RESEARCH ARTICLE – Drug Discovery–Development Interface

Kinetics and Mechanism of the Base-Catalyzed Rearrangement and Hydrolysis of Ezetimibe 2 ˇ ˇ ´ 1 ALESˇ IMRAMOVSKY, ´ 1 JOSEF HAJ ´ ´ICEK, ´ ´ 2 JIRˇ ´I HANUSEK1 JANA BATOV A, LUDMILA HEJTMANKOV A, 1

Institute of Organic Chemistry and Technology, Faculty of Chemical Technology, University of Pardubice, Pardubice 532 10, The Czech Republic 2 Zentiva a.s, 102 37 Praha 10, The Czech Republic Received 21 October 2013; revised 1 April 2014; accepted 4 June 2014 Published online 1 July 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24070

ABSTRACT: The pH-rate profile of the pseudo-first-order rate constants for the rearrangement and hydrolysis of Ezetimibe giving (2R,3R,6S)N,6-bis(4-fluorophenyl)-2-(4-hydroxyphenyl)-3,4,5,6-tetrahydro-2H-pyran-3-carboxamide (2) as the main product at pH of less than 12.5 and the mixture of 2 and 5-(4-fluorophenyl)-5-hydroxy-2-[(4-fluorophenylamino)-(4-hydroxyphenyl)methyl]-pentanoic acid (3) at pH of more than 12.5 in aqueous tertiary amine buffers and in sodium hydroxide solutions at ionic strength I = 0.1 mol L−1 (KCl) and at 39◦ C is reported. No buffer catalysis was observed and only specific base catalysis is involved. The pH-rate profile is more complex than the pH-rate profiles for the hydrolysis of simple ␤-lactams and it contains several breaks. Up to pH 9, the log kobs linearly increases with pH, but between pH 9 and 11 a distinct break downwards occurs and the values of log kobs slightly decrease with increasing pH of the medium. At pH of approximately 13, another break upwards occurs that corresponds to the formation of compound 3 that is slowly converted to (2R,3R,6S)-6-(4-fluorophenyl)-2-(4-hydroxyphenyl)-3,4,5,6-tetrahydro-2H-pyran-3-carboxylic acid (4). The kinetics of base-catalyzed C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association J hydrolysis of structurally similar azetidinone is also discussed.  Pharm Sci 103:2240–2247, 2014 Keywords: ezetimibe; rearrangement; azetidinone; NMR; kinetics; Liquid chromatography; acid-base catalysis; pH rate profile; chemical stability; degradation products; UV-VIS spectroscopy

INTRODUCTION  R

 R

Ezetimibe (Zetia , Ezetrol ), (3R,4S)-1-(4-fluorophenyl)-3[(3S)-3-(4-fluorophenyl)-3-hydroxypropyl]-4-(4-hydroxyphenyl )azetidin-2-one (1), is a selective cholesterol absorption inhibitor, which significantly lowers levels of the biliary and dietary cholesterol in the small intestine.1 The inherent stability of Ezetimibe has been determined under a variety of conditions,2–4 but there is no information about the structure of degradation products in those reports. Cited articles2–4 mainly report on the relative or absolute retention times of degradation products under various HPLC conditions. The formation of (2R,3R,6S)-N,6-bis(4-fluorophenyl)-2-(4-hydroxyphenyl)3,4,5,6-tetrahydro-2H-pyran-3-carboxamide (2) was briefly mentioned by Singh et al.,2 but no proper characterization of this compound was described. Moreover, the correct structure of the main alkaline degradation product has been the subject of controversy. At the beginning of 2011, Gajjar and Shah5 reported 5-(4-fluorophenyl)-2-[(4-fluorophenylamino)(4-hydroxyphenyl)methyl]-pent-4-enoic acid (Scheme 1) to be the major alkaline degradation product. As early as 1 week after the publication of this article, Barhate and Mohanraj6 pointed out in their Letter to the Editor, that the interpretation of spectral data in the original paper5 was wrong. The correct

Correspondence to: Jiˇr´ı Hanusek (Telephone: +420-466-037-015; Fax. +420466-037-068; E-mail: [email protected]) ´ cek on the occasion of his 70th birthDedicated to Professor Vladim´ır Machaˇ day. This article contains supplementary material available from the authors upon request or via the Internet at http://wileylibrary.com. Journal of Pharmaceutical Sciences, Vol. 103, 2240–2247 (2014)  C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association

2240

structure of the alkaline degradant—compound (2)—was ´ et al.8 published a few months later by Filip et al.7 and Santa (Scheme 1). The formation of the same compound (2) under strong acid conditions was reported earlier and its absolute structure was confirmed by X-ray analysis.9 However, the determination of the structure of other degradation products formed in alkaline medium and the dependency for their formation at different pH were not described. To get a better insight into the reaction mechanism of the unusual Ezetimibe transformation to (2) and other not yet published degradation products (3) and (4) (Scheme 1), we performed a detailed kinetic study whose results we wish to report in this paper.

MATERIALS AND METHODS Materials and Reagents Tertiary amines (triethylamine, N-methylmorpholine, Nmethyldiethanolamine, and N,N-dimethylaminoethanol) and hydrochloric acid used for the preparation of buffer solutions and KCl for ionic strength adjustment were of extra pure grade and were obtained from commercial suppliers (Sigma–Aldrich Company LLC, Prague, The Czech Republic, and Acros Organics, part of Thermo Fisher Scientific, Geel, Belgium). All the buffer solutions were freshly prepared just before kinetic measurements. The water used for the kinetic and product studies was distilled twice. Pure (>99.9%) Ezetimibe was obtained from Zentiva-Prague, Czech Republic (part of the Sanofi-Aventis group). All other solvents for preparative/analytical chromatography were purchased from Merck or Sigma–Aldrich and were of HPLC grade.

Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

2241

RESEARCH ARTICLE – Drug Discovery–Development Interface

Scheme 1. Ezetimibe alkaline degradation pathways.

Nuclear Magnetic Resonance Spectroscopy 1

13

The H and C NMR (nuclear magnetic resonance) spectra were recorded on a Bruker Avance 3–400 MHz (Billerica, Massachusetts, USA) instrument in hexadeuteriodimethyl sulfoxide (DMSO-d6 ) solution. Chemical shifts * are referenced to the solvent residual peaks *(DMSO-d6 ) = 2.50 (1 H) and 39.6 (13 C) ppm. Coupling constants J are quoted in Hz. Proton– proton connectivities were found by gs-COSY. Protonated carbon atoms were assigned by inspection of gs-HSQC spectra. 13 C NMR spectra were also measured in a standard way and by means of the APT (attached proton test) pulse sequence to distinguish CH, CH3 and CH2 , and Cquart . All NMR experiments were performed with the aid of the manufacturer’s software. HPLC Measurements The samples were taken directly from the reaction mixture containing 5 mg of Ezetimibe and 6 mL of 0.1 or 0.01 M NaOH, respectively, and analyzed on the LaChrom Elite HPLC System (Hitachi High Technologies America, Inc., Schaumburg, Illinois, USA) equipped with a diode array UV–Vis detector MultiChrom 5 (Hitachi High Technologies America, Inc., Schaumburg, Illinois, USA) using chromatographic column Purospher RP8e, 250 × 4.0 mm2 , 5 :m (Merck KGaA, Darmstadt, Germany) and a mobile phase composed of phosphate buffer (0.02 M KH2 PO4 , pH 2.7 ± 0.05) and acetonitrile (both Merck KGaA). The mobile phase flow rate was kept at 1 mL/min using gradient elution [gradient ranging from 70% (v/v) buffer +30% (v/v) acetonitrile to 30% (v/v) buffer +70% (v/v) acetonitrile during 25 min). Representative chromatograms can be found in the Supplementary Information.

nium formate pH 6.3 and acetonitrile (gradient ranging from 30% to 100% acetonitrile in 18 min). For ionization of eluted analytes, an APCI ion source operated in positive mode was employed (vaporizer temperature 400◦ C, capillary temperature 300◦ C, discharge current 4 :A, and tube lens voltage 40 V). Chiral HPLC The optical purity of Ezetimibe and the alkaline degradation product (2) were determined with a HPLC system consisting of a high-pressure pump (LCP 4000; ECOM-Prague, The Czech Republic), an autosampler and a UV detector (LCD 2082; ECOM-Prague), and employing a Chiralcel OD-H (Chiral Technologies Europe, Illkirch, Cedex, France) column (250 × 4.6 mm2 ). The mobile phase consisted of hexane– propan-2-ol mixture (85:15, v/v). The flow rate was kept at 0.8 mL min−1 and UV detection was at 230 nm. The retention time of Ezetimibe was 20.72 min, and the retention time of (2R,3R,6S)-N,6-bis(4-fluorophenyl)-2-(4-hydroxyphenyl)3,4,5,6-tetrahydro-2H-pyran-3-carboxamide (2) was 16.82 min. Both samples provided single peaks that proved their chemical as well as optical purity. R

Optical Rotatory Power The optical rotatory power of (2R,3R,6S)-N,6-bis(4fluorophenyl)-2-(4-hydroxyphenyl)-3,4,5,6-tetrahydro-2Hpyran-3-carboxamide (2) ["]D 20 = −114.4◦ was measured in methanol on a Perkin Elmer 341 (Perkin Elmer, Inc., Alameda, California, USA) instrument at the concentration 1.2 g/100 mL.

Liquid Chromatography–Mass Spectrometry Measurements High-resolution liquid chromatography–mass spectrometry (LC–MS) analysis of the sample taken from reaction mixture after 4 h in 0.1 M NaOH was performed on an linear trap quadrupole (LTQ) Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) coupled to a HPLC system (CTC, Basel, Switzerland). LC analysis of the prepared samples was performed on a Kinetex C18 column, 150 × 4.6 mm2 , 2.6 :m (Phenomenex, Inc., Torrance, California, USA ) using 0.6 mL min−1 flow rate. The mobile phase consisted of 10 mM ammoDOI 10.1002/jps.24070

Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) Mass spectra were recorded on a MALDI LTQ Orbitrap XL (Thermo Fisher Scientific) equipped with nitrogen UV laser (337 nm, 60 Hz, 8–20 :J) in positive ion mode. For the CID experiment using the LTQ helium was used as the collision gas and 2,5-dihydroxybenzoic acid or trans-2-[3-(4-tertbutylphenyl)-2-methylprop-2-en-1-yliden]malononitrile as the MALDI matrix. Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

