Jurnal Succinimide

A Mechanism-Based Kinetic Analysis of Succinimide-MediatedDeamidation, Racemization, and Covalent Adduct Formation in aModel Peptide in Amorphous Lyophiles

MICHAEL P. DEHART, BRADLEY D. ANDERSON

Department of Pharmaceutical Sciences, College of Pharmacy, University of Kentucky, Lexington, Kentucky 40536-0082

Received 8 August 2011; revised 1 December 2011; accepted 4 January 2012

Published online 23 January 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23061

ABSTRACT: The succinimide intermediate generated during deamidation of asparagine-containing peptides and proteins has been implicated as having a role in the formation ofmultiple types of degradants in addition to hydrolysis products, including racemization prod-ucts and, more recently, amide-linked, nonreducible protein and peptide aggregates. The for-mation of alternative degradants may be particularly important in solid-state formulations.This study quantitatively examines the role of the succinimide intermediate in hydrolysis,racemization, and covalent, amide-linked adduct formation in amorphous lyophiles. The degra-dation of a model peptide, Gly–Phe–L-Asn–Gly, and its L- or D-succinimide intermediates wereexamined in lyophiles containing hydroxypropyl methylcellulose and varying amounts of ex-cess Gly–Val. Disappearance of the starting reactants and formation of up to 10 degradantswere monitored when lyophiles were exposed to either 27◦C/40% relative humidity (RH) or40◦C/75 RH using a stability indicating high-performance liquid chromatography method. Ter-minal degradant profiles were the same when the starting reactant was either Gly–Phe–L-Asn–Gly or its succinimide intermediate. Nucleophilic attack occurred preferentially at the"-carbonyl of the succinimide intermediate at ratios of approximately 2:1 for both water andthe N-terminus of Gly–Val as the attacking nucleophiles. A mechanism-based kinetic modelanalysis indicates that hydrolysis, racemization, and covalent, amide-linked adduct forma-tion all proceed via the succinimide intermediate. © 2012 Wiley Periodicals, Inc. and theAmerican Pharmacists Association J Pharm Sci 101:3096–3109, 2012Keywords: nonreducible aggregates; amorphous solids; chemical stability; deamidation;lyophilization; solid-state kinetics; peptide stability; protein aggregation; solid state stability

INTRODUCTION

Proteins and peptides are often formulated as amor-phous lyophiles to maximize stability compared withaqueous solutions and to provide solids that canbe readily reconstituted prior to administration. Al-though processes involved in physical and chemicaldegradation may be significantly slower in the amor-phous solid state, they are not completely arrested.The development of reliable quantitative methods topredict long-term stability of drugs in lyophilized andother amorphous solid formulations from kinetic datagenerated over short periods of time under acceler-ated conditions continues to be an active area of in-vestigation.

Correspondence to: Bradley D. Anderson (Telephone: +859-218-6536; Fax: +859-257-2489; E-mail: [email protected])Journal of Pharmaceutical Sciences, Vol. 101, 3096–3109 (2012)© 2012 Wiley Periodicals, Inc. and the American Pharmacists Association

For reactions in solution, mechanism-based kineticmodels and the rate equations derived therefrom areoften used to describe the influence of various fac-tors on reactivity, including drug concentration, pH,temperature, and formulation components. In con-trast, publications in the pharmaceutical literaturethat attempt to address reaction kinetics in amor-phous systems quantitatively typically assume thatmolecular mobility is an overriding factor, such thatreactivity will be coupled to one or more indica-tors of structural relaxation such as the glass tran-sition temperature (Tg) or relaxation times gener-ated from methods such as dielectric analysis1 or nu-clear magnetic resonance.2 For example, Sun et al.3

found that Williams et al.4 empirical equation re-lating relaxation processes in amorphous polymersto temperature4 was useful in describing deviationsfrom the Arrhenius equation observed in the in-activation of glucose-6-phosphate dehydrogenase at

3096 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 9, SEPTEMBER 2012

SUCCINIMIDE-MEDIATED PEPTIDE REACTIONS IN AMORPHOUS LYOPHILES 3097

temperatures below or above Tg.3 Yoshioka et al.5 ob-served that the rates of the Maillard reaction and acyltransfer processes in lyophilized formulations con-taining various polymeric excipients increase with adecrease in Tg of the formulations. Below Tg, the vari-ation in reaction rates with temperature was found tocorrelate with the temperature dependence of struc-tural relaxation times calculated using the Adam–Gibbs–Vogel equation. Several groups have foundthat the empirical Kohlrausch–Williams–Watts equa-tion used to describe heterogeneous relaxation inamorphous solids can be applied to chemical degrada-tion in amorphous solids as well as physical processessuch as protein aggregation.6–9

Although reactant and matrix mobility are clearlyimportant considerations, the degree to which chem-ical degradation may be coupled to structural re-laxation is likely to depend on the nature of therate-determining step.10 Such observations serve asa reminder that despite the increased importance ofmolecular mobility in solid-state formulations, theunderlying reaction chemistry and mechanistic de-tails are still critical in determining both the rates ofdegradation and the degradants formed. Most chem-ical degradation pathways of pharmaceutical rele-vance involve the generation of one or more re-active intermediates. Consider, for example, someof the important pathways for peptide and proteindegradation. Peptide and protein deamidation occursthrough a reactive succinimide intermediate.11 Cova-lent protein aggregate formation may involve amidecross-linking mediated by succinimide12 or anhydrideintermediates,13–15 thiol–disulfide rearrangement in-volving free thiol intermediates,16 or lysinoalaninecross-linking via dehydroalanine residues generatedduring $-elimination of cysteine.17 The Maillard reac-tion involves a Schiff base intermediate,18 ultimatelyleading to a variety of degradation products.

Although the level of mechanistic understanding ofthese reactions currently available originates largelyfrom studies in aqueous solution, the same reac-tion pathways may be operative in the amorphoussolid state. Thus, a firm understanding of the mecha-nism of formation and the ultimate fate of reactiveintermediates may prove useful in predicting andexplaining differences in reaction kinetics in amor-phous solids when compared with solutions. This hasbeen a pursuit in several recent studies publishedfrom the authors’ laboratories. For example, Luo andAnderson19 demonstrated that the amino acid cys-teine forms a reactive sulfenic acid intermediate inthe presence of hydrogen peroxide in aqueous so-lutions resulting in the ultimate formation of thedisulfide cystine when the intermediate reacts withanother molecule of cysteine.20 In amorphous solidformulations containing hydrogen peroxide, addi-

tional degradants were observed that could be tracedto the same reactive intermediate. Their preferen-tial formation in an amorphous polymer glass wasattributed to competing reactions of the sulfenic acidintermediate with additional molecules of hydrogenperoxide, which, due to its small size, has a mo-bility advantage in amorphous glasses.19 A detailedmechanism-based kinetic model was successfully em-ployed to account for both the decline in reactant con-centrations and formation of products as a function oftime in various amorphous formulations.19 Strickleyand Anderson14 demonstrated that covalent, amide-linked aggregates of insulin formed in certain amor-phous lyophiles via nucleophilic attack of a secondinsulin molecule on a reactive cyclic anhydride inter-mediate. Despite the fact that covalent aggregate for-mation was in competition with hydrolysis, increasingwater content increasingly favored covalent aggrega-tion, which was attributed to the plasticizing effectsof water.15

It is now well established that deamidation of as-paragine residues in peptides and proteins occurs byway of a cyclic imide intermediate, both in aqueoussolutions11,12 and in amorphous lyophiles,21,22 result-ing in the formation of aspartate- and isoaspartate-containing peptides. In both amorphous solids andsolutions, deamidation favors formation of isoaspar-tate over aspartate in approximately a 3:1 ratio.11,22

