Pharmaceutical Research, Vol. 27, No. 4, April 2010 (# 2010)
`DOI: 10.1007/s11095-009-0045-6
`
`Expert Review
`
`Stability of Protein Pharmaceuticals: An Update
`
`Mark Cornell Manning,1,4 Danny K. Chou,2 Brian M. Murphy,1 Robert W. Payne,1 and Derrick S. Katayama3
`
`Received October 6, 2009; accepted December 27, 2009; published online February 9, 2010
`Abstract. In 1989, Manning, Patel, and Borchardt wrote a review of protein stability (Manning et al.,
`Pharm. Res. 6:903–918, 1989), which has been widely referenced ever since. At the time, recombinant
`protein therapy was still in its infancy. This review summarizes the advances that have been made since
`then regarding protein stabilization and formulation. In addition to a discussion of the current
`understanding of chemical and physical instability, sections are included on stabilization in aqueous
`solution and the dried state, the use of chemical modification and mutagenesis to improve stability, and
`the interrelationship between chemical and physical instability.
`
`KEY WORDS: formulation; protein stability; protein stabilization.
`
`INTRODUCTION
`
`In 1989, Manning, Patel and Borchardt wrote a review
`summarizing what was known at the time about the stability
`and stabilization of protein pharmaceuticals (1), an article
`that has been referenced almost 500 times. In the late 1980s,
`there were only three recombinant protein products on the
`US market: Orthoclone (OKT-3), human insulin, and tissue
`plasminogen activator. If one included plasma-derived prod-
`ucts, the number of approved proteins only numbered about
`a dozen. Clearly, recombinant DNA technology has drasti-
`cally changed the pharmaceutical market. Now there are
`nearly twenty antibody products and almost 150 approved
`protein-based products that are commercially available in the
`US alone. In addition, our knowledge regarding protein
`stability and formulation has increased dramatically. The
`purpose of this review is to provide an update regarding
`what we have learned in the past 20 years. In addition to
`updating the sections of the original review article, some
`discussion is provided regarding topics that were not found in
`the literature at the time, such as the interrelationship of
`chemical and physical
`instability,
`instabilities that occur
`during bioprocessing, the impact of lyophilization cycle on
`protein stability, and the importance of packaging in main-
`taining protein stability.
`One can separate protein instabilities into two general
`classes: chemical instability and physical instability. Chemical
`instabilities involve processes that make or break covalent
`bonds, generating new chemical entities. A list of the more
`commonly observed chemical degradation processes is listed
`in Table I. Conversely, there are physical
`instabilities for
`proteins in which the chemical composition is unaltered, but
`
`1 Legacy BioDesign LLC, Johnstown, Colorado 80534, USA.
`2 Genzyme, Framingham, Massachusetts 01701, USA.
`3 Amylin Pharmaceuticals, San Diego, California 92121, USA.
`4 To whom correspondence should be addressed. (e-mail: manning@
`legacybiodesign.com)
`
`the physical state of the protein does change. This includes
`denaturation, aggregation, precipitation, and adsorption
`(Table I). The term precipitation is used here to denote
`insolubility rather than insoluble aggregate formation.
`Our knowledge of all protein degradation pathways is
`markedly greater than it was 20 years ago. Therefore, the
`emphasis of this review is on the progress that has been made
`since 1989. In addition, there were degradation processes and
`topics that were barely discussed or observed at that time. Those
`are now included as separate sections below. For example, there
`have been many articles on increasing conformational stability
`of proteins with various excipients, both in aqueous solution and
`in the dried state. In addition, a brief overview is provided of
`protein stabilization methods, including various drying methods,
`chemical modification, and site-directed mutagenesis. Finally, a
`discussion of the interrelationship between chemical and
`physical instability is provided.
`
`CHEMICAL INSTABILITY
`
`Deamidation
`
`it was already appreciated that
`Twenty years ago,
`deamidation, which involves the hydrolysis of Asn and Gln
`side chain amides, was a common degradation pathway for
`proteins and peptides. It is still regarded as the most common
`chemical degradation pathway for peptides and proteins.
