res., Fundam. Mol. Mech. Mutagen., 1994, 305, 47–61.
36. J. E. Ridings, M. D. Barratt, R. Cary, C. G. Earnshaw, C. E. Eggington, M. K. Ellis, P. N. Judson, J. J. Langowski, C. A. Marchant and M. P. Payne, Toxicology, 1996, 106, 267–279.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra10641a

---