However, this ratio is dependent on several factorsincluding primary sequence,23 higher order proteinstructure,24 solvent choice,25 and apparent pH inamorphous solids.22 The results of experiments eval-uating the effects of various factors on asparaginedeamidation in model peptides have been summa-rized in several reviews26,27 and books.28,29

Lately, there has been an increase in publicationscorrelating succinimide formation and protein aggre-gation. Severs and Froland30 observed succinimideintermediates and dimerization of a pituitary adeny-late cyclase-activating peptide in anhydrous dimethylsulfoxide (DMSO).30 The authors hypothesized thatthe dimerization occurs via nucleophilic attack of afree amine (e.g., N-terminus, lysine, and arginine)on the succinimide carbonyl groups, resulting in amyriad of dimers. Using a more quantitative ap-proach, Desfougeres et al.31 prepared hen egg-whitelysozyme with various amounts of succinimide inter-mediates. They demonstrated a linear dependence be-tween dimer formation and the percent of succinimideintermediate present.31 Similar to the experiments inDMSO, the authors proposed that a free amine reactswith a carbonyl of a succinimide intermediate on anadjacent protein.

Quite recently, we conducted kinetic studies todescribe the decomposition of a model asparagine-containing peptide in amorphous lyophiles containing

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 9, SEPTEMBER 2012

3098 DEHART AND ANDERSON

an excess of a second peptide (Gly–Val).32 A total of10 degradation products were monitored, includinghydrolysis and racemization products as well as fourcovalent adducts resulting from the attack of Gly–Valon the D- and L-succinimide intermediates. The samedegradant profiles were found when the succinimidewas the starting reactant in lyophile formulations,implicating the cyclic imide as the intermediate forformation of all degradants.

In this manuscript, a comprehensive kineticmodel has been developed to describe the kineticsof degradant formation in two different Gly–Phe–L-Asn–Gly lyophile formulations to explore the generalapplicability of a mechanism-based model for ra-tionalizing both the degradants formed as well astheir relative amounts as a function of time. TheGly–Phe–L-Asn–Gly formulations used herein werethe same as those employed in a recent publication,although the storage conditions were different. Bothformulations were previously shown to be in theirglassy state after lyophilization.32 This analysishighlights the potential utility of applyingmechanism-based models to reactions in amor-phous glass formulations and, more specifically, thepotential importance of alternative pathways forbreakdown of succinimide intermediates formed insolid-state peptide and protein drug formulations,which can lead to a variety of covalent adducts inaddition to the normally encountered products ofimide hydrolysis.

MATERIALS AND METHODS

Reagents

The peptide Gly–Phe–L-Asn–Gly and its hydroly-sis products were synthesized by GenScript (Pis-cataway, New Jersey) as trifluoroacetate salts withpurities greater than 95% as determined by high-performance liquid chromatography (HPLC). Thesecompounds were further characterized as previouslydescribed.32 The synthesis of the succinimide in-termediates (Gly–Phe–D-Asu–Gly and Gly–Phe–L-Asu–Gly) and the characterization of covalent adductstandards are described elsewhere.32 Hydroxypropylmethylcellulose (HPMC, Methocel E5) was receivedas a free sample from Dow Chemical (Midland, Michi-gan) and used as received. Gly–Val was purchasedfrom Bachem (Torrance, California) as the zwitte-rion (purity 99.0% by thin layer chromatography).HPLC grade acetonitrile was purchased from FisherScientific (Springfield, New Jersey). Sodium bicar-bonate (ACS grade) was purchased from EM Science(Gibbstown, New Jersey) and succinic acid was pur-chased from Aldrich (St. Louis, Missouri). Deionizedwater was used throughout the experiments.

Preparation of Lyophile Formulations

Gly–Phe–L-Asn–Gly Formulation A

Two types of lyophile formulations were prepared asdescribed previously.32 Formulation A lyophiles wereprepared from a solution containing the trifluoroac-etate salt of Gly–Phe–L-Asn–Gly (1 mM) and a 7.5-fold excess of Gly–Val (7.5 mM) in an aqueous ex-cipient mixture consisting of 0.5% (w/v) HPMC and0.39% (w/v) sodium bicarbonate. The formulation wasadjusted to pH 9.5 with dilute sodium hydroxideprior to lyophilization. Aliquots (100:L) of the solu-tion were quickly transferred into HPLC autosamplervials, which were frozen by placing them on prechilled(−40◦C) shelves in a tray freeze dryer (Virtis Advan-tage, Stoneridge, New York) for approximately 5 min.Primary drying was carried out at −40◦C for 1420 minafter which the shelf temperature was increased at arate of 0.2◦C/min. Secondary drying was performedat 40◦C for 120 min. At each step during the freeze-drying process, a constant pressure of 100 mTorr wasmaintained.

Gly–Phe–L-Asn–Gly Formulation B1

A separate formulation contained the same concen-tration of Gly–Phe–L-Asn–Gly (1 mM) but with agreater molar excess of Gly–Val (30-fold) in 0.5%HPMC. Sodium bicarbonate was omitted from thisformulation in order to compensate for the increasedamount of Gly–Val and thereby maintain the sameconcentration of Gly–Phe–L-Asn–Gly in the solid for-mulation after lyophilization as that in formulationA. Again, pH was adjusted to 9.5 with dilute sodiumhydroxide.

Succinimide Formulations

Lyophiles were also prepared from 0.1 mM solutionsof either Gly–Phe–L-Asu–Gly (formulation B2) orGly–Phe–D-Asu–Gly (formulation B3) in 0.5% HPMCand 30 mM Gly–Val to approximate the compositionof the Gly–Phe–L-Asn–Gly formulation B1. The con-centrations of succinimide were reduced in formula-tions B2 and B3 by 10-fold relative to the Gly–Phe–L-Asn–Gly concentrations in formulation B1 in an at-tempt to partially account for the fact that cyclicimide concentrations in formulations of Gly–Phe–L-Asn–Gly remain at a low-steady state relative toGly–Phe–L-Asn–Gly. All three formulations (i.e., B1,B2, and B3) contained approximately a 300-fold molarexcess of Gly-Val relative to the succinimides.

Lyophile Characterization

Water Content

After lyophilization, cakes were placed in a vacuumoven and dried at 50◦C under a maximum vacuumto achieve a constant weight. A portion of the cake

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 9, SEPTEMBER 2012 DOI 10.1002/jps


was analyzed by suspending the cake in anhydrousDMSO. The water content was measured by injectingthe suspension into the cell of a Karl Fisher appa-ratus. Carbonate buffers contribute a molar amountof water to the measured water content in the KarlFisher analysis, which was subtracted from the mea-sured water content to obtain initial water contentsfor each formulation. The dried cakes were thenplaced at various relative humidities and monitoredfor weight change. The change in mass after equili-bration was attributed to the absorption of water.

Reconstituted Solution pH

The pH value for lyophiles reconstituted with 100:Lof water was determined at the beginning and endof the kinetic experiments. The average of the twomeasurements is reported. The measurements wereobtained on a Beckman pHI 40 pH Meter (Brea, Cal-ifornia) with a MI-40 combination micro-pH probe(Microelectrodes, Inc., Bedford, New Hampshire).

Polarized Light Microscopy

Samples of each formulation were analyzed by po-larized light microscopy at the beginning and end ofthe kinetic studies. Cakes were suspended in siliconeoil, placed on a glass slide, and then covered witha glass coverslip. Samples were then examined forcrystallinity using a polarizing microscope (OlympusBX51, Center Valley, PA) equipped with a 522 nmfilter.