`From a regulatory perspective, deamidation generates proc-
`ess-related impurities and degradation products. In addition,
`it may contribute to increased immunogenicity (2).
`At the time of the original review article, there were a
`few examples of deamidation in pharmaceutically relevant
`proteins,
`including human growth hormone (hGH) (3,4),
`insulin (5), γ-globulin (6), and hemoglobin (7). Moreover, the
`effect of extrinsic factors, such as pH, temperature, and ionic
`strength, were known as well (8). Since that time, the amount
`of information now available on deamidation and related
`
`0724-8741/10/0400-0544/0 # 2010 Springer Science+Business Media, LLC
`
`544
`
`1 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`Stability of Protein Pharmaceuticals: An Update
`
`545
`
`Table I. Chemical Instabilities Reported for Proteins of Therapeutic
`Interest
`
`Deamidation
`
`Asp–isoAsp interconversion/isomerization
`Racemization
`Proteolysis
`Beta-elimination
`Oxidation
`Metal-Catalyzed Oxidation (MCO)
`Photooxidation
`Free radical cascade oxidation
`Disulfide exchange
`DKP formation
`Condensation reactions
`pGlu formation
`Hinge region hydrolysis
`Trp hydrolysis
`
`reactions has increased significantly, as can be found in a
`number of excellent review articles (9–12) as well as entire
`books on the subject (13,14). There is even a web site devoted
`to this topic (www.deamidation.org).
`
`Asn Deamidation
`
`For those unfamiliar with this reaction, deamidation of Asn
`residues under acidic conditions takes place by direct hydrolysis
`of the Asn side chain amide to form only Asp. Under these
`conditions, deamidation is subject to acid catalysis. Similarly,
`Gln residues are converted to Glu (as is described in more detail
`
`below). However, this mechanism is rarely observed, as the pH
`must be less than 3. In neutral to basic solution (i.e., above pH
`6), the mechanism changes to an intramolecular cyclization
`reaction. The first step involves nucleophilic attack of the n+1
`nitrogen of the protein backbone on the carbonyl group of the
`Asn side chain (Fig. 1). This step is base catalyzed, since
`abstraction or partial abstraction of the backbone amide proton
`makes the nitrogen more nucleophilic, accelerating the reaction.
`A cyclic imide (also called succinimide or Asu) intermediate is
`formed (Fig. 1) with loss of ammonia. Since ammonia is a gas
`and is typically not retained in solution, this step is effectively
`irreversible. While the Asu intermediate often can be detected
`as a degradation product in its own right (see below), it is readily
`hydrolyzed in aqueous solution to form the Asp and isoAsp
`products (Fig. 1). Formation of the Asu five-membered ring
`intermediate is thought to be the reason that Asn deamidation is
`more prevalent than Gln deamidation, as five-membered
`heterocyclic rings are more stable than the six-membered rings
`associated with Gln deamidation.
`Consequently, deamidation generates two degradation
`products (Asp and isoAsp) at the site of the original Asn
`residue. Coupled with the possibility for racemization (15), four
`possible products (L-Asp, D-Asp, L-isoAsp, and D-isoAsp)
`could be formed. It is now known that racemization does not
`occur to any appreciable extent from the Asu intermediate, as
`was previously thought. Instead, it appears to be a parallel
`degradation pathway (16). Dehart and Anderson have provided
`a detailed kinetic description of the intramolecular cyclization
`(17). The same observation of a lack of racemization via the
`cyclic imide intermediate has been made for larger proteins as
`well (18).
`
`0
`■ C-Nfia
`~
`-NH-~~M-<-&.- ~
`
`R
`
`0
`
`0
`
`L- Aspereginyl peptide
`
`N~ alow
`'i
`t ;rc,
`-N-t~ -C-
`I
`
`fl
`I
`N-C:U-c-
`II
`0
`
`~L-Cyol~ lmloo~~O
`~o
`
`D-laoAaparty? poptlde u
`..:::=: 0 • Cyolb lmidc
`H
`~~ D-Aepartyl pepHd.t
`R
`0
`I
`I
`c;;fr c-r-H-Oi-C-
`~
`
`o
`
`t.t o-
`
`H
`I
`· I
`-N-f--a-l-c-l'lli-D-1- c-
`I
`l
`0
`0
`
`-N-1-~C-O-
`I
`0
`
`Fig. 1. General mechanism for deamidation of Asn residues and isomerization of Asp to isoAsp (taken from
`reference 1). Direct hydrolysis occurs below pH 4 while the cyclic imide pathway predominates at pH 6 and above.
