ANALYTICAL SCIENCES DECEMBER 1998, VOL. 14 1169 1998 © The Japan Society for Analytical Chemistry

Notes Identification of the Major Degradation Products of by Frit-FAB LC/MS and NMR

Kazuo YAMAGUCHI†, Tohru YASUZAWA and Satoshi KOBAYASHI

Analytical & Pharmacokinetic Laboratories, Pharmaceutical Research Institute, Kyowa Hakko Kogyo Co., Ltd., 1188 Shimotogari, Nagaizumi-cho, Shizuoka 411, Japan

Keywords Frit-FAB LC/MS, NMR, structure determination, vinorelbine, degradation product

Navelbine¨(Vinorelbine ditartrate) is a semi-synthetic (LKB 2249, Bromma, Sweden), an injector (7125, vinca alkaloid first obtained by Mangeney et al.1 It has Rheodyne, Cotati, CA) equipped with a 20 µl loop, a a broader anti-tumor spectrum and less toxicity than the Develosil ODS HG-5 column (150 mm×4.6 mm i.d., naturally occurring vinca alkaloids, and vin- Nomura Chemical, Tokyo, Japan) and a UV detector blastine, and is currently used clinically.2 Vinorelbine (2151, LKB). It was connected to a Frit-FAB interface (5′-nor-anhydrovinblastine, 1, Fig. 1) shares a similar (MS-FF, JEOL) attached to a MS spectrometer (JMS- dimeric structure composed of and vindo- SX102, JEOL). The HPLC elution was carried out line moieties with vincristine and . These with a mixture of 50 mM ammonium acetate (pH 4.5) vinca alkaloids undergo chemical degradation under and methanol (1:1, v/v) containing 1% of glycerol as several kinds of stresses, but knowledge of the struc- the FAB-MS matrix. The flow rate was 1 ml/min and tures of the degradation products is limited3, although the UV monitoring was done at 267 nm. The FAB-MS the drugs have been long and widely used clinically. measurements were done at a xenon acceleration volt- Investigation of stability profiles after various stresses age of 5 kV and an ion acceleration voltage of 8 kV. revealed that vinorelbine is susceptible to degradation under highly humid conditions, giving several kinds of NMR analysis of degradation products products. In this note we describe their structures Isolation of the major degradation products was car- newly identified by using Frit-FAB (fast atom bom- ried out on the above-described HPLC system eluted bardment)4 liquid chromatography / mass spectrometry with the same solvent. Peak fractions identified by UV (LC/MS) and nuclear magnetic resonance spectrometry monitoring were collected, neutralized with 0.1 M (NMR). NaOH and extracted with chloroform. The chloroform layer was dried under vacuum, reconstituted with DMSO-d6 (ca. 30 mg/ml) and submitted to NMR mea- Experimental surements. 1H- and 13C-NMR spectra were obtained at 303 K by using a Fourier transform NMR spectrometer Material (AMX-400, 9.4 T, 1H: 400 MHz 13C: 100.6 MHz, Vinorelbine was synthesized and provided by Pierre Bruker, Karlsruhe, Germany) and a data processing Fabre Medicament (Castres, France). Dimethyl- system (Aspect X32, Spectrospin, Faellanded, sulfoxide-d6 (DMSO-d6) was purchased from CEA (Commisariat a l′Energie Atomique, Gif-sur Yvette, France). Other chemicals and solvents were of reagent grade and commercially available.

Frit-FAB LC/MS analysis of vinorelbine stored under high humidity Vinorelbine (1 g) was stored at 40ûC under 75% rela- tive humidity for 7 d. The degradation products were dissolved in 100 ml of methanol/water (1:1 v/v) and an aliquot of 20 µl was injected onto a Frit-FAB LC/MS composed of a high performance liquid chromatograph (HPLC) and a mass spectrometer. The HPLC system consisted of a gradient pump

† To whom correspondence should be addressed. Fig. 1 Chemical structure of vinorelbine. 1170 ANALYTICAL SCIENCES DECEMBER 1998, VOL. 14