Kinetic Studies

Lyophiles containing Gly–Phe–L-Asn–Gly in formu-lation A (7.5-fold molar excess of Gly–Val) werepreequilibrated at 75% relative humidity (RH; oversaturated sodium chloride solution) and ambient tem-perature for 15 min then placed in a dessicator at thesame RH and 40◦C. A lower temperature and RH werenecessary to monitor the reactivities of formulationsB1, B2, and B3 to ensure that the breakdown of thesuccinimide intermediates would be slow enough toallow the collection of multiple samples from whichcomplete concentration versus time profiles could begenerated. Therefore, lyophiles of formulations B1,B2, and B3 (∼30-fold molar excess of Gly–Val) wereplaced directly in a dessicator at 27◦C containing asaturated solution of potassium carbonate to providean RH of 40%.

At specific time points, samples were removed fromthe chamber and reconstituted with the original fillvolume of water. Hydrochloric acid (1 N) was used toadjust the pH of the reconstituted samples to quenchthe reactions and provide a similar pH value as theHPLC mobile phase. Samples were either analyzedimmediately by HPLC or frozen until the time ofanalysis.

For kinetic analyses, only data up to 25% dis-appearance of Gly–Phe–L-Asn–Gly starting peptidewere utilized, whereas the succinimide intermediateswere allowed to degrade almost to completion. Indi-vidual peak areas for all hydrolysis products, L- orD-succinimides, and covalent adducts were monitoredand converted to concentrations using response fac-tors as obtained in a previous publication.32

HPLC Analyses

The identification, quantitation, and separation ofGly–Phe–L-Asn–Gly and 10 of its degradants are de-scribed elsewhere.32 Briefly, separation was achievedusing a Waters Alliance LC system and a SupelcosilABZ+ column (15 cm × 3.5 mm, 5-:m pore size).To separate Gly–Phe–L-Asn–Gly, the hydrolysis prod-ucts, and the succinimide intermediates, the mobilephase consisted of succinic acid (20 mM) in pure wa-ter with the pH adjusted to 4.1 with dilute sodiumhydroxide. After 16 min, a step gradient to 5% ace-tonitrile and 95% buffer was necessary to elute thecovalent adducts. A 5 min wash step (25:75 buffer–acetonitrile) was implemented at the end of each runto remove HPMC that was retained on the column, fol-lowed by re-equilibration for 15 min with 100% buffer.

Mechanism-Based Kinetic Model Development

A comprehensive reaction scheme based on the knownmechanism of deamidation of Asn-containing pep-tides coupled with the assumption that racemizationand covalent, amide-linked adduct formation also pro-ceed through a succinimide intermediate is shown inFigure 1. Differential equations based on the variouspathways depicted in Figure 1 were derived to simul-taneously fit the concentration versus time profilesof each analyte (Eqs. 1–11). To simplify the differen-tial equations, the primary sequences of the individ-ual compounds have been removed and replaced withRoman numerals and subscripts to denote stereo-chemistry as listed in Figure 1.

The reaction scheme shown does not include pos-sible additional, potential decomposition productsformed, for example, by reactions between the succin-imide intermediate and a second Gly–Phe–L-Asn–Glymolecule or with excipient (i.e., HPMC) molecules, orsecondary degradation of the degradants describedin Figure 1 to produce other covalent adducts or hy-drolysis products. In cases where mass balance wasnot achieved (e.g., in formulations containing succin-imide as the initial reactant), an additional pathwayleading from either succinimide to unknown productswas included as designated by a rate constant kx. In-cluded in kx were rate constants for any pathways thatwere not explicitly included in fitting the data becauseof the lack of ability to detect or quantify degradantconcentrations.



Figure 1. Mechanism-based model for the degradation of Gly–Phe–L-Asn–Gly [I] to formGly–Phe–L- Asu-Gly [IIL], which then racemizes to Gly–Phe–D-Asu–Gly [IID]. Both succinimideintermediates can then react with water to form aspartates [III] and [IV], with Gly–Val to formamide-linked adducts [V] and [VI], or possibly with other excipients in the formulation.

Simultaneous fits of all concentration versus timeprofiles were performed by nonlinear least squares re-gression analysis using commercially available com-puter software (Scientist, Micromath Scientific Soft-ware, St. Louis, Missouri). Parameters for the initialconcentrations of each degradant were initially in-cluded in the parameter set but those for which the95% confidence limits for the parameter estimate in-cluded zero were eliminated and the initial concen-trations in those cases were fixed at zero.

∂[I]∂T

= −k1 × [I] (1)

∂[IIL]∂T

= k1 × [I] − (k2 + k4 + k6 + k8 + k10 + kx)

×[IIL] + k11 × [IID] (2)

∂[IID]∂T

= −(k3 + k5 + k7 + k9 + k11 + kx)

×[IID] + k10 × [IIL] (3)

∂[IIIL]∂T

= k2 × [IIL] (4)

∂[IIID]∂T

= k3 × [IID] (5)

∂[IVL]∂T

= k4 × [IIL] (6)

∂[IVD]∂T

= k5 × [IID] (7)



∂[VL]∂T

= k6 × [IIL] (8)

∂[VD]∂T

= k7 × [IID] (9)

∂[VIL]∂T

= k8 × [IIL] (10)

∂[VID]∂T

= k9 × [IID] (11)

RESULTS

Lyophile Characterization

After lyophilization, the cakes obtained were white incolor and occupied the same volume as the original fillheight. After storage at 75% RH and 40◦C, lyophiles offormulation A exhibited partial collapse as indicatedby some retraction of the cakes from the vial walls.However, no evidence of crystallization was observedas indicated by the lack of birefringence under a po-larizing microscope. Upon exposure to storage condi-tions of 27◦C and 40% RH, lyophiles of formulationB underwent partial collapse to approximately halfof their original height. Evidence of partial recrys-tallization of Gly–Val was previously reported whenlyophiles stored at 40◦C and 40% RH were examinedby polarized light microscopy (data not shown). Be-cause of the overwhelming excess of Gly–Val in rela-tion to succinimide in these formulations, the impactof partial recrystallization of Gly–Val on the reactionkinetics of interest in this study was considered to beinconsequential.

The average pH of samples reconstituted with theoriginal volume of water was 9.58 ± 0.02 for formula-tion A and 9.36 ± 0.06 for formulation B. The watercontent for formulation A lyophiles was 2.4 ± 0.6%and increased to 13.6 ± 0.8% when exposed to 40◦Cand 75% RH. The water content of formulation Blyophiles was 3.5 ± 0.7% after lyophilization and4.53 ± 0.99% after storage at 40% RH and 27◦C.

Degradant Profiles from Gly–Phe–L-Asn–Gly inFormulation A

A total of 10 degradants were observed over timewhen lyophiles containing Gly–Phe–L-Asn–Gly informulation A were incubated at 40◦C and 75%RH (Fig. 2). These degradants were identifiedpreviously32 as four covalent, amide-linked adductsof Gly–Val, four hydrolysis products (D- and L-aspartates and isoaspartates), and the D- and L-succinimide intermediates. The four covalent adductand hydrolysis product peaks result from nucleophilicattack of either Gly–Val or water at the "- or $-carbonyl of either the D- or L-succinimide interme-

Figure 2. HPLC chromatogram showing the formationof four hydrolysis products (IIIL, IIID, IVL, and IVD), twosuccinimide intermediates (IIL and IID) and four cova-lent, amide-linked adducts (VL, VD, VIL, and VID) whenlyophiles of I in formulation A were stored at 75% RH and40◦C for 2 h. Exact structures of each peak are depicted inFigure 1. Changes in the baseline due to the step gradientwere removed and baselines adjusted.

diate. In this formulation, which contained only a7.5-fold excess of Gly–Val in relation to Gly–Phe–L-Asn–Gly and had a relatively high water content(13.6%) as a result of its storage at 75% RH, hydroly-sis products were the dominant degradants with theminor degradants being covalent Gly–Val adducts.