2242

RESEARCH ARTICLE – Drug Discovery–Development Interface

pKa Measurement The acidity constant of Ezetimibe was determined from its initial UV–Vis spectra in tertiary amine buffers in an 1-cm closable cell using an Agilent 8453 (Thermo Fisher Scientific) diode array spectrophotometer at 39 ± 0.1◦ C. Methanolic solution (10 :L) of Ezetimibe (c = 9.92 10−3 mol L−1 ) was injected into the thermostated cell containing 2 mL of individual buffer solution, and after quick shaking, the initial spectrum was scanned. The absorbance at 247 nm was plotted against the pH of the individual buffer (see Supplementary Information, Part II), and from the inflexion point of the resulting sigmoid curve, the pKa = 9.74 was calculated. This value corresponds to a dissociation of a proton from the phenolic group. Kinetic Measurements All of the kinetic measurements were carried out in an 1cm closable cell using an Agilent 8453 (Thermo Fisher Scientific) diode array spectrophotometer at 39 ± 0.1◦ C. The observed pseudo-first-order rate constants kobs were calculated from absorbance-time (A/t) dependences at the wavelengths of 246–250 nm and at the substrate concentrations of approximately 5 × 10−5 mol L−1 using the equation A–A∞ = (A0 – A∞ ) e−k obs t , where A, A0 , and A∞ are absorbances at time t, at the beginning, and at the end of the reaction, respectively. The concentration of base buffer component (cB ) was always 0.025 mol L−1 and the ionic strength I = 0.1 mol L−1 was adjusted by addition of KCl. In all the kinetic runs, the standard deviation in the fit was always less than 1% of the quoted value and was more usually between 0.2% and 0.4% of the quoted value. The pH of individual buffers was measured at 39◦ C using a PHM 93 Radiometer (Copenhagen, Denmark) apparatus equipped with a glass electrode (calibration with standard IUPAC buffers at 39◦ C). Alkaline Transformation of Ezetimibe in Dilute NaOH Solution and in Buffer Solutions Ezetimibe (0.5 g; 1.2 mmol) was treated with sodium hydroxide (0.24 g, 6 mmol) in 60 mL of redistilled water for 12 h at 40◦ C. Then, the solution was cooled to room temperature and neutralized with 1 mol L−1 aqueous HCl. Precipitated crude product (2) (0.31 g) was filtered off, dried, and directly analyzed by means of 1 H and 13 C NMR spectroscopy and HPLC. Both techniques confirmed the formation of one unified product (2) with purity of approximately 95%. Analytically pure compound (2) was obtained after crystallization from ethyl acetate–toluene (1:1); melting point 242◦ C–244◦ C. The structure of (2) was confirmed by 1 H NMR and 13 C NMR spectra and X-ray analysis (see Supplementary Information) that were consistent with those published elsewhere.7–9 Another portion (0.05 g) of compound (2) was isolated from the aqueous filtrate after its evaporation, extraction (dry ethanol), and column chromatography [silica gel 60 :m; dichloromethane–ethyl acetate (2:1)]. Finally, the chromatographic column was washed with methanol to give a small amount (∼0.05 g) of a mixture containing according to MS analysis 5-(4-fluorophenyl)-5-hydroxy-2[(4-fluorophenylamino)-(4-hydroxyphenyl)methyl]-pentanoic acid (3) and 6-(4-fluorophenyl)-2-(4-hydroxyphenyl)-3,4,5,6tetrahydro-2H-pyran-3-carboxylic acid (4) as the main components. Unfortunately, their isolation and purification was unsuccessful. The same compound (2) was obtained when Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

Ezetimibe (0.1 g) was treated with triethylamine (0.13 g) solution in redistilled water (13 mL). Alkaline Degradation of Ezetimibe in Concentrated NaOH Methanolic Solution Ezetimibe (0.7 g; 1.7 mmol) was dissolved in 60 mL of redistilled water–methanol mixture (75/25, w/w) and 12.6 mL of 10 M aqueous sodium hydroxide solution was added in one portion so that the final concentration of NaOH was 2 M. The reaction mixture was stirred for 1 h at 40C and then cooled and neutralized by addition of aqueous-concentrated hydrochloric acid. The precipitated product (2) (76 mg; 11% of theory) was filtered off and dried. The filtrate was evaporated and the solid residue was extracted using dry ethanol. The extract was evaporated and the organic residue was chromatographed [silica gel 60 :m; dichloromethane–ethyl acetate (2:1)] and recrystallized from the same solvent mixture to give 270 mg (50%) of off-white crystalline (2R,3R,6S)-6-(4-fluorophenyl)-2-(4-hydroxyphenyl)3,4,5,6-tetrahydro-2H-pyran-3-carboxylic acid (4) with m.p. 232◦ C–258◦ C (decomp.). 1 H NMR (DMSO-d6 , 400 MHz): * = 1.51–1.61 (m, 1H, CH2 ); 1.87–2.02 (m, 2H, CH2 ); 2.12–2.18 (m, 1H, CH2 ); 2.53–2.59 (m, 1H, CH–CO); 4.48 (d, 1H, J 10.4, OCH); 4.58 (d, 1H, J 10.4, OCH), 6.70 (m, 2H, J 8.4, ArOH H3,5 ); 7.12– 7.19 (m, 4H, ArOH-H2,6 + ArF-H2,6 ); 7.37–7.41 (m, 2H, ArFH3,5 ); 9.40 (s, 1H, OH); 12.13 (s, 1H, OH). 13 C NMR (DMSOd6 , 100 MHz): * = 28.1 (CH2 ); 32.2 (CH2 ); 48.6 (CH–CO); 78.3 (OCH); 80.6 (OCH); 114.7 (ArOH-C3,5 ); 114.8 and 115.0 (d, J 20.9, ArF C3,5 ); 127.8 and 127.9 (d; J 8.2, ArF-C2,6 ); 128.6 (ArOH-C2,6 ); 131.2 (ArOH-C1 ); 139.1 (d; J 2.8; ArF-C1 ); 157.1 (ArOH-C4 ); 160.2 and 162.6 (d; J 242.8; ArF-C4 ); 174.6 (C=O). HR MALDI–MS: C18 H17 FO4 [M+Na]+ : measured m/z 339.0996; required m/z 339.1003. Attempt to Further Degrade Compound (2) Under Strongly Alkaline Conditions Compound (2) (0.5 g; 1.2 mmol) was dissolved in 60 mL of 2 M NaOH and the mixture was heated at 40◦ C for 1 h. After cooling and neutralization by addition of aqueous-concentrated hydrochloric acid, only precipitated starting compound (2) (0.45 g) was recovered. (2R*, 3S*)-(4-Fluorophenyl)-4-(4-Hydroxyphenyl)3-Methylazetidin-2-One (5) This compound was prepared in two steps according to a general procedure in Ref. 10 First step: 4-Benzyloxybenzyliden-N-(4-fluorophenyl) imine11 (6.8 g, 22.3 mmol) was dissolved in dry toluene (40 mL) under an argon atmosphere at 80◦ C and tributylamine was added in one portion. Then, a solution of propionyl chloride (2.1 g, 22.5 mmol) in toluene (40 mL) was added dropwise during 30 min, and the reaction mixture was stirred for 2 days at 80◦ C. After cooling, 1 M aqueous hydrochloric acid solution (80 mL) was added, and the resulting solution was stirred for 15 min. The resulting mixture was transferred to a separating funnel, diluted with ethyl acetate (80 mL), washed with 1 M HCl (2 × 100 mL), saturated NaHCO3 (100 mL), water and brine, dried over anhydrous Na2 SO4 , and evaporated. The residue was crystallized from petroleum ether–diethylether (2:1) to give 4.9 g (61%) of white crystalline (3R*,4S*)-4-[4 -(benzyloxy)phenyl]-1-(4-fluorophenyl)-3-methylazetidin-2-one DOI 10.1002/jps.24070