`
`L-l&oA$pQrtyl peplldo
`
`2 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`546
`
`Manning et al.
`
`Effect of Sequence on Asn Deamidation
`
`Some work had been done by 1989 on the effects of primary
`sequence on deamidation, especially recognizing that Asn–Gly
`were particularly prone to deamidation. Subsequently, the
`influence of sequence was examined by Robinson and coworkers
`in detail (19,20). Ultimately, their work on sequence effects
`resulted in effective predictive schemes (14,21–25). In general,
`two trends are apparent. First, having amino acids with smaller
`side chains after the Asn residues leads to faster deamidation,
`presumably due to lack of steric hindrance of the initial cyclization
`reaction. Second, succeeding amino acids that have side chains
`that can act as hydrogen bond donors tend to accelerate the
`reaction, likely due to intramolecular hydrogen binding to the
`carbonyl oxygen of Asn, making it more electrophilic and thereby
`more reactive to nucleophilic attack.
`As a result, one does not need to be concerned about
`deamidation at every Asn residue. Only those Asn residues
`followed by a small or hydrogen bond-donating (e.g., Ser, Asn,
`or Asp) residues are found to exhibit deamidation on a time scale
`relevant to the pharmaceutical scientist. For example, Chelius
`et al. found that Asn deamidation in monoclonal antibodies
`(MAbs) occurred at Asn–Gly and Asn–Asn sequences (26),
`while Xiao and Bondarenko found deamidation at Asn–Asp
`sequences (27). Overall, Asn–Gly is the most reactive sequence
`in polypeptides, consistent with the schemes of Robinson and
`Robinson (Table II). For the most part, the preceding residue
`has little or no effect on deamidation rate, at least in solution.
`However, Li et al. have shown that Gln or Glu in that position
`appears to accelerate deamidation in the solid state, presumably
`by increasing hydration around the Asn residue (28).
`For deamidation that occurs at acidic pH, the mechanism
`does not involve cyclic imide formation at all. Instead, the
`protonated amide side chain undergoes direct nucleophilic
`attack by water. Therefore, it is not surprising that sequence
`
`Table II. Relative Deamidation Rates for Asn–Xaa where Xaa is the
`Succeeding Amino Acid (Taken from Reference 21)
`
`Residue
`
`% deamidation after
`100 d (Tris buffer)
`
`% deamidation after
`100 d (phosphate buffer)
`
`Gly
`Ser
`His
`Ala
`Asp
`Glu
`Asn
`Thr
`Lys
`Gln
`Cys
`Lys
`Gln
`Arg
`Phe
`Met
`Tyr
`Trp
`Leu
`Val
`Ile
`
`38
`9.8
`7.7
`5.6
`4.8
`3.0
`1.5
`2.0
`1.8
`1.7
`1.1
`1.8
`1.7
`1.6
`1.1
`0.9
`0.9
`0.4
`0.3
`
`87.3
`34.6
`33.2
`17.9
`19.0
`13.1
`13.5
`12.3
`10.4
`10.1
`7.3
`10.4
`10.1
`9.4
`7.3
`5.4
`5.4
`5.0
`5.4
`2.8
`1.3
`
`has been found to play a minimal role in controlling
`deamidation rates (29).
`
`Effect of Higher Order Structure on Asn Deamidation
`
`In 1989, the ability of higher order structure to influence
`deamidation rates was just starting to be appreciated. In 1988,
`Kossiakoff demonstrated that polypeptide chain flexibility
`impacted deamidation rates (30). Other studies have since arrived
`at the same conclusions, examining the relative deamidation rates
`for Asn residues dispersed across a given globular protein
`structure (31,32). In addition, a number of studies have shown
`that placement of the reactive Asn residue within an ordered
`secondary structure slows the reaction rate. This has been found
`for α-helices (33,34), β-sheets (35), and β-turns (36,37).
