Pharmaceutical Research, Vol. 13, No. 12, 1996
`
`Report
`
`Characterization of a Solid State
`
`Reaction Product from a
`
`Lyophilized Formulation of a Cyclic
`Heptapeptide. A Novel Example of
`an Excipient-Induced Oxidation
`
`David C. Dubost,1 Michael J. Kaufman,"3
`Jeffrey A. Zimmerman,1 Michael J. Bogusky,2
`Arthur B. Coddington,2 and Steven M. Pitzenberger2
`
`Received April 25, I996; accepted August 31, I 996
`
`Purpose. To elucidate the structure of a degradation product arising
`from a lyophilized formulation of a cyclic heptapeptide, and to provide
`a mechanism to account for its formation.
`Methods. Preparative HPLC was used to isolate the degradate in quan-
`tities sufficient for structural studies. A structure assignment was made
`on the basis of the compounds spectroscopic properties (UV, MS,
`NMR) and the results of amino acid analysis.
`Results. The degradate was identified as a benzaldehyde derivative
`arising from the oxidative deamination of an aminomethyl phenylala-
`nine moiety. The extent of formation of this product is influenced by
`the amount of mannitol used as an excipient in the formulation. A
`mechanism is proposed whereby reducing sugar impurities in mannitol
`act as an oxidizing agent via the interrnediacy of Schiff base adducts
`which subsequently undergo tautomerization and hydrolysis.
`Conclusions. Reducing sugar impurities in mannitol are responsible
`for the oxidative degradation of the peptide via a mechanism that
`involves Schiff base intermediates. This mechanism may be a potential
`route of degradation of other arylmethyl amines in mannitol-based
`formulations.
`
`KEY WORDS: peptide stability; peptide formulation; oxidation;
`fibrinogen receptor antagonist; mannitol.
`
`INTRODUCTION
`
`acetylcysteine-asparagine-(5,5-dimethyl-4-
`L-367,073,
`thiazolidinecarbonyl)-4-aminomethyl—phenylalanine)—glycine-
`aspartic acid-cysteine cyclic disulfide, is a potent and specific
`fibrinogen receptor antagonist with potential utility in the treat-
`ment of a variety of cardiovascular conditions ( 1). Structurally,
`the compound is based on the known affinity of the GP IIb/
`IIIa receptor for the RGD (arginine-glycine-aspartic acid)
`sequence in fibrinogen (2). L-367,073 is a synthetic heptapep—
`tide which is cyclized through a disulfide linkage (Figure l).
`The compound lacks significant oral activity and must be
`administered by the intravenous route. Preformulation studies
`have demonstrated that the peptide does not possess sufficient
`solution stability to be formulated as a pre-made solution, and
`therefore a lyophilized formulation for reconstitution was devel-
`oped for clinical use. The stability of the drug in the lyophilized
`
`‘ Pharmaceutical Research and Development, Merck Research Labora-
`tories, West Point, Pennsylvania 19486.
`2 Medicinal Chemistry, Merck Research Laboratories, West Point,
`Pennsylvania 19486.
`3 To whom correspondence should be addressed.
`
`formulation was monitored and found to be entirely satisfactory
`upon storage of the dosage form for up to 14 weeks at 5°
`and 30°C; however, longer term testing indicated that the drug
`exhibits instability at temperatures greater than 5°. In particular,
`storage of the lyophilized formulation for 1 year at 30°C results
`in the formation of a late eluting (by reverse phase HPLC)
`degradation product that had not been previously observed in
`stability studies using the drug as a neat solid or in buffered
`aqueous solution. In this report, we describe the isolation and
`identification of this degradate, and propose a mechanism for
`its formation which involves the pharmaceutical excipient (or
`more precisely, an impurity present in the excipient) acting as
`an oxidizing agent.