RESEARCH ARTICLE – Drug Discovery–Development Interface

with m.p. 107◦ C–109◦ C. 1 H NMR (CDCl3 -d1 ): * = 1.46 (d; 3H; J 8.0; CH3 ); 3,12 (qd, 1H, CH); 4.51 (d; 1H; J 4.0; CH); 5.05 (s, 2H, OCH2 ); 6.90–6.99 (m, 4H); 7.23–7.27 (m, 4H); 7.33–7.43 (m, 5H); 13 C NMR (CDCl3 -d1 ): * = 13.0 (CH3 ); 55.5 (CH); 62.5 (NCH); 70.1 (OCH2 ); 115.4 (ArOR, C3,5 ); 115.6 a 115.9 (d; J 22.5; ArF-C3,5 ); 118.3 a 118.4 (d; J 7.9; ArF-C2,6 ); 127.2 (2 × CH; ArH); 127.5 (2 × CH; ArH); 128.1 (ArH-C4 ); 128.6 (ArOR-C2,6 ); 129.7 (ArOR-C1 ); 134.0 a 134.1 (d; J 3.0; ArF-C1 ); 136.6 (ArH-C1 ); 157.7 a 160.1 (d; J 243.0; ArF-C4 ); 159.0 (ArOR-C4 ); 168.2 (C=O). HR MALDI–MS: C23 H20 FNO2 [M+H]+ : measured m/z 362.1548; required m/z 362.1551. Second step: (3R*,4S*)-4-[4-(benzyloxy)phenyl]-1-(4fluorophenyl)-3-methylazetidin-2-one (4.8 g, 13.3 mmol) was dissolved in ethyl acetate (60 mL) and 0.93 g of 5% Pd-C was added. The suspension was hydrogenated in an autoclave under 20 bar pressure of hydrogen for 27 h at room temperature and then filtered. The filtrate was evaporated and the residue was crystallized from petroleum ether–diethylether (1:1) to give 2.5 g (69%) of white crystalline product (5) with m.p. 113◦ C–115◦ C. 1 H NMR (CDCl3 -d1 ): * = 1.46 (d; 3H; J 8.0; CH3 ); 3.11 (qd, 1H, CH); 4.50 (d; 1H; J 4.0; CH); 5.21 (s, 1H, OH); 6.83–6.86 (m, 2H); 6.90–6.96 (m, 2H); 7.21 (m, 2H); 7.26 (m, 2H); 13 C NMR (CDCl3 -d1 ): * = 12.9 (CH3 ); 55.3 (CH); 62.7 (NCH); 115.7 a 115.9 (d; J 22.6; ArF-C3,5 ); 116.1 (ArOR-C3,5 ); 118.5 (d; J 7.7; ArF-C2,6 ); 127.3 (ArOR-C2,6 ); 128.9 (ArOR-C1 ); 133.8 (d; J 3.0; ArF-C1 ); 156.3 (ArOR-C4 ); 157.8 a 160.2 (d; J 243.6; ArF-C4 ); 168.8 (C=O). HR MALDI–MS: C16 H14 FNO2 [M+H]+ : measured m/z 272.1082; required m/z 272.1081.

3-[(4-Fluorophenyl)Amino]-3-(4-Hydroxyphenyl)2-Methylpropanoic Acid (6) $-Lactam (5) (0.4 g; 1.5 mmol) was reacted with sodium hydroxide (0.25 g, 6 mmol) in water (30 mL) under an argon atmosphere for 2 days at 40◦ C. Then, the reaction mixture was cooled and neutralized by addition of concentrated hydrochloric acid. The precipitated product was purified by column chromatography (silica gel; DCM:EtOAc, 1:1) and by crystallization to give 50 mg (12%) of off-white product (6) (m. p. 179◦ C–181◦ C). Two sets of signals were found in the 1 H NMR spectrum corresponding to a diastereomeric mixture. Such racemization was previously observed in the literature.10 Major diastereoisomer (58%): 1 H NMR (CD3 OD-d4 ): * = 0.91 (d; 3H; J 6.8; CH3 ); 2.70–2.79 (m, 1H, CO–CH); 4,64 (d; 1H; J 9.2; NH–CH); 6,78 (m; 2H; J 6.8; ArOH-H3,5 ); 7.02–7.07 (m; 2H; ArF-H2,6 ); 7.18–7.22 (m, 2H, ArOH-H2,6 ); 7.58–7.62 (m; 2H; ArF-H3,5 ); 13 C NMR (CD3 OD-d4 ): * = 15.6 (CH3 ); 50.6 (CH–CO); 77.5 (NCH); 116.3 (ArOH-C3,5 ); 116.2 a 116.4 (d; J 22.1; ArFC3,5 ); 123.4 (d; J 8.0; ArF-C2,6 ); 129.4 (ArOH-C2,6 ); 135.1 (ArOHC1 ); 136.0 (d; J 3.0; ArF-C1 ); 158.4 (ArOH-C4 ); 159.7 a 162.1 (d; J 242.5; ArF-C4 ); 176.8 (C=O). Minor diastereoisomer (42%): 1 H NMR (CD3 OD-d4 ): * = 1.34 (d; 3H; J 6.8; CH3 ); 2.70–2.79 (m, 1H, CO–CH); 4.64 (d; 1H; J 9.2; NH–CH); 6.71 (m; 2H; J 6.8; ArOH-H3,5 ); 6.93–6.97 (m; 2H; ArF-H2,6 ); 7.18–7.22 (m, 2H, ArOH-H2,6 ); 7.25–7.29 (m; 2H; ArF-H3,5 ); 13 C NMR (CD3 OD-d4 ): * = 14.9 (CH3 ); 51.2 (CH–CO); 76.9 (NCH); 116.0 (ArOH-C3,5 ); 116.1 a 116.3 (d; J 23.1; ArFC3,5 ); 123.8 a 123.9 (d; J 8.0; ArF-C2,6 ); 129.0 (ArOH-C2,6 ); 135.4 (ArOH-C1 ); 135.7 (d; J 3.0; ArF-C1 ); 158.1 (ArOH-C4 ); 159.6 a 162.0 (d; J 242.5; ArF-C4 ); 175.9 (C=O). DOI 10.1002/jps.24070