`Combining information about primary sequence along
`with the location of an Asn residue within a three-dimen-
`sional structure leads to improved predictive accuracy for
`deamidation rates (22). Moreover, alterations in the three-
`dimensional structure can affect deamidation rates. For
`example, addition of ligands that induce α-helical structure
`in insulin slow deamidation at AsnB3 (38).
`
`Deamidation in Monoclonal Antibodies (MAbs)
`
`Our knowledge regarding the stability and structure of
`MAbs has increased exponentially over the past 20 years.
`This includes detailed studies of deamidation in these
`pharmaceutically important molecules. In general, deamida-
`tion is responsible for much of the heterogeneity observed in
`MAbs along with other kinds of chemical
`instability and
`glycosylation differences (39).
`In 1992, Kroon et al. reported that OKT-3, the first
`marketed monoclonal antibody product, undergoes deamida-
`tion (40). Subsequently, there were sporadic reports of deami-
`dation in MAbs over the next decade (41–43). In the last 5 years,
`the number of reports on deamidation in MAbs has increased
`significantly. Some focus on the effect of extrinsic factors (44),
`some on sequence effects (26), while others emphasize the
`analytical methods used to monitor and quantify deamidation,
`which is primarily done by some type of mass spectrometry
`(26,45–57). These studies provide a solid basis for monitoring
`and quantifying deamidation in any protein or peptide. Other
`groups have reported using charge separation methods to detect
`and quantify deamidation in peptides and proteins (58–61),
`while others have employed RP HPLC (62,63), peptide
`mapping (64), and even Raman spectroscopy, which was
`reported to detect deamidation (65). However, the latter is
`quite insensitive, requiring deamidation to exceed 10%.
`Prolonged storage of a human MAb resulted in deamida-
`tion at both Asn and Gln residues, as well as other chemical
`instabilities, such as fragmentation and pGlu formation (66–68).
`Those other degradation pathways are discussed below. What
`appears to be true is that the factors controlling deamidation
`rate (primary sequence, temperature, pH, etc.) in peptides and
`smaller proteins are equally important in MAb degradation.
`
`Deamidation of Other Protein Pharmaceuticals
`
`In addition to the large amount of work on MAbs, a
`number of other studies have appeared describing deamida-
`
`3 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`Stability of Protein Pharmaceuticals: An Update
`
`547
`
`Table III. Protein of Pharmaceutical Interst Where Deamidation has
`Been Observed
`
`Compound
`
`tPA
`IL-11
`rhGH
`hGH
`Tetanus vaccine
`Anthrax vaccine
`Anthrax protective antigen
`Fc fusion protein
`Glucagon
`Lymphotoxin
`Protein G
`Hemoglobin
`NGF
`Hirudin
`IL-1 receptor (type II)
`
`Reference
`
`(69)
`(47,70)
`(71)
`(72)
`(73)
`(74)
`(75)
`(76,77)
`(78)
`(31)
`(79)
`(80)
`(81)
`(82,83)
`(84)
`
`tion in peptides and protein of pharmaceutical relevance,
`including vaccines and antigens (47,69–84). These include the
`entries in Table III. In general, any protein or peptide that
`contains one of these reactive Asn–Xaa sequences will be
`prone to deamidation over time.
`
`Control of Deamidation Rates
`
`A number of formulation approaches have been described
`to slow deamidation. The most effective approach is to control the
`pH. Deamidation for a single reactive Asn displays a V-shaped
`pH-rate profile, with the minimum being between pH 3 and 6. In
`addition, being a chemical reaction, it displays typical Arrhenius
`behavior, provided the protein or peptide does not change
`conformation appreciably across the temperature range of study.