Kinetic Analysis of the Degradation ofGly–Phe–L-Asn–Gly and Product Formationin Formulation A

Previously, a simplified kinetic analysis was per-formed for the degradation of Gly–Phe–L-Asn–Glyin formulation A at 40◦C and at lower humidity(40% RH) by pooling the concentrations of hydroly-sis products and Gly–Val adducts to generate a singlerate constant for hydrolysis and another for covalentadduct formation.32 Data generated herein from thedegradation of Gly–Phe–L-Asn–Gly in formulation Aat 40◦C and 75% RH were used to evaluate a morecomprehensive kinetic model that included a rate con-stant for the formation of each degradant. This dataset was selected because all 10 degradants formedin appreciable amounts and therefore reliable rateconstants could be estimated. The model describedin Figure 1 assumes that the L-succinimide under-goes racemization to the D-succinimide. The L- andD-succinimides are then susceptible to nucleophilicattack at the "- and $-carbonyls by either water orGly–Val to yield hydrolysis products or covalent,amide-linked adducts, respectively. The nonlinearsimultaneous fits of Eqs. 1–11 to the data are seenin Figure 3 and a summary of the rate constants andstatistics is listed in Table 1. The initial concentra-tions for three of the degradants (i.e., the normal as-partates, IVL and IVD, and the L-succinimide IIL) dif-fered significantly from zero and parameters for theseimpurities were therefore also included in the model.Although they are not included in Table 1, their valuescan be estimated from Figures 3b and 3d. Clearly, the



kinetic model satisfactorily accounts for the simulta-neous disappearance of Gly–Phe–L-Asn–Gly and theformation of all 10 degradants as a function of time inthis formulation, at least during the period in whichapproximately 25% of the starting compound has de-graded.

The concentrations of both succinimides versustime are consistent with their role as reactive inter-mediates in that the concentrations of the L- and D-succinimide rise to maxima of approximately 3%–4%and approximately 1% of the Gly–Phe–L-Asn–Glyconcentration, respectively, followed by gradual de-clines (Fig. 3d). Individual rate constants for theconversion of the L- to the D-succinimide and viceversa (k10 and k11) were set equal to each other.More reliable estimates of the rate constants for in-terconversion of Gly–Phe–L-Asu–Gly and Gly–Phe–D-Asu–Gly were determined in formulations B2 andB3 when the individual succinimides were used asstarting reactants (Table 2) and these individual de-

Table 1. Summary of the Calculated Rate Constants WhenEqs. 1–11 Were Used to Fit the Data for Lyophiles ofGly–Phe–L-Asn–Gly in Formulation A Stored at 75% RH and40◦C

Parameter Reaction Step Calculated Rate Constant (h−1)

k1 I→IIL 0.11 (0.08–0.14)k2 IIL→IIIL 1.5 (1.0–1.9)k3 IID→IIID 2.7 (1.8–3.5)k4 IIL→IVL 0.7 (0.0–1.4)k5 IID→IVD 1.0 (0.1–2.0)k6 IIL→VL 0.024 (0.016–0.031)k7 IID→VD 0.018 (0.011–0.025)k8 IIL→VIL 0.024 (0.016–0.032)k9 IID→VID 0.045 (0.03–0.06)k10,k11 IIL→IID, IID→IIL 1.8 (1.1–2.5)

For this analysis, the rate constants k10 and k11 representingL-/D-succinimide interconversion were assumed to be equal to each other.Numbers in parentheses represent 95% confidence intervals.

terminations support the approximation in Table 1that k10 = k11. A kinetic model that included the

Figure 3. Results from the simultaneous fitting of concentration versus time data obtainedfrom Gly-Phe-L-Asn-Gly degradation in formulation A lyophiles at 40◦C and 75% RH. (a) Degra-dation of I ( *); (b) formation of IIIL (�), IIID (�), IVL (�), and IVD (�); (c) formation of VL (�),VD (�), VIL (�), VID (); d) IIL (×), and IID (+). Lines represent nonlinear regression analysesof concentration versus time profiles based on Eqs. 1–11.



possible formation of Gly–Phe–D-Asu–Gly directlyfrom Gly–Phe–L-Asn–Gly was evaluated but showedno improvement in the fits and negatively impactedthe statistics.

Apparent lag times are evident in the formationof both hydrolysis products and covalent adducts(Figs. 3b and 3c), consistent with their formation froma reactive intermediate. As seen in Figure 3b, normalaspartyl-containing peptides were present as impu-rities at tzero but the concentrations of isoaspartyl-containing peptides showed a rapid increase in con-centration over time and eventually surpassed theaspartyl-containing peptides in concentration. Isoas-partyl peptide formation was favored over aspartylpeptides by a 2.2:1 ratio when generated from theL-succinimide intermediate and by 2.6:1 when gener-ated from the D-succinimide as determined by theirrespective rate constants. The same trend was ob-served for the attack by the N-terminus of the Gly–Val on the D-succinimide where "-carbonyl attack wasfavored by approximately 2:1, although no signifi-cant difference could be shown for the rates of for-mation of isoadducts and normal adducts from theL-succinimide from this data set.

Degradant Profiles of Gly–Phe–L-Asn–Gly,Gly–Phe–D-Asu–Gly, and Gly–Phe–L-Asu–Gly inLyophiles of Formulation B

Comparisons of the degradation of Gly–Phe–L-Asn–Gly, Gly–Phe–D-Asu–Gly, and Gly–Phe–L-Asu–Gly were conducted in formulation B, whichcontained a greater excess of Gly–Val than formula-tion A and under storage conditions (27◦C and 40%RH) that led to a reduced water content in theselyophiles in comparison with formulation A (i.e., 4.5%vs. 13.6%). As a consequence, only trace quantitiesof hydrolysis products totaling less than 5% ofthe initial reactant concentration were observedwhen the succinimides were the starting compounds

and no hydrolysis products were observed whenGly–Phe–L-Asn–Gly was the starting reactant. Asshown in Figure 4, the four major degradants thatwere observed when Gly–Phe–L-Asn–Gly and thesuccinimide intermediates were the starting com-pounds in formulations B1, B2, and B3 were the D-and L-diastereomers, resulting from Gly–Val attackat the "- and $-carbonyls of the D- and L-succinimideintermediates.

The major degradants formed from the L-succinimide were the respective L-adducts of Gly–Valwith relatively smaller amounts of the D-adducts(Fig. 4d). Similarly, incubation of the D-succinimideyielded mostly D-adducts with L-adducts as minordegradants (Fig. 4e). The interconversion betweenthe D- and L-succinimides was relatively rapid asshown in Figure 4c, albeit slower than the over-all conversion of succinimide to covalent adducts.Both covalent adduct diastereomers were observedwhen Gly–Phe–L-Asn–Gly was the starting reactant(Fig. 4b), although the L-adducts dominated.