`
`MATERIALS AND METHODS
`
`Preparation of Lyophilized Formulations
`
`L-367,073 was obtained in >98% purity from the Depart-
`ment of Process Research, Merck Research Laboratories. Man-
`nitol, USP was obtained from ICI Specialty Chemicals, and
`conformance to USP specifications was verified by internal
`testing prior to use. Lyophilized formulations were prepared
`using a Usifroid Model SMH 101 lyophilizer. For dosage form
`manufacture, a solution containing the peptide (5 mg/ml) and
`mannitol USP (20 mg/ml) was adjusted to pH 5.0 with sodium
`hydroxide and filled in 2.0 ml aliquots into 3 ml molded glass
`vials. The solutions were frozen at - 38°C for 5 hrs, the chamber
`pressure was adjusted to 30 mTorr, and primary drying was
`conducted at -30°C for 5 hrs and 0°C for 10 hrs. The shelf
`temperature was then raised to 30°C at a rate of 6°C/hr, and
`secondary drying was conducted at 30°C for 10 hrs after which
`time the vials were stoppered under vacuum (<30 mTorr).
`The determination of residual moisture levels in the lyophi-
`lized product was determined by Karl Fischer analysis using
`a Model 447 Coulomatic K-F Titrimeter (Fisher Scientific).
`Samples were taken up in methanol dried over molecular sieves
`for the analysis, and the results were corrected for the water
`content of the sieve—dried methanol as determined in a blank run.
`
`Degradate Isolation
`
`A pure sample of the unknown peptide degradation product
`was isolated from stressed (60°C/4 weeks) samples of the lyoph-
`ilized dosage form using preparative scale HPLC. The prepara-
`tive separation was accomplished on a Waters PrepLC 4000
`instrument using a Vydac Prep 218TP column (250 X 22.5
`mm) and a mobile phase consisting of 0.1% aqueous trifluoro-
`acetic acid and acetonitrile. Using a linear gradient from 10%
`to 50% acetonitrile over 30 minutes at a flow rate of 20 ml/
`min, the intact peptide elutes at 6.5 min and the degradate elutes
`at 27 min. A sample loading of 25 mg (5 ml X 5 mg/ml)
`per injection was employed. The degradate fractions from 10
`injections were combined, most of the acetonitrile and volatile
`trifluoroacetic acid was removed on a rotary evaporator, and
`the remaining solution was lyophilized to provide a white solid
`that was 97% pure by analytical HPLC.
`
`1811
`
`0724-8741/96/1200-1811309.50/0 © 1996 Plenum Publishing Corporation
`
`PFIZER, INC. v. NOVO NORDISK AIS - IPR2020-01252, Ex. 1062, p. 1 of 4
`
` PFIZER, INC. v. NOVO NORDISK A/S - IPR2020-01252, Ex. 1062, p. 1 of 4
`
`

`

`1812
`
`Dubost et al.
`
`L-367 ,073
`
`' Fig. 1. Structure of L—367,073 and the oxidative degradation product.
`
`Degradate Characterization
`
`1H NMR Spectmscopy
`
`1H NMR spectra were acquired on a Varian Unity—500
`NMR spectrometer operating at 499.772 MHz. Samples for
`NMR measurements contained approximately 0.82 mg of the
`degradate dissolved in 0.65 ml DMSO-dé. All spectra were
`acquired at 25°C and referenced internally to the residual proton
`of DMSO-ds at 2.49 ppm. Proton chemical shift assignments
`were made with the combined use of 2D TOCSY and NOESY
`
`experiments (3,4). All two-dimensional data sets were acquired
`in the hypercomplex mode (5) for phase-sensitive presentation.
`The TOCSY experiment was acquired with 1K complex points
`in t2 and 512 points in t1 consisting of 8 transients per increment.
`Spin-locking was performed with a MLEV-17 mixing sequence
`(6) for a duration of 50 ms. Data were zero-filled in t1 to 1K
`points and multiplied by a Gaussian apodization function in both
`dimensions prior to Fourier transformation. Baseline correction
`was performed in F2 by fitting the baseline to a second order
`polynomial. A NOESY spectrum was recorded with a mixing
`time of 500 ms. Solvent suppression was achieved by selective
`saturation of the residual water resonance during the 1.0 second
`recycle delay. Data sets consisted of 1K complex points in t2
`and 512 points in t1 with 64 transients per increment using a
`relaxation delay of 1.0 seconds. Data sets were zero-filled to
`1K points in t1 and multiplied by a shifted Gaussian apodization
`function in both dimensions prior to Fourier transformation.