`Interestingly, it is possible to slow deamidation rates by
`altering the conformation of the protein. Even in 1989, it was
`known that a specific set of phi/psi angles is needed to allow the
`intramolecular nucleophilic attack to form the Asu intermedi-
`ate (85). Phi and psi refer to the dihedral angles for the Cα–N
`bond and the C(O)-Cα bond, respectively. Therefore, limiting
`the flexibility of the peptide chain should and does slow
`deamidation. This is the basis for slower deamidation rates in
`well-defined and rigid higher-order structures (see above). It is
`possible to alter polypeptide chain flexibility using excluded
`solutes. Addition of sucrose to a flexible peptide caused it to
`adopt a β-turn conformation, thereby slowing deamidation
`(86). Sugars and polyols compact the structure of alcohol
`dehydrogenase and thereby slow deamidation in both the apo
`and holoenzymes (87). Similarly, removal of C-terminal amino
`acids in histidine-containing protein allows deamidation to
`proceed, presumably by removing steric constraints (88).
`Finally, one can imagine that formulations that lower NH
`acidity would slow deamidation rates. This has been done
`using nonaqueous solvents (33,89), although these same
`solvents can also affect conformation, viscosity, and solvent
`dielectric, so the effect might not be entirely due to
`modulation of acid-base properties. The effect of viscosity
`has been described for model peptides (90,91). Similarly,
`dielectric and viscosity effects have been examined for Asp
`isomerization in MAbs (92,93). In that case,
`increased
`
`chemical stability was obtained at the expense of reduced
`conformational stability. Therefore, such approaches using
`nonaqueous solvents may not be viable for many globular
`proteins but could work for peptides, where solution con-
`formation is less important to maintaining biological activity.
`Prior to 1989, it was known that certain buffers exhibited
`buffer catalysis of Asn deamidation. Most buffers had been
`shown to exhibit some degree of buffer catalysis. Therefore,
`limiting the amount of buffer used should slow deamidation
`rates. In the last 20 years, relatively little has been done on this
`topic. Tyler-Cross and Schirch (29) demonstrated that deami-
`dation of model peptides exhibited general base catalysis, but
`they did not observe specific base catalysis in their studies. So,
`apart from some observations on buffer effects, little has been
`done on mechanistic aspects of catalysis of deamidation. As for
`more recent observation on buffer effects, Girardet et al.
`reported that phosphate buffer increased deamidation rates in
`α-lactalbumin faster than tris buffer at pH 7.4 (94). Zheng and
`Janis conducted a detailed study on buffer effects on deamida-
`tion in a MAb,
`looking at tartrate, citrate, succinate, and
`phosphate (44). They found that citrate was the best choice,
`while the pH had to be less than 5.
`
`Deamidation in the Solid State
`
`The propensity of peptides and proteins to degrade
`chemically while in the solid state has been reviewed by Lai
`and Topp (95). Briefly, many of the reactions described here,
`including deamidation, have been observed for polypeptides
`in the solid state as well. For example, the deamidation rates
`of both cyclic and linear peptides were investigated in the
`solid state (37). A comparison of deamidation rates between
`solution and in the solid state can be found as well (96).
`Finally, Houchin and Topp (97) have recently reviewed the
`chemical degradation of peptides and proteins,
`including
`deamidation, encapsulated within PLGA microspheres.
`
`Gln Deamidation
`
`Our knowledge base regarding deamidation of Gln has
`increased tremendously over the past 20 years. It is still true that
`deamidation of Gln residues is less common than for Asn.
`Recall that cyclization of Asn residues leads to a five-membered
`ring. With Gln, that same intermediate is a six-membered ring,
`which is less favorable thermodynamically than the smaller ring.
`Certainly, Gln deamidation was known in 1989 (19). Yet, so little
`was reported that it was not discussed in our previous review.
`Since then, Joshi and Kirsch have reported some detailed
`mechanistic studies on Gln deamidation in peptides (78,98,99).
`A number of reports have found Gln deamidation in larger
`proteins, such as crystallins (100) and MAbs (101).
`
`Theoetical Studies on Deamidation
`
`In addition to the explosion of experimental studies on
`deamidation in peptides and proteins, a number of theoretical
`studies have emerged as well. These include molecular dynamics
`(MD) simulations (102) and ab initio calculations (103–105). Of
`note, Radkiewicz et al. 2001 showed that backbone conforma-
`tion (i.e., phi–psi angles) affect acidity of the NH group (106).
`Gly, being able to sample more conformational space, shows
`
`4 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`548
`
`Manning et al.