2243

HR MALDI-MS: C16 H16 FNO3 [M+Na]+ : measured m/z 312.1006; required m/z 312.1006.

RESULTS Initially, the Ezetimibe degradation was studied qualitatively in sodium hydroxide solutions (pH 12, c = 0.01 M; pH 13, c = 0.1 M) at 40◦ C using HPLC and LC–MS. In 0.01 M, NaOH only degradation product (2) was found even after 24 h, whereas in 0.1 M NaOH Ezetimibe was completely degraded after 4 h to give a mixture of (2) and (3) in the ratio 4:1. If this mixture was further left to stand at 60◦ C for another 24 h, then the amino acid (3) completely disappeared and new substituted tetrahydropyrancarboxylic acid (4) was formed. This result confirms that the compound (4) is formed from acid (3) and not through the hydrolysis of amide (2) (see Scheme 1). The two main degradation products—amide (2) and acid (4) were isolated and fully characterized (see Materials and Methods section and Supplementary Information). Unfortunately, all attempts to isolate pure acid (3) failed. It was only possible to characterize it using LC–MS (see Supplementary Information). The basecatalyzed transformation of Ezetimibe giving amide (2) as the only product at pH of less than 12.5 and the mixture of amide (2) and acid (3) at pH of more than 12.5 was further studied spectrophotometrically under pseudo-first-order conditions in aqueous sodium hydroxide solutions (c = 0.02–2 mol L−1 ) and in aqueous tertiary amine buffers (pH 7–12) at constant ionic strength (I = 0.1 mol L−1 ) and at 39◦ C. All time-resolved spectra recorded in NaOH solutions showed sharp isosbestic points (Fig. 1) that indicate that no stable intermediate accumulates to detectable levels and that the subsequent transformation of

Figure 1. Time-resolved spectra for the transformation of Ezetimibe in 0.02 mol L−1 aqueous sodium hydroxide solution and kinetic curve obtained at 247 nm (inset, solid circles). Note: some time resolved spectra were omitted for clarity of the Figure 1 but corresponding absorbances at 247 nm are shown in kinetic curve as open circles. Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

2244

RESEARCH ARTICLE – Drug Discovery–Development Interface

Table 1. The Dependence of the Observed Rate Constants kobs on the Concentrations of Triethylamine (TEA) Buffers with [B]/[BH] = 1:1, N,N-Dimethylaminoethanol (DMAE) Buffer with [B]/[BH] = 1:1, and N-Methylmorpholine (NMM) Buffer with [B]/[BH] = 4:1 Measured at 39◦ C and at Ionic Strength I = 0.1 M. Amine TEA

DMAE

NMM

cB (mol L−1 )

cBH + (mol L−1 )

pH

104 ·kobs (s−1 )

0.05 0.10 0.15 0.20 0.0125 0.0250 0.0500 0.0750 0.1000 0.0100 0.0125 0.0250 0.0350 0.0500

0.05 0.10 0.15 0.20 0.0125 0.0250 0.0500 0.0750 0.1000 0.0025 0.00313 0.00625 0.00875 0.0125

11.11 11.12 11.10 11.10 9.41 9.42 9.41 9.40 9.39 8.21 8.22 8.23 8.22 8.23

7.27 7.25 7.30 7.00 8.43 8.22 8.03 8.22 8.25 1.23 1.20 1.12 1.32 1.13

aminoacid (3) to acid (4) does not interfere with its formation from Ezetimibe. However, replacing sodium hydroxide solutions by the tertiary amine buffers, the buffer components absorption partially overlaps that of the substrate/product, so that the isosbestic point is no longer observable. Fortunately, the kinetic curves measured at the maximum of the absorption band of Ezetimibe (247 nm) were still single exponentials, from which pseudo-first-order rate constants kobs were easily accessible. The tertiary amines were adopted because they cannot give any aminolysis products and they behave as bases. First, buffer catalysis was examined and it was found that increasing the concentration of both buffer components had only negligible influence (Table 1) on the reaction rate from which it can be concluded that there is no buffer catalysis and the rate of transformation is controlled by the pH of the medium only (specific acid-base catalysis). From the measurements of rates in the individual buffers and in sodium hydroxide solutions, the corresponding pH profile (Fig. 2) was generated. There are two distinct breaks in the pH profile in Figure 2. In neutral and weakly basic medium, the log kobs increases linearly with pH up to approximately 9 and the slope of line is close to unit value. Between pH 9 and 11, a break downwards occurs and the values of log kobs decrease with increasing pH of the medium. Then, another break upwards occurs, and from pH of approximately 13, a linear increase of log kobs is observed again with slope approaching to unity. For comparison, the hydrolysis (Scheme 2) of the structurally similar 1-(4-fluorophenyl)-4-(4-hydroxyphenyl)-3methylazetidin-2-one (5) giving 3-[(4-fluorophenyl)amino]-3-(4hydroxyphenyl)-2-methylpropanoic acid (6) was also studied in sodium hydroxide solutions. For this compound, the rearrangement is impossible because of the absence of the side chain containing the hydroxyl group, and the pH profile in alkaline medium is a simple straight line with unit slope (see Fig. 2). Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