Switzerland). Chemical shifts were referenced to inter- also gave fragment ions, m/z 443 and m/z 323. The for- nal tetramethylsilane (TMS: 0.00 ppm). mer is 14 Da smaller than that of the vindoline moiety. Thus 3 was considered to be a product produced from vinorelbine by demethylation at its vindoline moiety. Results and Discussion Compound 4 gave a mass spectrum very similar to that of vinorelbine suggesting that it is a structural isomer Figure 2 shows a total ion chromatogram of the or a multiply modified form of 1. In order to determine degradation product of vinorelbine on the FRIT-FAB the structures of the major degradation products 2, 3 LC/MS in comparison with a UV chromatogram simul- and 4, they were further purified by HPLC and investi- taneously monitored. Both chromatograms showed a gated by NMR. similar pattern, confirming that the HPLC separation The 1H and 13C-NMR spectra of 2 resembled those of was not impaired during the FRIT-FAB measurement. 1. Observation of only a few impurity signals suggested Table 1 summarizes the characteristic ion peaks that 2 had been adequately purified. Close and compar- observed in the mass spectra of the LC/MS peaks. ative examination of the difference of the NMR spectra Vinorelbine gives a prominent quasimolecular ion (Table 2; signal assignments were done by extensive (MH+) at m/z 779 and characteristic fragment peaks at 2D-NMR measurements including COSYDQFPH, m/z 457 and m/z 323 which could be assigned to the NOESYPH, HMBC and HSQC, and by comparison of vindoline and catharanthine moieties respectively (Fig. chemical shifts with published values5) revealed that 1). The most abundant degradation product, 2, gave multiple ion peaks at m/z 869, 809, 795 and 777. Compound 3 gave a quasimolecular ion [MH+], m/z 765 which is 14 Da smaller than that of vinorelbine. It Table 1 LC/MS peaks of degradation products and character- istic FAB-MS ion peaks

LC/MS Mass number (m/z) Tentative assignment peak

a 795, 485, 323 1-Nb′-oxide b 751, 441, 323 20-deethyl-1 c 765, 455, 323 ? d 765, 455, 443, 323 3 e 779, 469, 457, 323 4 f 809, 793, 483, 323 mixture g 869, 809, 781, demethyl-2 763, 441, 323 h 809, 499, 323 methanol adduct of 2 i 809, 499, 323 isomer of f j 869, 809, 795, 2 777, 455, 323 Fig. 2 Chromatograms of degradation products of vinorel- k 791, 323 methyl-2 bine: (A) UV chromatogram (268 nm), (B) total ion LC/MS l 779, 457, 323 1 chromatogram.

Table 2 Analysis of the 1H and 13C NMR spectral data for vinorelbine(1) and the major degradation products Proton/carbon Vinorelbine(1) 23 4

H-3a 2.68(br.d, J=ca.16 Hz) 5.94(d, J=ca.7 Hz) 2.64(br.d, J=16.2 Hz) 2.30(br.d, J=ca. 9 Hz) H-3b 3.20(br.dd, J=16.7, 4.6 Hz) — 3.22(br.dd, J=16.2 Hz) ca.2.5(br.) H-14a 5.76(ddd, J=10.1, 5.1, 1, ca.3 Hz) 4.78(br.dd, J=6.5 Hz, ca.7 Hz) 5.76(br.dd, J=10.2, 5.1 Hz) 1.58(br.dt, J=ca. 3, ca.13 Hz) H-14b — — — 1.80(br.d, J=ca.13 Hz) H-15 5.24(br.d, J=10.1 Hz) 4.03(br.d, J=6.5 Hz) 5.23(br.d, J=10.2 Hz) 3.89(br.s) H-21 2.58(s) 3.36(br.s) 2.51(s) 3.08(br.s) 16-COOCHÐ 3 3.63(s) 3.70(s) — 3.73(s)