`
`increased NH acidity, which would contribute to increased Asn
`deamidation rates. Therefore, the increased reactivity of Asn–
`Gly sequences might not be entirely due to lack of steric
`hindrance to intramolecular nucleophilic attack.
`
`Succinimide Formation
`
`In general, deamidated forms (Asp and isoAsp), as well
`as the corresponding cyclic imide (Asu) intermediates, have
`been isolated and identified, especially in peptides. The cyclic
`imide intermediate has been repeatedly isolated and charac-
`terized in monoclonal antibodies. Groups from Amgen used
`hydrophobic interaction chromatography (HIC), cation
`exchange chromatographjy (CEX), and liquid chromatography-
`mass spectrometry (LC–MS) to identify Asu formation in
`MAbs, especially IgG2s, that were stored at elevated temper-
`atures (56,57,107,108). The primary degradation product
`appears to be the cyclic imide (Asu) intermediate at position
`30 of the light chain (LC). Other studies have reported Asu
`formation at position LC32 (109) and residue 102 of the heavy
`chain (43).
`Succinimide formation has been reported in other
`systems. For example, stressed samples of hGH form a
`succinimide product at an Asp–Gly site that was isolated
`and quantified using reversed-phase HPLC (110). Similar
`degradation has been reported for glial cell
`line-derived
`neurotrophic factor, which forms a succinimide product at
`position 96 (111). The degraded form was identical to the
`native protein in structure, pharmacokinetics and activity.
`Lysozyme has also been reported to form a succinimide
`product at a Asp–Gly site as well (112).
`
`Asp Isomerization
`
`Once the cyclic imide intermediate forms, it can open to
`form either Asp or isoAsp products (Fig. 1). Such a
`mechanism indicates that Asp itself could cyclize to form
`the same succinimide (Asu) species, thereby allowing con-
`version from Asp to isoAsp. This reaction has been called
`Asp–isoAsp interconversion, but is more commonly referred
`to as Asp isomerization. The rate-limiting step is the same for
`both deamidation and Asp isomerization, that is, the rate is
`controlled by formation of the cyclic imide intermediate.
`Consequently, the same approach can be taken to slow each
`reaction. In other words, pH provides the greatest degree of
`control by slowing deprotonation that leads to intramolecular
`cyclization. Early work on this reaction indicated that only
`the protonated form of Asp isomerizes, i.e., there is much
`lower reaction rate above pH 5 (113). In fact, above pH 8, the
`reaction is independent of pH and buffer concentration.
`Below pH 3, only hydrolysis is observed. The size of the C-
`terminal amino acid retards the formation of the cyclic imide
`intermediate (114), thereby slowing Asp isomerization. Steric
`constraints affect cyclization rates, as with deamidation (88).
`Since the original review was published, Asp isomerization
`has been reported in many systems, especially monoclonal
`antibodies (27,43,55,92,93,107,108,115). Some of the same LC-
`MS methods used to identify deamidation in MAbs have been
`used to monitor Asp isomerization as well (27,55,116). Both
`degradation pathways have been observed in MAbs (43,55).
`Specifically, Asp isomerization has been reported at position 32
`
`in the light chain (93,109) and position 102 in the heavy chain
`(43). For Asp–Asp sequences in MAbs, both Asp isomerization
`and Asp-assisted hydrolysis were observed (27).
`Racemization (which is discussed in more detail below)
`has been observed concomitantly with Asp isomerization
`(63), similar to the observations with deamidation (13,15).
`This emphasizes once again how interconnected many of
`these chemical degradation pathways can be.
`Other proteins of pharmaceutical interest besides MAbs
`have been reported to undergo Asp isomerization. For
`example, Asp93 isomerization has been shown to be the
`primary degradation pathway for NGF (81), while Asp
`isomerization (at Asp45 and Asp47) has been found in IL-11
`as well (70). Dette and Wätzig were able to resolve the
`isoAsp product of Asp isomerization in recombinant hirudin
`using capillary electrophoresis (117).