Figure 2. pH profile for Ezetimibe transformation measured in N-methylmorpholine (•), N-methyldiethanolamine (◦), N,Ndimethylaminoethanol (), and triethylamine buffers () and in sodium hydroxide solutions (NaOH; ) and for the hydrolysis of (2R*, 3S*)-1-(4-fluorophenyl)-4-(4-hydroxyphenyl)-3-methylazetidin-2one (5, ×) in sodium hydroxide solutions. The line for Ezetimibe represents the best fit of all data points using Eq. (7) with parameters given in the text. The concentration of the basic buffer component cB was always 0.025 mol L−1 .

DISCUSSION Ezetimibe is not expected to undergo facile hydrolysis because of the lack of functional groups that hydrolyze under neutral conditions. Even in strongly alkaline media, azetidin-2-ones ($-lactams) undergo slow hydrolysis to give $-amino acids12 at room or slightly elevated temperature. It means that Ezetimibe preferentially undergoes base-catalyzed rearrangement of the $-lactam ring to give a tetrahydropyran ring instead of possible hydrolysis of the amide group (Scheme 1) at pH of less than 12.5. The rearrangement proceeds with intramolecular nucleophilic displacement of the N-4-fluorophenylamide group by the alkoxide formed in a very fast pre-equilibrium from (1). It is clear that the reaction takes place at the chiral carbon C-4 of the $-lactam ring with the inversion in absolute configuration (S → R), and this inversion was confirmed by X-ray analysis of a sample of (2) (see Supplementary Information). It is well known that the inversion of absolute configuration is an inherent and consistent characteristic of the bimolecular nucleophilic substitution SN 2. Therefore, the rearrangement of (1) to (2) must follow this reaction mechanism. The preference of the above-mentioned nucleophilic displacement in comparison with simple hydrolysis of $-lactam ring can be explained by comparison of the molecularity of both reactions. Intramolecular reactions (1 → 2) are always much faster as compared with their intermolecular counterparts because of the formation of an entropically and thermodynamically favorable sixmembered ring.13 One would expect that the formation of product (4) in strongly basic solutions occurs via consecutive hydrolysis of the DOI 10.1002/jps.24070

RESEARCH ARTICLE – Drug Discovery–Development Interface

2245

Scheme 2. Alkaline hydrolysis of azetidinone 5.

amide bond in (2), that is, after the rearrangement. However, when it was tried to independently prepare compound (4) from compound (2) under the same reaction conditions (see Material and Methods) it was surprisingly found that no reaction takes place and only amide (2) was recovered. This observation was also proven spectrophotometrically under pseudo-firstorder conditions. No spectral change of the starting amide (2) was observed in aqueous–methanolic (or even aqueous) NaOH solution during 1 h at 39◦ C. From this observation and from above-mentioned HPLC experiment, it can be concluded that the acid (4) has to be formed by a different reaction pathway involving another reactive intermediate. From the shape of the pH profile, several key points can be concluded for the reaction mechanism. The first transition from unit to zero slope in the pH profile at pH 9.5–10 must correspond to the formation of the conjugated base (phenoxide) of the starting Ezetimibe, and the break around pH 10 gives the value of the apparent ionization constant pKa(1). This is completely consistent with pKa(1) = 9.74 measured spectrophotometrically (see Material and Methods and Supplementary information). The second transition from zero to negative slope means that the formation of some unreactive (or less reactive) intermediate occurs and the overall rate of transformation decreases. The slope value is not an integral number (ca.−0.3 only), although a value around −1 would be expected.14 However, in a relatively narrow pH region (approximately two pH units), the slopes can change from one integral value to another, giving intermediate values, for example, during hydrolysis of aspirin15,16 or rearrangement17 of S-(2-oxotetrahydrofuran-3yl)-N-(4-methoxyphenyl)isothiuronium bromide. In our case, at pH 10–12, it is likely the slope in pH profile changes from 0 to −1 and back to 0 giving the intermediate value. The last transition to positive unit value means that some new reaction pathway gradually opens. This is completely consistent with the appearance of a new Ezetimibe hydrolysis product 3 in the reaction mixture in concentrated sodium hydroxide solutions. According to the kinetic measurements, the following Scheme 3 can be suggested for the rearrangement of Ezetimibe to (2) and (3) and transformation of (3) to (4). On the basis of Scheme 3, the following rate Eq. (1) can be derived: v = k1 [SH− ] + k2 [S2− ] + k3 [S2− ][OH− ]

initial substrate concentration cS (cS = [SH2 ]+[SH− ]+[S H− ] + [S2− ]). K A1 =

[SH− ][H2 O] [SH2 ][OH− ]

(2)

K F1 =

[S H− ][H2 O] [SH2 ][OH− ]

(3)

K AF 2 =

[S2− ][H2 O] [SH− ][OH− ]

(4)

K FA 2 =

[S2− ][H2 O] [S H− ][OH− ]

(5)

(1)

All the reactive species (SH− , S H− , and the dianion S2− ) are formed in very fast acid-base pre-equilibria (specific base catalysis is involved), so that their equilibrium concentrations can be derived from Eqs. (2–5) and their sum including the concentration of the original substrate (SH2 ) gives the total DOI 10.1002/jps.24070

Scheme 3. Ezetimibe alkaline degradation mechanism.