`lH chemical shift assignments are shown in Table 1.
`
`Mass Spectrum
`
`The fast atom bombardment mass spectrum of the degra-
`date was obtained on a VG—70E magnetic sector mass spectrom-
`eter operating at an accelerating potential of 6 kVolts. The
`analyte (ca.
`1 mg) was dissolved in 50 ul of methanol and
`1—2 11.1 of the solution was mixed with an equal amount of
`dithiothreotol/dithioerythritol (magic bullet) matrix in metha-
`nol. Magnet scans were performed over the mass/charge range
`of 1200 to 140 at a scan rate of 6 second/decade.
`
`‘H NMR Chemical Shifts“ for the Oxidation Product
`Table I.
`-————_——
` Residue” NH Hm H3 Other
`
`
`
`Cys 1
`8.18
`4.62
`2.76, 2.91
`Asn
`8.31
`4.86
`2.79
`Dmt
`—
`3.96
`-—
`
`
`
`7.41, 7.91 (NHZ)
`4.74, 5.15 (H5)
`0.57, 1.30 (B'CHg)
`7.48, 7.79 (Ar—H)
`
`2.92, 3.36
`
`2.57, 2.75
`2.88, 3.04
`
`7.97
`7.52
`8.05
`8.50
`
`4.48
`3.77
`4.61
`4.63
`
`AmF
`G1)!
`Asp
`Cys 2
`1.81
`CH3CO
`9.94
`CHO
`———_—_—
`
`" Expressed in ppm referenced to internal DMSO-ds at 2.49 ppm.
`b Abbreviations: Cys, cysteine; Asn, asparagine; Dmt, dimethylthiopro-
`line; AmF, aminomethyl phenylalanine; Gly, glycine.
`
`UV Spectra
`
`Ultraviolet spectra were recorded on a Perkin Elmer
`Lambda 6 spectrophotometer. All samples were dissolved in
`water at a concentration of ca. 15 ug/ml and spectra were
`recorded from 200 to 400 nm at ambient temperature.
`
`Amino Acid Analysis
`
`Samples of the intact drug or the degradate (ca. 0.5 mg
`each) were hydrolyzed for 20 hrs in 6 M hydrochloric acid at
`110°C, then evaporated to dryness under vacuum. The residues
`were dissolved in commercial citrate buffer preparation (Beck-
`man NaS diluent) and analyzed on a Beckman 6300 High
`Performance Amino Acid Analyzer. Results were calculated
`using response factors determined from authentic samples of
`the amino acids run just prior to the sample assay.
`
`Stability Studies
`
`Accelerated stability studies were conducted at 40° and
`60°C. Solutions containing 5 mg/ml of peptide and either 5,
`10, or 20 mg/ml of mannitol were adjusted to pH 5 with concen—
`
` PFIZER, INC. v. NOVO NORDISK A/S - IPR2020-01252, Ex. 1062, p. 2 of 4
`
`

`

`Excipient-Induced Oxidation in a Lyophilized Peptide Formulation
`
`1813
`
`trated sodium hydroxide and filled and lyophilized as previously
`described. The freeze dried samples were stored in the dark in
`ovens maintained at the stated temperature to :1°C. After 12
`weeks of storage, the contents of the lyophilized samples were
`quantitatively transferred to a 10 mL volumetric flask and
`diluted to the mark with water. Analytical HPLC was performed
`on a Beckman Ultrasphere C18 column (250 X 2.4 mm, 5 pm)
`using a mobile phase of acetonitrile vs 0.01 M phosphate buffer
`(pH 3.5), and a gradient elution profile varying from 10%
`acetonitrile initially to 35% acetonitrile at 20 minutes was
`employed. The flow rate was set at 1.5 ml/min and detection
`was by UV at 260 nm.