`Outside of controlling pH and temperature (see above),
`little has been reported on formulation strategies to slow Asp
`isomerization. The use of excluded solutes to provide
`conformational stability in a MAb actually decreased chem-
`ical stability by accelerating Asp isomerization (92). Presum-
`ably, changing the succeeding amino acid (in the n+1
`position) would also slow the reaction, but no detailed studies
`of that type have been reported.
`
`Asp Hydrolysis
`
`There is a third reaction that is associated with degradation
`at As/Asp residues and that is Asp-associated hydrolysis of the
`peptide backbone (also known as proteolysis). Unfortunately,
`there are few reviews available on the topic, with the most
`extensive dating back to 1983 (118). Since this reaction also
`involves intramolecular cyclization,
`it is not surprising that
`proteolysis shows the same pH-rate profile and sensitivity to
`buffer catalysis as deamidation (119). The mechanism was
`delineated in detail by Joshi and Kirsch (78), with nucleophilic
`attack occurring at the ionized side chain of Asp on the
`protonated carbonyl of the peptide backbone. This produces
`an anhydride species and release of the N-terminal portion of
`the peptide chain. There is some information available on the
`effect of primary structure on Asp hydrolysis. The presence of
`Ser or Tyr at position n+1 can accelerate reaction (98,99).
`Similarly, having Ser or Val at position n+1 accelerates
`hydrolysis relative to Asp isomerization (114).
`Other similar hydrolysis reactions have been reported.
`For example, the Asn–Pro bond appears to be particularly
`labile in the presence of ammonia (120). A similar degradation
`process has been reported for the Asp60–Pro61 bond in NGF
`(81). The peptide linkages in either side of Pro and Trp were
`found to hydrolyze in spantide II, a bioactive peptide (121).
`
`Hinge Region Hydrolysis
`
`Hydrolysis of the peptide backbone has been seen in
`antibodies even when Asp is not present. This reaction occurs
`most frequently within the hinge region of the antibody, so it
`is known as hinge region hydrolysis. However, it can occur at
`the CH2–CH3 interface as well (67). Typically, it occurs in
`IgG1s, so the reaction is likely influenced by the flexibility of
`the peptide chain. This reaction is distinct from the enzymatic
`hydrolysis that can occur in this region with antibodies (39).
`
`5 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`Stability of Protein Pharmaceuticals: An Update
`
`549
`
`There have been a number of detailed studies on this
`process. The first study reported cleavage in the hinge region
`of mouse MAbs (122), showing that the reaction can occur
`under basic pH conditions (122). Fragmentation, along with
`other chemical instabilities, was reported in OKT3, which is a
`mouse IgG2a antibody (40,123). Using MALDI-TOF and
`capillary electrophoresis, Alexander and Hughes found hinge
`region hydrolysis to occur in chimeric mouse/human IgGs
`(124), as was also reported by Paborji et al. (125).
`The general nature of this reaction was shown by Cordoba
`et al., who showed that hinge region hydrolysis occurred in four
`different human IgG1s (126). The observed fragmentation
`pattern indicated that the hydrolysis reaction is not specific to
`a particular peptide bond, but occurs within a narrow range of
`residues. In this case, hydrolysis was limited to the heavy chain
`sequence Ser-Cys-Asp-Lys-Thr-His-Thr. Similarly, descriptions
`of hinge region hydrolysis, detected in the course of mass
`spectrometry studies on MAbs, were reported as well (127,128).
`While chain flexibility appears to be important, recently it was
`demonstrated that conformational instability of Fab region leads
`to increased rates of hinge region hydrolysis as well (129).
`The pH-rate profile for hinge region hydrolysis is V-shaped
`(130), with a minimum near pH 6. The rate increases linearly
`with pH above pH 6. The study by Cordoba et al. indicated that
`EDTA and protease inhibitors have no effect on hydrolysis rates
`(126). In addition to the more general hinge region hydrolysis
`described above, there have been reports of metal-assisted
`hydrolysis of MAbs in the same region (131,132). In these cases,
`chelating agents have some ability to slow degradation.
`
`Trp Hydrolysis
`
`In addition to these better-known degradation processes,
`other functional groups are also sensitive to hydrolysis. For
`
`example, the side chain of Trp is known to undergo hydrolysis.