Simultaneous Kinetic Analysis with Variation in theInitial Reactant in Formulation B

Data generated from the degradation of Gly–Phe–L-Asu–Gly, Gly–Phe–D-Asu–Gly, or Gly–Phe–L-Asn–Gly in amorphous formulation B lyophiles at27◦C and 40% RH were used to quantitatively testthe self-consistency of the mechanism-based modelproposed for covalent adduct formation. As reportedpreviously,32 the degradant profiles were the samewhen either succinimide intermediate (formulationsB2 and B3) or the asparagine-containing peptide (for-mulation B1) were the starting reactants supportinga central role for the succinimide in covalent adductformation. To quantitatively evaluate the kineticsof formation of the four diastereomeric adductsfrom their respective succinimide intermediates,degradant concentration versus time profiles from

Table 2. Comparison of Rate Constants for Various Reaction Steps in Lyophile Formulations B1, B2, and B3 at 40% RH and 27◦C.Gly–Phe–L-Asn–Gly was the Initial Reactant in Formulation B1 while the L- or D- Succinimides were the Initial Reactants inFormulations B2 and B3. The Left-Hand Column Represents Results from Simultaneous Fits of All Formulations with Shared RateConstants. The Right-Hand Column Displays Results from Simultaneous Fits of Only the Succinimide Formulations (B2 & B3).

Calculated Rate Constant (h−1)

Parameter Reaction Step All Formulations (B1, B2, and B3)Succinimide Formulations Only (B2 and

B3)

k1 I→IIL 3.3 × 10−3 (2.3 × 10−3 − 4.2 × 10−3) –k6 IIL→VL 0.095 (0.065–0.13) 0.082 (0.047–0.118)k7 IID→VD 0.10 (0.060–0.14) 0.093 (0.049–0.137)k8 IIL→VIL 0.15 (0.11–0.19) 0.14 (0.084–0.20)k9 IID→VID 0.27 (0.18–0.36) 0.27 (0.17–0.37)k10 IIL→IID 0.21 (0.11–0.31) 0.20 (0.090–0.30)k11 IID→IIL 0.20 (0.12–0.27) 0.15 (0.074–0.23)kx IIL or IID→Unknown degradants 0.21 (0.10–0.32) 0.17 (0.022–0.33)

Numbers in parentheses represent 95% confidence intervals.



Figure 4. Results from the simultaneous fitting of concentration versus time data obtainedfrom formulations B1, B2, and B3 stored at 40% RH and 27◦C. Gly-Phe-L-Asn-Gly (I) in for-mulation B1 was the initial reactant in the upper panels (a and b) while the lower panels(c-e) display data from formulations B2 and B3 containing Gly-Phe-L-Asu-Gly (IIL) or Gly-Phe-D-Asu-Gly (IID) as the initial reactant. Symbols: (a) I (formulation B1), *; (b) adduct(VL, �; VD, ; VIL, �; VID, �) and succinimide (IIL, +; IID, gray line (predicted)) in formu-lation B1; c) IIL (formulation B2) as starting reactant, �; IID (formulation B3) as startingreactant, �; IIL from IID (formulation B3), ; IID from IIL (formulation B2), �; (d) adduct forma-tion in formulation B2 (same symbols as in (b)); (e) adduct formation in formulation B3 (samesymbols as in (b)).

the Gly–Phe–L-Asu–Gly and Gly–Phe–D-Asu–Glykinetic studies were fit simultaneously over the timeframe required for complete disappearance of thestarting compounds. Initially, the model accountingfor only the four covalent Gly–Val adducts did apoor job of fitting the disappearance curves for thestarting succinimides. Evidently, this model did notfully account for all of the decomposition productsforming from the succinimides. To account for thisdiscrepancy, it was necessary to add an additionalrate constant, kx, which incorporated any unknowndegradation pathways along with the four hydrolysisrate constants (k2, k3, k4, and k5 in Fig. 1) that had notbeen included in the initial analysis. Incorporatingkx into the model (which accounted for ∼40% of theoverall degradation) provided acceptable fits of theconcentration versus time profiles for disappearanceof the succinimides as starting reactants along withthe formation of the corresponding succinimidediastereomer (Fig. 4c) and all four covalent adductsfrom Gly–Phe–D-Asu–Gly (Fig. 2d) and Gly–Phe–L-Asu–Gly (Fig. 2e). The calculated values for the rateconstants generated from the two succinimide datasets in formulations B2 and B3 are listed in Table 2(right-hand column).

As a final test to confirm that the succinimide isthe point of racemization and the exclusive path-way for covalent, amide-linked adduct formation fromGly–Phe–L-Asn–Gly in amorphous lyophiles of formu-lation B1, concentration versus time profiles gener-ated in formulations B1, B2, and B3 with Gly–Phe–L-Asn–Gly and the D- and L-succinimides as the start-ing reactants, respectively, were fit simultaneouslyto the mechanism-based model that included the ad-ditional rate constant kx. The results are shown bythe solid lines in the five panels of Figure 4 andthe rate constants are summarized in Table 2. Therate constants for product formation obtained fromfitting only the succinimide-containing formulations(B2 and B3) were not significantly altered when theGly–Phe–L-Asn–Gly data set was included and fitsimultaneously with the succinimide data sets. Thisself-consistency lends further support to the centralrole of the succinimide intermediates in the formationof covalent peptide adducts.

As observed in the results for formulation A, nucle-ophilic attack by the N-terminus of Gly–Val occurredat both the "- and $-carbonyls of the succinimide in-termediates. The average ratios of rate constants for"/$ attack in formulations B1, B2, and B3 were 1.6



for the L-adducts and 2.6 for the D-adducts. Prefer-ential nucleophilic attack at the "-carbonyl is consis-tent with previous literature for hydrolysis in aqueoussolutions11 and amorphous solids,22 and when ammo-nia is the nucleophile in aqueous solutions.12

DISCUSSION

As mentioned at the outset, many of the reaction path-ways most prominent in pharmaceutical systems arerelatively well understood mechanistically, at leastin aqueous solution, as a result of many detailed in-vestigative studies over several decades. Given theimportance of reactive intermediates in a majorityof these pathways, an underlying hypothesis of thepresent work was that this wealth of understandingof the mechanisms of formation and the ultimate fateof reactive intermediates based on solution studiesmay prove useful in predicting or at least rationaliz-ing differences in reaction kinetics in amorphous solidformulations when compared with solutions.

Deamidation of asparagine residues in peptidesand proteins, for example, is well known to occur viaa reactive cyclic imide intermediate, both in aque-ous solutions11,12 and in amorphous lyophiles,21,22

leading to both isoaspartyl and aspartyl hydrolysisproducts accompanied by racemization. Recently, wedemonstrated in aqueous solution that a more di-verse set of heretofore unidentified degradants mayform from Asn-containing peptides and proteins whenother nucleophiles such as amines are present atsufficiently high effective concentrations to competewith water for the succinimide intermediate.12 Morerecently, this finding was extended to reactions ofa model asparagine-containing peptide (Gly–Phe–L-Asn–Gly) in amorphous lyophiles containing an ex-cess of a second peptide (Gly–Val).32 A total of10 degradation products were identified, includinghydrolysis and racemization products as well asfour covalent adducts resulting from the attack ofGly–Val on the D- and L-succinimide intermediates.The same degradant profiles were found when thesuccinimide was the starting reactant in lyophile for-mulations, implicating the cyclic imide as the inter-mediate for formation of all degradants. The rationalefor the incorporation of a second peptide (Gly–Val) inthese studies was to reduce the number of potentialdegradants that might form. Because both Gly–Valand any amide-linked covalent adducts produced inthe reaction of these two peptides would no longerhave an Asn residue, the tendency of the adducts toform higher order aggregates was reduced.