By combination of all above-mentioned equations, the extended rate Eq. (6) can be derived from which the relationship Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

2246

RESEARCH ARTICLE – Drug Discovery–Development Interface

(Eq(7). between the kobs and the hydroxide concentration (or pH—if logarithmical scale is used) is apparent: − 2

v=

− 3

[OH ] [OH ] A AF k1 K A1 cS [OH− ] + k2 K A1 K AF 2 cS [H O] + k3 K 1 K 2 cS [H O] 2

2

− − [H2 O] + (K A1 + K F1 + K A1 K AF 2 [OH ])[OH ]

= kobs cS

(6)

− 2

kobs =

− 3

[OH ] A AF [OH ] k1 K A1 [OH− ] + k2 K A1 K AF 2 [H O] + k3 K 1 K 2 [H O] 2

2

− − [H2 O] + (K A1 + K F1 + K A1 K AF 2 [OH ])[OH ]

(7)

The excellent agreement between the calculated line and the experimental data in Figure 2 supports the validity of the suggested kinetic model shown in Scheme 2. The following parameters were optimized from the experimental data in Figure 2 using Eq. (7): k1 K1 A = (2.09 ± 0.16)·103 s−1 , k2 = (2.71 ± 0.35)·10−2 s−1 , k3 = (5.63 ± 0.58)·10−2 L mol−1 s−1 , K1 F = (1.96 ± 0.28)·106 , and K1 A K1 AF = (2.78 ± 0.28)·108 . From the optimized value of K1 F , it is possible to calculate (using equation Ka = KW ·K1 F /[H2 O]) an apparent pKa(1) = 8.98 for phenolic group of Ezetimibe. The value of the ionic product of water18 pKW = 13.5348 and concentration of water in water [H2 O] = 55.1 mol L−1 at 40◦ C were adopted. The calculated value pKa(1) = 8.98 is somewhat lower than the value determined spectrophotometrically from the initial spectra of starting Ezetimibe [pKa(1, spect.) = 9.74] that is probably caused by the less precise kinetic determination from the multiparameter kinetic equation. Under presumption that the pKa of alcoholic group in Ezetimibe is close to the value known for simple benzylalcohol (pKa = 15.4), it is possible to estimate (using equation K1 A = Ka ·[H2 O]/KW = 0.75) from the product k1 ·K1 A the value k1 ≈ 2780 s−1 . This very high value is in accordance with intramolecular attack of alkoxide giving thermodynamically favorable13 six-membered ring. On the contrary, the low-value k2 shows that the dianion S2− is much less reactive because of delocalization of negative charge that lowers the electrophilicity of carbon C4 toward the internal nucleophile– alkoxide. The value k3 = 5.63 10−2 L mol−1 s−1 corresponding to hydrolysis giving (3) is similar to the values12,19 observed for hydroxide-catalyzed hydrolyses of various $-lactams. For example, the second-order rate constant measured for structurally related 1,4-diphenylazetidin-2-one with hydroxide in water at 35◦ C is19 kOH = 4.8 10−3 L mol−1 s−1 and for our azetidin-2-one (5) kOH = 9.82 10−4 L mol−1 s−1 (Fig. 2). To get better insight into the reaction mechanism at pH of more than 12.5, solvent kinetic isotope effect was also studied that can give more information about the nature of the intermediate or transition state. An inverse solvent kinetic isotope effect is typical20,21 for simple hydrolysis of $-lactams where nucleophilic attack of hydroxide ion to a $-lactam carbonyl giving tetrahedral intermediate is rate limiting. On the contrary, the normal solvent kinetic isotope effect is occasionally observed21 in those cases when the rate-limiting step involves water-catalyzed decomposition of a tetrahedral intermediate. For Ezetimibe, the rate of transformation in NaOD–D2 O solutions is slower than in NaOH–H2 O solutions and kH2O –kD2O = 1.29, whereas for azetidin-2-one (5), the expected inverse solvent kinetic isotope effect kH2O –kD2O = 0.79 was determined. From this observation, it could be concluded that for Ezetimibe, the rate-limiting Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014

step involves decomposition of tetrahedral intermediate (In3 − in Scheme 2) and the observed solvent kinetic isotope effect arises22 from reordering of several water molecules in the C–N bond cleavage. However, the observed solvent kinetic isotope effect for azetidin-2-one (5) is inverse, although the structural change in side chain at position 3 should not affect the nature of the effect. Therefore, another explanation should be taken into account. The theory suggests that the deprotonation of the benzylic hydroxy group should have a solvent isotope effect of +2 that is followed by a rate-limiting formation of the In3 − for which the normal value around 0.8 would be expected. Therefore, the combined solvent isotope effect for a pathway involving attack of OH− on In3 − would have value 2 × 0.79 = 1.58 that is close to the observed value (1.29). A similar positive value of solvent isotope effect was also found during cyclizations of amides to imidazolinones.23

CONCLUSIONS In our article, we have characterized the structure of the degradation products and provided a kinetic study of the degradation of Ezetimibe in weakly and strongly alkaline solutions. We have determined the pH profile for Ezetimibe degradation and identified three degradation products formed at high pH whose structures were not previously been published. On the basis of the measured kinetic data and HPLC analysis, we propose the reaction mechanism for the formation of individual degradation products and provide a kinetic equation that fits all the measured data. From our results, it is clear that between pH 7 and 12.5 the main degradation product is amide (2) but from pH 12 two new degradation products—$-aminoacid (3) and substituted tetrahydropyrancarboxylic acid (4)—gradually start to form. Surprisingly, it was found that the tetrahydropyrancarboxylic acid (4) is not formed from pyran-carboxamide (2) via the expected amide-bond hydrolysis, but from the $-aminoacid (3).

ACKNOWLEDGMENT Institutional support by the Ministry of Education, Youth and Sports of the Czech Republic is gratefully acknowledged.