`The primary degradation product is called kynurenine (133–135),
`which fluoresces at much longer wavelength (450 nm) than Trp
`itself. Kynurenine and related substances can also form during
`oxidative degradation of Trp as well (see below).
`
`Racemization and β-Elimination
`
`These two degradation pathways are interrelated, as the
`initial step is the same: deprotonation of the hydrogen on the
`α-carbon (Fig. 2). Usually, C–H bonds have little acid-base
`reactivity, but the C–H bond of an amino acid does have
`some acidic character. As a result, racemization is usually a
`very slow process, so slow that it can be used to date artifacts.
`In vivo, a number of proteins have been reported to racemize,
`as in crystallins from the lens of the eye (136,137) and myelin
`in muscle (138).
`Typically, the racemization occurs at Asp residues (138),
`although racemization at Asn127 in murine lysozyme has
`been reported (139). Why this residue is more reactive is
`not yet known. A more extensive summary of amino acid
`racemization can be found in the review by McCudden and
`Kraus (140).
`Once the Cα–H bond ionizes, recombination can lead to
`racemization (Fig. 2). On the other hand, the resulting
`carbanion can rearrange and eject a leaving group from the
`β-carbon, producing a double bond between the alpha- and
`beta-carbon. This is β-elimination. At high temperatures, it
`appears that β-elimination of Cys residues occurs readily in a
`number of proteins (141). Among proteins of pharmaceutical
`interest, β-elimination has been reported for IL-1ra (142) and
`insulin (143). It has also been shown that β-elimination occurs
`under conditions causing hinge region hydrolysis (144).
`
`(,
`X
`I
`CHR
`CHR
`I
`Ii
`-1-N-c-c- -
`--HN-C=C-
`1;
`-~ ~)
`Carbanion intermediate
`
`I
`
`CHR
`II
`-HN-C-C- +x·
`II
`0
`
`X
`I
`CHR
`_
`I
`-1-N-c-c- -
`l _} ~
`HJ
`
`OH"
`
`L-amino acid residue
`
`11
`
`H+
`
`-H+
`
`H
`I
`-HN-C-C-
`II
`1
`CHR 0
`I
`X
`
`D-amino acid residue
`Fig. 2. General mechanism for racemization and β-elimination in proteins (taken from reference 1).
`
`6 of 32
`
`Fresenius Kabi
`Exhibit 1015
`
`

`

`550
`
`DKP Formation
`
`One other N-terminal cyclization process has been
`described in some detail. Note that the N-terminal amino
`group can be a potent nucleophile, especially above pH 8. If
`the amine attacks the second carbonyl group in the peptide
`backbone, a diketopiperazine (DKP) ring is formed. Degra-
`dation of the N-terminus of a peptide or protein by DKP
`formation has been commonly observed during long-term
`storage and during peptide synthesis (145–147).
`This reaction was initially observed in peptides (148),
`where the DKP ring can rearrange, either with loss of the first
`two amino acids or reversal of their positions in the chain.
`The extent of DKP formation depends on percentage of
`terminal amino groups in the free base form (17,149,150).
`Under acidic conditions, the reaction is quite slow and pH-
`independent. Kinetic analyses of DKP formation in peptides
`have examined the effects of pH, buffer type and concen-
`tration and temperature (17,149–151). The first-order rate
`constant generally increases with increasing buffer concen-
`tration, except for carbonate, which shows no concentration
`dependence (150). Degradation caused by DKP formation
`was shown to be responsible for the N-terminal heterogeneity
`observed in hGH (145) and substance P (152). Further details
`of
`the reaction kinetics of DKP formation have been
`presented recently (17).
`To the extent that DKP formation leads to reduction in the
`length of the polypeptide chain, it can be considered a proteolytic
`reaction. Rearrangement of a DKP from the first two amino
`acids, via cleavage of the peptide bond C-terminal to the second
`amino acid, produces a clipped protein reduced in molecular
`weight by the mass of the two amino acids. In solution, DKP
`formation is common for proteins with the N-terminal sequence
`NH2-Gly-Pro (153).
`
`pGlu Formation
`
`This reaction was not covered in the original 1989 review,
`although there were some literature references prior to