In this manuscript, studies of the same model pep-tides and the D- and L-succinimide intermediateshave been conducted with the aim of developing acomprehensive, mechanism-based model to describethe kinetics of degradant formation in two different

Gly–Phe–L-Asn–Gly lyophile formulations and stor-age conditions that produce dramatic disparities indegradant profiles. This approach allowed us to testthe general applicability of a mechanism-based modelfor rationalizing both the degradants formed as wellas their relative amounts as a function of time whenchanges in formulation and/or storage conditions sig-nificantly alter the relative amounts of degradantsproduced. The analysis highlights the potential util-ity of applying mechanism-based models to reactionsin amorphous glass formulations and also reinforcesthe potential importance of alternative pathways forbreakdown of succinimide intermediates formed insolid-state peptide and protein drug formulations. Al-though the findings in the present study are hope-fully generally applicable in describing reactions thatasparaginyl peptides and proteins may undergo inthe solid state, clearly the primary structure canhave a significant quantitative impact on deamida-tion rates in both solution and solid state. For exam-ple, Li et al.33 reported that the deamidation ratesof AcGQNEG and AcGQNDG in the solid state weresimilar to those in solution. They suggested that thecharges on Glu and Asp residues in these compoundsmay facilitate local hydration in the solid state, a fac-tor that may also have contributed to the reactivitiesof the model tetrapeptides in the present study.

Although the simple kinetic model employed in ourpreceding study in which all hydrolysis products andall covalent adducts were pooled32 was effective indemonstrating the role of the succinimide intermedi-ate in the formation of both hydrolysis products andcovalent adducts, it did not provide kinetic informa-tion on each individual reaction pathway. The authorsare unaware of any previous attempts to apply such acomplex, mechanism-based kinetic model to reactiv-ity in amorphous solid formulations in a quantitativemanner.

The first set of experiments monitored the simul-taneous disappearance of Gly–Phe–L-Asn–Gly, suc-cinimide formation, and conversion to both hydrol-ysis products and covalent adducts in a formulation(formulation A) containing a reduced ratio of excessGly–Val stored at 40◦C and 75% RH. In this for-mulation and at these storage conditions, a total of10 reactive intermediates or final degradants formedin appreciable amounts and could be monitoredquantitatively to allow nearly all of the relevant re-action parameters to be reliably determined.

Recognizing that generating rate constants for theformation and, in some cases, breakdown of 10 re-action products with reliable precision and accuracymay be difficult, we also examined a formulationcontaining a greater excess of Gly–Val (formulationB) stored at 27◦C and 40% RH. Under these con-ditions, the degradant profile shifted to predomi-nantly the succinimide intermediates and covalent



adducts. In this formulation, the initial starting re-actant was either Gly–Phe–L-Asn–Gly or one of thetwo succinimide intermediates, Gly–Phe–L-Asu–Glyor Gly–Phe–D-Asu–Gly. The question to be addressedwith this set of data was whether or not the same ki-netic model could simultaneously fit the complete con-centration versus time profiles generated from eachof the three initial reactants. This exercise would alsotest in a more quantitative fashion the hypothesisthat the formation of covalent adducts proceeds ex-clusively through succinimide intermediates.

Fate of the Succinimide Intermediate in AmorphousLyophiles

Hydrolysis Products

The lyophilized formulations in the present studywere prepared from pH 9.5 solutions. Deamidationof asparaginyl peptides in amorphous lyophiles isknown to proceed through a succinimide intermedi-ate when the effective lyophile pH is in the basicregion,22 and this was also found in the present study.Although the formation of isoaspartyl and aspartylhydrolysis products resulting from the attack ofwater at the succinimide "- or $-carbonyl in the sec-ond step of the reaction would be expected to proceedat a slower rate in amorphous solids where the watercontent is substantially reduced in relation to aque-ous solutions, it is important to bear in mind thatthe rate-determining step for deamidation is the for-mation of the cyclic imide rather than its breakdown.Consequently, consideration of factors affecting nucle-ophilic attack at the succinimide may be more impor-tant for ascertaining the ultimate reaction productsrather than the overall degradation rate.

Succinimide breakdown generally favors isoas-partyl (attack at the "-carbonyl) over aspartyldegradants in a 3:1 to 4:1 ratio in aqueoussolutions,11,34 but the ratio can depend on the typeof solvent,25 the solution environment,35 primary se-quence and, in the case of proteins, higher orderstructure.24 A preference for isoaspartyl over aspartyldegradants has also been reported in lyophiles.31,36

The present results obtained in formulation A, wherehydrolysis products dominated, are consistent withthese previous observations. Hydrolysis products gen-erated from Gly–Phe–L-Asn–Gly favored isoaspartateover aspartate by approximately 2.5:1. Ratios of isoas-partate/aspartate closer to one have been previouslyreported in native proteins with higher order struc-ture where proximal amino acids are thought to influ-ence the stability of transition states involved in suc-cinimide breakdown.24 Primary sequence of a peptidecan also affect the relative amounts of isoaspartyl-and aspartyl-containing peptides. For example, ly-sine located at the N-1 position in small peptides in-creases the relative amounts of aspartyl-containing

peptides. The slightly reduced preference for isoas-partyl over aspartyl hydrolysis products of 2.5:1 ob-served here likely also reflects influences of the localmicroenvironment in the vicinity of the succinimideon the stability of the transition states leading to theisoaspartyl or aspartyl products.

Succinimide Involvement in Covalent, Amide-LinkedAdducts

In solution, the succinimide intermediate generatedduring the deamidation of Asn-containing peptidesmay be susceptible to attack by nucleophiles otherthan water.12 In aqueous solutions, intramolecularreactions involving neighboring amine residues withsuccinimides to form diketopiperazines can be par-ticularly important.12 In nonaqueous solutions, co-valent, amide-linked dimers and higher aggregatesmay predominate.30 The recent literature also con-tains hints that succinimide intermediates may beinvolved in the formation of amide-linked peptide orprotein aggregates in lyophile formulations. For ex-ample, Simons et al.37 observed the formation of co-valent dimers, trimers, and tetramers when lyophilesof RNase A were stored at high temperature, whichthey attributed to the formation of amide bonds be-tween RNase molecules located in close proximity toeach other. Later, Maroufi et al.38 reported amide-linked dimer formation in amorphous hen egg-whitelysozyme. Desfougeres et al.31 demonstrated a lin-ear correlation between the percentage of succin-imide and the amounts of covalent dimers formedin lysozyme lyophiles stored at high temperature. Inthe present study, covalent Gly–Val adduct formationin lyophiles of Gly–Phe–L-Asn–Gly containing excessquantities of Gly–Val has been shown to compete withwater for reaction with the succinimide intermediate.

In formulation A lyophiles containing Gly–Phe–L-Asn–Gly and a 7.5-fold excess of Gly–Val in a matrixconsisting of HPMC and sodium bicarbonate stored at40◦C and 75% RH, hydrolysis products were found todominate but both hydrolysis products and covalentGly–Val adduct formation proceeded through succin-imide intermediates. The involvement of the succin-imide intermediate in the formation of the observeddegradants is confirmed in Figure 3 by the successof the kinetic model defined by Eqs. 1–11 in simulta-neously fitting the concentration versus time profilesfor Gly–Phe–L-Asn–Gly (Fig. 3a), the succinimide in-termediates (Fig. 3d), the L- and D- isoaspartyl andaspartyl hydrolysis products (Fig. 3b), and the fourGly–Val adducts (Fig. 3c).

Increasing the molar excess of Gly–Val to approxi-mately 30-fold in formulation B and removing sodiumbicarbonate to compensate for the additional massof Gly–Val in the lyophilized solids along with stor-age at 27◦C and a lower humidity of 40% RH re-sulted in a shift in the succinimide intermediate



partitioning toward Gly–Val adducts as the predom-inant products. Again, the central role of the suc-cinimide intermediate was confirmed by the demon-stration that the mechanism-based kinetic model de-scribed by the set of parameter values in Table 2was successful in simultaneously fitting the kineticsof Gly–Phe–L-Asn–Gly disappearance (Fig. 4a), suc-cinimide and Gly–Val adduct formation regardless ofthe starting reactant [Gly–Phe–L-Asn–Gly (Fig. 4b),L-succinimide (Fig. 4d), or D-succinimide (Fig. 4e)],and L- and D-succinimide disappearance and inter-conversion when these compounds were the initialreactants (Fig. 4c).