`
`Derivatizations with 2,4-Dinitrophenylhydrazine
`
`2,4-Dinitrophenylhydrazine (DNPH) was purchased from
`Aldrich and was recrystallized from absolute ethanol prior to
`use. In a typical derivatization reaction, 4 m1 of a stock DNPH
`solution (0.3 mg/ml in acetonitrile) was combined with 0.3 ml
`of 2 M aqueous hydrochloric acid and 3 ml of a stock solution
`of mannitol (15—30 mg/ml) in water. The mixture was incubated
`at ambient room temperature for 90 min, then brought up to a
`total volume of 10 ml with distilled water and analyzed by
`HPLC. The HPLC analyses used a Beckman Ultrasphere C18
`column, with a mobile phase of 65% 0.01 M hydrochloric acid
`and 35% acetonitrile at a flow rate of 1 ml/min. Detection was
`by UV absorbance at 365 nm.
`Two additional derivatization reactions were conducted.
`
`A blank run utilizing distilled water in place of the mannitol
`solution served as a control. In addition, a solution of marmose
`(Sigma Chemical Co., used as received) was derivatized and
`analyzed as described above. The mannose—DNPH adduct was
`used to estimate the levels of reducing sugar impurities in
`mannitol by assuming equal chromatographic response factors
`at the wavelength used for detection.
`
`RESULTS AND DISCUSSION
`
`Degradate Identification
`
`Storage of the peptide in the lyophilized formulation for
`extended time periods at 30°C or for shorter time periods at
`higher temperatures results in the formation of a late-eluting
`degradation product. Significantly, this degradation product is
`not observed in stability studies with the peptide as a neat solid
`or in aqueous buffered solutions. It therefore seems likely that
`the excipient mannitol which is used in the formulation as a
`bulking agent is inducing the degradation of the active drug.
`The degradation product was isolated by preparative HPLC
`in quantities sufficient for spectroscopic structure determina-
`tion. The UV spectrum of the degradate exhibits an absorption
`maximum at 257 nm; in contrast, the drug has no significant
`absorbtivity above 230 nm. The shift to longer wavelength is
`indicative of additional conjugation in the molecule, and tends
`to rule out many of the commonly observed degradative mecha-
`nisms of peptide molecules such as amide bond cleavage/isom-
`erization, disulfide bond reduction, or ester hydrolysis (7—9).
`These mechanisms are also rendered unlikely by mass spectral
`data which indicate a reduction in molecular mass of 1 for the
`degradation product relative to the intact peptide.
`Amino acid analyses for aspartic acid (Asp), glycine (Gly),
`and p-arninomethyl phenylalanine (AmF) proved to be extremely
`
`useful in deducing a structure for the degradate. After 20 hrhydro-
`lysis in 6 N hydrochloric acid, analysis of the intact drug gave
`Asp-2.06, Gly-1.05, AmF-1.02 mmol/mg (note that the aspara-
`gine residue appears as Asp in the assay). Under the same hydro-
`lysis conditions, the Asp and Gly content of the degradate was
`not significantly different (Asp—1.93, Gly-0.94 mmol/mg), but
`the Amf content was reduced to 0.03 mmol/mg.
`Consistent with the amino acid analysis, the 1H-NMR data
`(Table 1) shows the dissappearance of the benzylic protons on
`the AmF residue. Of particular note in the spectrum is the
`appearance of a new resonance at 9.94 ppm which integrates
`for one proton. This region of the spectrum is generally diagnos-
`tic for acidic or aldehydic protons (10). The only structure that
`is consistent with the UV, mass spectral, and NMR results is
`the benzaldehyde derivative arising from oxidative deamination
`of the drug, as shown in Figure 1.
`
`Mechanistic Aspects
`
`The identification of the aldehyde degradate raises the
`question-how does the oxidation of the drug occur in the solid
`state in a lyophilized formulation sealed under vacuum? The
`oxidation occurs at a benzylic methylene carbon, and it is
`well established that benzylic hydrogens are susceptible to free
`radical oxidation (11). In such oxidations, however, the oxidiz-
`ing species is molecular oxygen. Although we cannot com-
`pletely rule out the possibility that small amounts of residual
`oxygen were present in the lyophilized samples, a free radical
`mechanism requiring a radical initiator and a stoichiometric
`amount of oxygen does not seem likely. Moreover, the oxidative
`degradate was not observed in stability studies with neat solid
`drug conducted under an air atmosphere. The implication is
`that the excipient mannitol is involved in the oxidation.