REFERENCES 1. Heek MV, Farley C, Compton D, Hoos L, Davies H. 2001. Ezetimibe selectively inhibits intestina cholesterol absorption in rodents in the presence and absence of exocrine pancreatic function. Br J Pharmacol 134:409–417. 2. Singh S, Singh B, Bahuguna R, Wadhwa L, Saxena R. 2006. Stress degradation studies on ezetimibe and development of a validated stability-indicating HPLC assay. J Pharm Biomed Anal 41:1037–1040. 3. Dixit RP, Barhate CR, Nagarsenker MS. 2008. Stability-indicating HPTLC method for simultaneous determination of Ezetimibe and simvastatin. Chromatographia 67:101–107. 4. Pandey S, Rathanand M. 2010. Development and validation of stability indicating reversed-phase HPLC method for the determination of ezetimibe in pharmaceutical dosage forms. Int J Pharm Sci Rev Res 1:53–57. 5. Gajjar AK, Shah VD. 2011. Isolation and structure elucidation of major alkaline degradant of Ezetimibe. J Pharm Biomed Anal 55:225– 229. 6. Barhate CR, Mohanraj K. 2011. What is the degradation product of ezetimibe? J Pharm Biomed Anal 55:1237–1238. DOI 10.1002/jps.24070

RESEARCH ARTICLE – Drug Discovery–Development Interface

´ ´ 7. Filip K, Bankowski K, Sidoryk K, Zagrodzka J, Łaszcz M, Trzcinska K, Szyprowska A, Cmoch P, Maruszak W. 2011. Physicochemical characterization of ezetimibe and its impurities. J Mol Struct 991:162– 170. ´ ´ 8. Santa Z, K´oti J, Sz˝oke K, Vukics K, Szantay C Jr. 2012. Structure of the major degradant of ezetimibe. J Pharm Biomed Anal 58:125– 129. 9. Swamy GYSK, Ravikumar K, Wadhwa LK, Saxena R, Singh S. 2005. (2R*,3R*,6S*)-N,6-Bis(4-fluorophenyl)-3,4,5,6-tetrahydro-2Hpyran-3-carboxamide. Acta Crystallogr E 61:o3608–o3610. 10. Browne M, Burnett DA, Caplen MA, Chen LY, Clader JW, Domalski M, Dugar S, Pushpavanam P, Sher R, Vaccaro W, Viziano M, Zhao H. 1995. Trans-diastereoselective synthesis of 3-alkyl substituted beta-lactams via the acid chloride-imine reaction of non-activated acid chlorides. Tetrahedron Lett 36:2555–2558. 11. Guillon J, Stiebing S, Robba M. 2000. First synthesis of N-(4fluorophenyl)-5-(4-fluorophenyl)-2-(4-hydroxybenzyl)pentanamide, a new potential cholesterol absorption inhibitor pharm. Pharm Comm 6:25–29. 12. Blackburn GM, Placket JD. 1972. Strain effects in acyl transfer reactions. Part I. The kinetics of hydrolysis of some N-aryl-lactams. J Chem Soc, Perkin Trans II 10:1366–1371. 13. Mandolini L. 1986. Intramolecular reactions of chain molecules. In Advances in physical organic chemistry; Gold V, Bethell D, Eds. London, UK: Academic Press, pp 1–111. 14. Birk JP. 1976. The use of log-log plots in the determination of reaction orders. J Chem Educ 53:704–707. 15. Edwards LJ. 1950. The hydrolysis of aspirin. A determination of the thermodynamic dissociation constant and a study of the reaction kinetics by ultra-violet spectrophotometry. Trans Faraday Soc 46:723– 735.

DOI 10.1002/jps.24070

2247

16. Garrett ER. 1957. The kinetics of solvolysis of acyl esters of salicyclic acid. J Am Chem Soc 79:3401–3408. ˇ J, Sedlak ´ M, Hanusek J. 2010. Kinetic evidence for the 17. Va´ na coexistence of zwitterionic (T± ), neutral (T0 ) and anionic (T− ) intermediates during rearrangement of S-(2-oxotetrahydrofuran-3-yl)N-(4-methoxyphenyl) isothiuronium bromide to 5-(2-hydroxyethyl)-2(4-methoxyphenylimino)-1,3-thiazolidin-4-one. J Org Chem 75:3729– 3736. 18. Lide DR, Ed. 2005. CRC handbook of chemistry and physics, internet version 2005. Boca Raton, Florida: CRC Press, pp 8– 85. 19. Butler AR, Freeman KA, Wright DE. 1977. A mechanistic study of the hydrolysis of $-lactams. In Recent advances in chemistry of $lactam antibiotics; Elks J, Ed. Vol. 28. London, UK: Chemical Society, pp 299–303. 20. Deraniyagala SA, Adediran SA, Pratt RF. 1995. $-Secondary and solvent deuterium kinetic isotope effects and the mechanisms of baseand acid-catalyzed hydrolysis of penicillanic acid. J Org Chem 60:1619– 1625. 21. Brown RS, Bennet AJ, Slebocka-Tilk H. 1992. Recent perspectives concerning the mechanism of H3 O+ - and hydroxide-promoted amide hydrolysis. Acc Chem Res 25:481–488. 22. Slebocka-Tilk H, Bennet AJ, Hogg HJ, Brown RS. 1991. Predominant 18 O exchange accompanying base hydrolysis of a tertiary toluamide: N-ethyl-N-(trifluoroethy1)toluamide. Assessment of the factors that influence partitioning of anionic tetrahedral intermediates. J Am Chem Soc 113:1288–1294. ´ M, Hanusek J, B´ına R, Kavalek ´ ´ cek V. 1999. Kinet23. Sedlak J, Machaˇ ics and mechanism of the formation and decomposition of imidazolin-4ones from 2-(N-benzoylamino)alkanamides in aqueous medium. Collect Czech Chem Comm 64:1629–1640.

Baˇtov´a et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:2240–2247, 2014