Simultaneous fitting of the degradation profilesgenerated from Gly-Phe-L- Asn-Gly, Gly-Phe-L- Asu-Gly, or Gly-Phe-D- Asu-Gly as the starting reactantprovided the same rate constants for formation forthe normal and isoadducts from the correspondingsuccinimide in formulations B1, B2, and B3, respec-tively (Table 2). The ratio of attack by the N-terminusof Gly–Val at the L-succinimide "-carbonyl versus the$-carbonyl was approximately 1.6:1, whereas for theD-succinimide this ratio was approximately 2.6 (seeTable 2), consistent with the preference for nucle-ophilic attack at the "-carbonyl found when water isthe reactant.

To successfully fit the formation of covalent, amide-linked adduct concentration profiles from the Gly-Phe-L-Asn-Gly, Gly-Phe-D-Asu-Gly, or Gly-Phe-L-Asu-Gly data sets described in Figure 4, the model didrequire an additional degradation pathway (kx) toachieve mass balance. This rate constant may in-corporate those pathways leading to isoaspartyl andaspartyl hydrolysis products that could not be indi-vidually quantified in formulation B kinetic studiesdue to their low concentrations, potential succinimidedimerization, downstream reactions of degradants tosecondary decomposition products, and possible reac-tion of the succinimides with other nucleophiles, suchas free hydroxyl groups on HPMC sugar moieties.39

Succinimide Racemization in Lyophiles

In addition to the formation of aspartyl and isoas-partyl peptides, aspartate racemization occurs alongthe deamidation pathway as evident by the formationof D- and L-hydrolysis products.11 In their now clas-sic 1987 study, Geiger and Clarke11 were able to at-tribute two-thirds of the total racemization observedin solution during deamidation of an asparagine-containing hexapeptide to the succinimide interme-diate. The remainder they attributed to possible di-rect abstraction of the proton in aspartate peptidesor to other racemization-prone reactive intermedi-ates. The conversion of naturally occurring L-aminoacids to D-amino acids has been reported at neu-

tral pH with aspartates having a higher propen-sity for racemization.40 However, racemization ofaspartic and isoaspartic acids cannot account forthe observations of Geiger and Clarke11 becausethe interconversion rates of aspartates are muchslower—on the order of years in solutions whenextrapolated to lower temperatures.41 Li et al.42

observed that Gly–Gln–L-Asn–Glu–Gly transientlyformed Gly–Gln–D-Asn–Glu–Gly and vice versa dur-ing the process of deamidation in aqueous solution atpH 10 and 70◦C, which they attributed to racemiza-tion at the tetrahedral intermediate stage.42

The extent of peptide and protein racemizationin the solid state and the factors that may affect ithave not been systematically explored, to our knowl-edge. In the present work, the model depicted inFigure 1 assuming racemization only at the succin-imide was able to account for the kinetics of forma-tion of all D-and L-degradation products observed.Alternative models were considered, such as racem-ization prior to the formation of the succinimide, butno improvement in fits could be demonstrated. How-ever, these results do not eliminate the possibilityfor racemization at other points along the deami-dation pathway.42 The calculated rate constants forsuccinimide interconversion of approximately 0.2 h−1

in formulations B2 and B3 at 27◦C and 40% RHare comparable with the pseudo-first-order rate con-stants for other reaction pathways involving succin-imide (Table 2). Consequently, the concentration ofthe D-succinimide in formulations containing the L-succinimide as the starting reactant rose to signif-icant concentrations in relation to other degradantsduring the first half-life of succinimide disappearance(Fig. 4c) and the same was true for the L-succinimidewhen the starting reactant was the D-succinimide.Geiger and Clarke11 reported a first-order rate con-stant for racemization in aqueous solutions at pH7.4 and 37◦C that was 16-fold less than that for hy-drolysis of the succinimide.11 Under the formulationand storage conditions reflected in Table 2, succin-imide racemization kinetics proceeded more rapidlythan hydrolysis and exceeded the rate constants forracemization reported by Geiger and Clarke11 by ap-proximately one order-of-magnitude, even though theoverall rate of deamidation of Gly–Phe–L-Asn–Glyin the present study was approximately an order-of-magnitude slower than that for the hexapeptideexamined by Geiger and Clarke.11 Several factors, in-cluding the reduced water content in lyophiles as wellas formulation composition, peptide structure, tem-perature, and pH differences may contribute to thesedisparities. Nevertheless, these results highlight thepotential significance of racemization in amorphouslyophiles of Asn-containing peptides or proteins.



CONCLUSION

A mechanism-based kinetic model that assigns a cen-tral role to the D- and L-succinimide intermediateshas been employed to quantitatively account for theformation of hydrolysis and covalent, amide-linkeddegradants resulting from the deamidation of a modeltetrapeptide in the presence of excess Gly–Val inlyophilized formulations. Terminal degradant profileswere the same when the starting reactant was eitherGly–Phe–L-Asn–Gly or its L-succinimide intermedi-ate, further indicating that hydrolysis, racemization,and covalent, amide-linked adduct formation all pro-ceed via the succinimide intermediate. Nucleophilicattack occurred preferentially at the "-carbonyl of thesuccinimide intermediate when the nucleophile waseither water or the N-terminus of Gly–Val. The bi-molecular reaction between the succinimide interme-diate and the free amine of another peptide reportedherein suggests a likely mechanism for the forma-tion of amide-linked nonreducible aggregates in solid-state formulations of proteins and peptides.

Future studies will examine in a more systematicfashion the kinetics of formation of hydrolysis andcovalent adducts in similar formulations as a func-tion of such variables as water content and excipientdilution.

ACKNOWLEDGMENTS

Partial support for this project was provided by theH.B. Kostenbauder Endowed Professorship.

REFERENCES

1. Duddu SP, Sokoloski TD. 1995. Dielectric analysis in the char-acterization of amorphous pharmaceutical solids. 1. Molecularmobility in poly(vinylpyrrolidone)–water systems in the glassystate. J Pharm Sci 84:773–776.

2. Yoshioka S, Aso Y, Kojima S. 1999. The effect of excipientson the molecular mobility of lyophilized formulations, as mea-sured by glass transition temperature and NMR relaxation-based critical mobility temperature. Pharm Res 16:135–140.

3. Sun WQ, Davidson P, Chan HS. 1998. Protein stability in theamorphous carbohydrate matrix: Relevance to anhydrobiosis.Biochim Biophys Acta 1425(1):245–254.

4. Williams ML, Landel RF, Ferry JD. 1955. Mechanical prop-erties of substances of high molecular weight. 19. The tem-perature dependence of relaxation mechanisms in amorphouspolymers and other glass-forming liquids. J Am Chem Soc77(14):3701–3707.

5. Yoshioka S, Aso Y, Kojima S. 2004. Temperature- andglass transition temperature-dependence of bimolecular re-action rates in lyophilized formulations described by theAdams–Gibbs–Vogel equation. J Pharm Sci 93:1062–1069.

6. Pikal MJ, Rigsbee D, Roy ML, Galreath D, Kovach KJ, WangB, Carpenter JF, Cicerone MT. 2008. Solid state chemistry ofproteins: II. The correlation of storage stability of freeze-driedhuman growth hormone (hGH) with structure and dynamicsin the glassy solid. J Pharm Sci 97:5106–5121.