`Of particular relevance to the mechanism of oxidation is
`the known oxidation of benzylamine to benzaldehyde in acidic
`aqueous formaldehyde solution (12). This reaction proceeds
`via an irnine intermediate and does not require molecular oxy-
`gen. An analogous reaction involving reducing sugar impurities
`present in commercial mannitol provides a reasonable mecha-
`nism for the formation of an aldehyde from the aminometh—
`ylphenyl moiety of the drug.
`It
`is thus proposed that
`the
`oxidative degradate is formed via (1) Schiff base formation
`between the peptide primary amine and the aldehydic group of
`a reducing sugar impurity, (2) tautomerization to move the
`double bond into a more stable configuration in which it is
`conjugated with the phenyl group, and (3) hydrolytic cleavage
`of the new Schiff base to generate the observed degradation
`product. The sequence is illustrated in Figure 2.
`The obvious requirement for this mechanism to be correct
`is the presence of reducing sugar impurities in the mannitol.
`The mannitol used in the lyophilized formulations conformed to
`pharmacopeal requirements; however, the USP test for reducing
`sugars gives only qualitative information (13). Therefore, the
`reducing sugar impurity level was determined chromatographi-
`cally after conversion to the phenylhydrazone derivatives (14).
`The results show that phenylhydrazine reactive impurities
`(assumed to be sugar aldehyde groups) are present at approxi—
`mately 0.1% w/w in the mannitol sample. This is a sufficient
`quantity to react with 2% of the amount of peptide present in
`the lyophilized formulation based on the relative amounts of
`peptide and mannitol and the molecular weights of 869 for the
`
` PFIZER, INC. v. NOVO NORDISK A/S - IPR2020-01252, Ex. 1062, p. 3 of 4
`
`

`

`1814
`
`Dubost et a1.
`
`previous studies have generally implicated free radical precur-
`sors as the initiators of oxidation. In contrast, the mechanism
`described herein does not rely on the involvement of radical
`or radical-precursor species, but does depend on the intermedi-
`acy of a Schiff base. In this respect, our mechanism is similar
`to that recently described by Stella et a1. (16) in which the
`degradation of benzylguanine in aqueous polyethylene glycol
`solution was ascribed to the presence of formaldehyde in the
`organic solvent.
`In summary, degradation of a cyclic peptide drug in a
`lyophilized formulation occurs to give an aldehyde as an oxida—
`tion product. The oxidative pathway does not require oxygen,
`but does depend on the presence of impurities in the pharmaceu—
`tical excipient used in the formulation. A mechanism involving
`Schiff base formation, double bond isomerization, and subse-
`quent hydrolysis has been proposed to account for the formation
`of the degradate. It is remarkable that this series of reactions
`occurs within a lyophilized solid, and further studies will hope-
`fully reveal additional details of the various reaction steps. Since
`mannitol finds widespread use in lyophilized pharmaceutical
`dosage forms, we mention that other compounds containing an
`arylmethylamine moiety may react in a fashion analogous to
`that described here.
`
`ACKNOWLEDGMENTS
`
`The authors would like to thank Mr. Robert Kenney for
`his assistance with the derivatization procedure and Ms. Mei-
`Jy Tang for performing the amino acid analyses. We also
`acknowledge helpful discussions with Drs. Ruth Nutt, Stephen
`Brady, and Daniel Veber.
`
`REFERENCES
`
`5"?
`
`9°.“
`
`1. D. R. Ramjit, J. J. Lynch, G. R. Sitko, M. J. Mellot, M. A.
`Holahan, I. I. Stabilito, M. T. Stranieri, G. Zhang, R. J . Lynch,
`P. D. Manno, C. T. C. Chang, R. F. Nutt, S. F. Brady, D. F. Weber,
`P. S. Anderson, R. J. Shebuski, P. A. Friedman, and R. J. Gould.
`J. Pharmacol. Exper. Therap. 266:1501—1511 (1993).
`2. G. D. Hartman, M. S. Egbertson, W. Halczenko, W. L. Laswell,
`M. E. Duggan, R. L. Smith, A. M. Naylor, P. D. Manno, R. J.