C-3 49.6 139.3 49.8 40.7 C-14 124.0 97.9 123.8 24.0 C-15 130.2 73.9 30.4 76.7 C-16 79.4 85.1 78.6 87.2 C-20 42.0 46.5 42.1 46.0 16-COOCHÐ 3 51.5 51.9 — 51.8 Numbers are chemical shifts(ppm) referenced to internal tetramethylsilane(0.00 ppm). Abbreviations of signal multiplicity in parenthe- ses are as follows. s, singlet; d, doublet; dd, double doublet; dt, double triplet; br, broad. ANALYTICAL SCIENCES DECEMBER 1998, VOL. 14 1171

one methine carbon signals were also observed in its 13C-NMR spectrum (40.7 ppm, 24.0 ppm and 76.7 ppm respectively). Thus 4 was presumed to be an analog of 2 reduced at the double bond between C-3 and C-14. The molecular weight determined by FAB-MS described above agreed with the structure (Fig. 3). The 15′,16′-ether linkage of 2 and 4 at its vindoline moiety was already reported by several researchers.6,7 The reaction was proposed to begin with one electron Fig. 3 Chemical structures of the degradation products 2, 3 oxidation at a nitrogen atom and to proceed via a reac- and 4. tive imminium intermediate. Given the structure of 2, its molecular weight was calculated to be 776. Thus the ion peaks at m/z 795, m/z 809 and m/z 865 observed the catharanthine moiety was intact in 2. The most in its FAB-MS spectrum were presumably those of arti- striking change of the 1H NMR was observed among facts that could be assigned to adducts of water, the signals assigned to protons of the unsaturated methanol, and glycerol respectively. Adduct ions pyrimidine ring of the vindoline moiety. The two caused by reaction with matrix components during FAB olefinic proton signals of vinorelbine at 5.76 ppm and were reported previously.8 5.24 ppm were shifted to 5.94 ppm and 4.78 ppm, and a The substructure of compound 4, a reduced form of new putatively oxygen-attached methine signal vindoline ether, was reported to be generated by appeared at 4.03 ppm. The adjacency of these three microorganism biotransformation.6 Although the reac- protons was confirmed by their COSY connectivity. tion mechanism under the conditions of the present These results suggested the presence of an ether linkage study is not clear, photochemical abstraction of a between C-15 and C-16 and a enamine-type double hydrogen atom might be responsible. bond between C-3 and C-14. The assignments of 13C signals at 139.3 ppm, 97.9 ppm and 85.1 ppm to C-3, C-14 and C-16 respectively, established by a CH- References COSY measurement, and a long-range coupling between C-17 and C-15 identified by a COLOC mea- 1. P. Mangeney, R. Z. Adriamialisoa, J. Y. Lallemand, N. surement, further supported the structure (Fig. 3). Lamgolois, Y. Lamgolois and P. Potier, Tetrahedron, 35, Comparison of the 1H-NMR spectrum of 3 with that 2175 (1979). of 1 (Table 2) revealed that a methyl group that had 2. K. L. Goa and D. Faulds, Drugs & Aging, 5, 200 (1994). been assigned to the 16-COOMe in the latter was 3. K. N. Thimmaiah and V. S. Sethi, Cancer Res., 45, 5382 (1985). absent from the former. This observation was also con- 4. R. M. Caprioli, “Continuous-Flow Fast Atom 13 firmed by C-NMR measurements. These results, Bombardment Mass Spectrometry”, p. 93, John Wiley and together with that of the FAB-MS measurement Sons, Chichester, New York, Brisbane, Toronto, Singapore, described above, implied that 3 is a 16-COOMe- 1990. demethylated form of 1 (Fig. 3). 5. T. D. Spitzer, R. C. Crouch and G. E. Martin, J. Comparison of the 1H-NMR spectrum of 4 with that Heterocyclic Chem., 294, 265 (1992). of 1 (Table 2) suggested that the catharanthine moiety 6. T. Nabih, L. Youel and J. P. Rosazza, J. Chem. Soc. Perkin is intact in 4. Only the signals of the piperidine ring of Trans I, 1978, 757. the vindoline moiety showed a difference. That is, the 7. S. A. Elmaracby, M. W. Duffel and J. P. Rosazza, J. Med. two olefinic protons of vinorelbine (5.76 ppm 5.24 Chem., 32, 2158 (1989). 8. S. Naylor, M. Kajbaf, J. H. Lamb, M. Jahanshahi and J. W. ppm) disappeared and two types of mutually coupled Gorrod, Biol. Mass Spectrom., 21, 165 (1992). methylene protons (2.30 ppm, 2.52 ppm and 1.80 ppm, 1.58 ppm), and one putatively oxygen-bound methine (Received June 5, 1998) proton (3.89 ppm) coupled with the methylene protons (Accepted August 25, 1998) were observed. The corresponding two methylene and