7. Yoshioka S, Aso Y, Kojima S. 2001. Usefulness of theKohlrausch–Williams–Watts stretched exponential function to

describe protein aggregation in lyophilized formulations andthe temperature dependence near the glass transition temper-ature. Pharm Res 18:256–260.

8. Yoshioka S, Tajima S, Aso Y, Kojima S. 2003. Inactivation andaggregation of $-galactosidase in lyophilized formulation de-scribed by Kohlrausch–Williams–Watts stretched exponentialfunction. Pharm Res 20:1655–1660.

9. Pikal MJ, Rigsbee DR. 1997. The stability of insulin in crys-talline and amorphous solids: Observation of greater stabilityfor the amorphous form. Pharm Res 14:1379–1387.

10. Shamblin SL, Hancock BC, Pikal MJ. 2006. Coupling betweenchemical reactivity and structural relaxation in pharmaceuti-cal glasses. Pharm Res 23:2254–2268.

11. Geiger T, Clarke S. 1987. Deamidation, isomerization, andracemization at asparaginyl and aspartyl residues in peptides.Succinimide-linked reactions that contribute to protein degra-dation. J Biol Chem 262:785–794.

12. Dehart MP, Anderson BD. 2007. The role of the cyclic imidein alternate degradation pathways for asparagine-containingpeptides and proteins. J Pharm Sci 96(10):2667–2685.

13. Oliyai C, Borchardt RT. 1993. Chemical pathways of peptidedegradation. IV. Pathways, kinetics, and mechanism of degra-dation of an aspartyl residue in a model hexapeptide. PharmRes 10:95–102.

14. Strickley RG, Anderson BD. 1996. Solid-state stability of hu-man insulin I. Mechanism and the effect of water on the kinet-ics of degradation in lyophiles from pH 2–5 solutions. PharmRes 13:1142–1153.

15. Strickley RG, Anderson BD. 1997. Solid-state stability of hu-man insulin II. Effect of water on reactive intermediate par-titioning in lyophiles from pH 2–5 solutions—Stabilizationagainst covalent dimer formation. J Pharm Sci 86:645–653.

16. Thing M, Zhang J, Laurence J, Topp EM. 2010. Thiol–disulfide interchange in the tocinoic acid/glutathione systemduring freezing and drying. J Pharm Sci 99(12):4849–4856.

17. Zale SE, Klibanov AM. 1986. Why does ribonuclease ir-reversibly inactivate at high temperatures? Biochemistry25:5432–5444.

18. Qiu Z, Stowell JG, Morris KR, Byrn SR, Pinal R. 2005. Ki-netic study of the Maillard reaction between metoclopramidehydrochloride and lactose. Int J Pharm 303:20–30.

19. Luo D, Anderson BD. 2006. Kinetics and mechanism for thereaction of cysteine with hydrogen peroxide in amorphouspolyvinylpyrrolidone lyophiles. Pharm Res 23:2239–2253.

20. Luo D, Smith SW, Anderson BD. 2005. Kinetics and mech-anism of the reaction of cysteine and hydrogen peroxide inaqueous solution. J Pharm Sci 94:304–316.

21. Oliyai C, Patel J, Carr L, Borchardt RT. 1994. Solid-state sta-bility of lyophilized formulations of an asparaginyl residuein a model hexapeptide. J Parenteral Sci Technol 48:67–173.

22. Song Y, Schowen RL, Borchardt RT, Topp EM. 2001. Effectof ‘pH’ on the rate of asparagine deamidation in polymericformulations: ‘pH’-Rate profile. J Pharm Sci 90:141–156.

23. Capasso S, Balboni G, Di Cerbo P. 2000. Effect of lysineresidues on the deamidation reaction of asparagine sidechains. Biopolymers 53(2):213–219.

24. Athmer L, Kindrachuk J, Georges F, Napper S. 2002. Theinfluence of protein structure on the products emerging fromsuccinimide hydrolysis. J Biol Chem 277(34):30502–30507.

25. Brennan TV, Clarke S. 1993. Spontaneous degradation ofpolypeptides at aspartyl and asparaginyl residues: Effects ofthe solvent dielectric. Protein Sci 2(3):331–338.

26. Wakankar AA, Borchardt RT. 2006. Formulation considera-tions for proteins susceptible to asparagine deamidation andaspartate isomerization. J Pharm Sci 95(11):2321–2336.

27. Lai MC, Topp EM. 1999. Solid-state chemical stability of pro-teins and peptides. J Pharm Sci 88(5):489–500.



28. Robinson NE, Robinson AB. 2004. Molecular clocks—Deamidation of asparaginyl and glutaminyl residues in pep-tides and proteins. 1st ed. Cave Junction, Oregon: AlthousePress, pp 419.

29. Aswad DW, Ed. 1994. Deamidation and isoaspartate formationin peptides and proteins. Boca Raton, Florida: CRC Press Inc.,pp 272.

30. Severs JC, Froland WA. 2008. Dimerization of a PACAP pep-tide analogue in DMSO via asparagine and aspartic acidresidues. J Pharm Sci 97(3):1246–1256.

31. Desfougeres Y, Jardin J, Lechevalier V, Pezennec S, NauF. 2011. Succinimidyl residue formation in hen egg-whitelysozyme favors the formation of intermolecular covalentbonds without affecting its tertiary structure. Biomacro-molecules 12(1):156–166.

32. DeHart MP, Anderson BD. Kinetics and mechanisms of deami-dation and covalent amide-linked adduct formation in amor-phous lyophiles of a model asparagine-containing peptide.Pharm Res [Epub ahead of print.]

33. Li B, Gorman EM, Moore KD, Williams T, Schowen RL, ToppEM, Borchardt RT. 2005. Effects of acidic N + 1 residues onasparagine deamidation rates in solution and in the solid state.J Pharm Sci 94(3):666–675.

34. Patel K, Borchardt RT. 1990. Chemical pathways of peptidedegradation. II. Kinetics of deamidation of an asparaginylresidue in a model hexapeptide. Pharm Res 7:703–710.

35. Brennan TV, Clarke S. 1995. Deamidation and iso-aspartateformation in model synthetic peptides: The effects of sequenceand solution environment. In Deamidate and iso-aspartate for-mation in peptides and proteins; Aswad DS, Ed. Ann Arbor,Michigan: CRC Press, pp 65–90.

36. Lai MC, Hageman MJ, Schowen RL, Topp EM. 1999. Chemicalstability of peptides in polymers. 1. Effect of water on peptidedeamidation in poly(vinyl alcohol) and poly(vinylpyrrolidone)matrices. J Pharm Sci 10:1073–1080.

37. Simons BL, King MC, Cyr T, Hefford MA, Kaplan H. 2002.Covalent cross-linking of proteins without chemical reagents.Protein Sci 11:1558–1564.

38. Maroufi B, Ranjbar B, Khajeh K, Naderi-Manesh H, YaghoubiH. 2008. Structural studies of hen egg-white lysozymedimer: Comparison with monomer. Biochim Biophys Acta1784:1043–1049.

39. Killion RB Jr., Stella VJ. 1990. The nucleophilicity of dextrose,sucrose, sorbitol, and mannitol with p-nitrophenyl esters inaqueous solution. Int J Pharm 66:149–155.

40. Bada JL. 1972. Kinetics of racemization of amino acids as afunction of pH. J Am Chem Soc 94(4):1371–1373.

41. Bada JL, Helfman PM. 1975. Amino acid racemization datingof fossil bones. World Archaeol 7(2):160–173.

42. Li B, Borchardt RT, Topp EM, VanderVelde D, Schowen RL.2003. Racemization of an asparagine residue during peptidedeamidation. J Am Chem Soc 125(38):11486–11487.


Jurnal Succinimide

Documents

Transcript of Jurnal Succinimide