`Lynch, G. Zhang, T. C. Chang, and R. J. Gould. J. Med. Chem
`35:4640—4642 (1992).
`3. L. Braunschweiler and R. R. Ernst. J. Magn. Reson. 53:521-
`525 (1983).
`A. Bax and D. G. Davis. J. Magn. Reson. 652355—360 (1985).
`D. J. States, R. A. Haberkom, and D. J. Ruben, J. Magn. Reson.
`48:286—292 (1982).
`6. M. H. Levitt, R. Freeman, and T. Frenkiel. J. Magn. Reson.
`47:328—330 (1982).
`T. Geiger and S. Clarke. J. Biol. Chem. 262:785—794 (1987).
`Y.-C. J. Wang and M. A. Hanson. J. Parenteral Sci. Tech. 42:53—
`526 (1988).
`9. M. C. Manning, K. Patel, and R. T. Borchardt. Pharm. Res.
`6:903—918 (1989).
`10. D. H. Williams and I. Fleming. Spectroscopic Methods in Organic
`Chemistry, McGraw Hill, London, 1973.
`J. A. Howard. Adv. Free Radical Chem. 4:49—174 (1972).
`J. Graymore and D. D. Davies.
`J. Chem. Soc. 293—294
`(1945).
`13. Official Monographs/Mannitol. The United States Phan'nacopeia.
`23:929 (1995).
`14. A. I. Vogel. A Textbook of Practical Organic Chemistry, Long-
`mans, Green and Co., London, 1957.
`J. W. McGinity, J. A. Hill, A. L. LaVia. J. Pharm. Sci. 64:356—
`357 (1975).
`16. D. S. Bindra, T. D. Williams, and V. J. Stella. Pharm. Res.
`11:1060—1064 (1994).
`
`11.
`12.
`
`15.
`
`o
`>711}!
`114'
`
`BIN—g
`0
`
`o '
`
`>—m-1
`2L,
`
`PIN—é
`
`°
`
`+
`
`314
`a
`
`- H o
`”2—.
`
`NH2
`
`2(2an
`
`l
`NH m‘?
`o
`
`o
`
`1’11.
`
`3
`
`o
`
`L-367,073
`>»NH FIN—g
`"1.,1
`
`o +
`
`R1‘\
`
`m2
`
`320
`
`\
`
`N——cnzi=.1
`o
`n
`Fig. 2. Proposed route of formation of the oxidative degradation
`product.
`
`peptide and 180 for a monosaccharide. Since we have not
`observed the degradate to form to an extent greater than this
`on a weight basis, we conclude that the mechanism cannot be
`ruled out based on mass balance.
`
`Further support for the postulated mechanism is provided
`by the results of accelerated stability studies which were con-
`ducted during dosage form optimization. In these studies, the
`drug content of the formulation was held constant (10 mg per
`vial) and the amount of mannitol was varied between 10 and
`40 mg/vial. Stability data were obtained after 12 weeks storage
`in constant temperature ovens at 40° and 60°C, and these results
`are shown in Table 2. The data show that the extent of oxidative
`
`degradation increases with increasing mannitol content at both
`temperatures studied. Moisture analyses on the freshly prepared
`lyophilized samples did not reveal any significant differences
`in water content (range 0.9 to 1.2% water), thereby eliminating
`varying water content as an explanation for the trends in the
`stability data.
`The role of excipient impurities in promoting drug oxida-
`tion reactions has been described in the literature (15); however,
`
`Table 11. Stability Data for Lyophilized L-367,073“ (Expressed as
`Weight Percent of Oxidative Degradate) as a Function of Mannitol
`Level After 12 Weeks Storage
`
`Wt % Degradate
`
`60°C
`40°C
`mg Mannitol/vial
`0.36
`0. 13
`10
`0.86
`0.37
`20
`
`
`0.5840 1.30
`
`‘1 Each vial contained 10 mg of peptide and was adjusted to pH 5 prior
`to lyophilization.
`
` PFIZER, INC. v. NOVO NORDISK A/S - IPR2020-01252, Ex. 1062, p. 4 of 4
`
`