Perspective
`
`Engineering Challenges of
`Protein Formulations
`
`Theodore W. Randolph
`Dept. of Chemical and Biological Engineering, University of Colorado, Boulder, CO 80309
`
`John F. Carpenter
`Dept. of Pharmaceutical Sciences, University of Colorado Health Sciences Center, Denver, CO 80262
`
`DOI 10.1002/aic.11252
`Published online June 25, 2007 in Wiley InterScience (www.interscience.wiley.com).
`
`Keywords: protein stability, protein aggregation, shelf life, accelerated degradation
`
`Introduction
`
`Formulation Challenge
`
`Protein-based pharmaceuticals are the fastest-growing
`
`class of new drugs. They not only offer promise for
`treatments to address major health challenges, such as
`cancer, but also a wealth of new engineering problems to
`solve. Chemical engineers have long been proficient at pro-
`ducing products that meet exacting specifications for chemical
`purity, but therapeutic proteins now bring additional chal-
`lenges: these products must not only be highly chemically
`pure, but also conformationally pure, and must remain so dur-
`ing manufacturing and through the drug’s entire shelf-life and
`delivery to patients.
`Proteins degrade through a variety of mechanisms. These are
`usually classified as either physical or chemical degradation
`pathways.1 Physical degradation pathways include unfolding,
`misfolding, and aggregation of the protein molecules. Chemical
`degradation pathways encompass a myriad of unwanted
`chemical reactions that proteins commonly undergo, such as
`oxidation, deamidation,
`racemization, hydrolysis, disulfide
`exchange, and carbamylation. Classification of degradation
`pathways as physical or chemical
`is somewhat artificial,
`because the two types of degradation often are closely linked.
`For example, we have shown that an oxidation process result-
`ing in crosslinking of tyrosine residues in a-synuclein (a protein
`that forms characteristic fibrils in Parkinson’s disease), is a pre-
`cursor to aggregation and fibrillogenesis.2 Addition of radical
`scavenging molecules, such as methionine to a-synuclein for-
`mulations delays onset of in vitro fibril formation by reducing
`the rate of tyrosine oxidation formation. Conversely, oxidation
`of the serine protease subtilisin can be inhibited by adding for-
`mulation excipients, such as sucrose that act to increase the
`thermodynamic conformational stability of the protein.3
`
`Correspondence concerning this article should be addressed to T. W. Randolph
`at theodore.randolph@colorado.edu.
`
`Ó 2007 American Institute of Chemical Engineers
`
`To allow proteins to be used as therapeutic agents, proteins
`must be placed in a formulation that confers suitable stability
`against physical and chemical degradation. In addition to sta-
`bilizing the pharmaceutically-active protein ingredients, for-
`mulation components, or excipients, also must be compatible
`with their intended use. For example, a formulation intended
`for parenteral use (e.g., subcutaneous injection) must be ster-
`ile, nontoxic, and exhibit acceptable viscosity and tonicity.
`Although these requirements place limits on the types and
`concentrations of excipients that practically can be used, there
`are still far too many possible sets of formulations to allow a
`purely empirical screening approach to be successful.
`For economic viability,
`therapeutic protein formulations
`typically require a shelf life of 18–24 months.4 Over the
`course of this time, when stored as directed on the product
`label (usually refrigerated at 2–8 8C), the protein must retain
`adequate chemical and conformational purity. Meeting the
`stringent requirements for stability during shelf-life is a daunt-
`ing task. Most of the common chemical degradation products
`(especially hydrolysis and oxidation byproducts) are signifi-
`cantly thermodynamically favored vs. the desired native state
`of the protein. Furthermore, the properly folded native state of
`most proteins is only marginally more stable (the free energy
`of unfolding DGumf, is about 20–60 kJ/mol) than the folded
`state,5 and appears to be unstable under most conditions with
`respect to aggregated forms of the protein.6,7
`Required chemical and conformational purity levels are dic-
`tated by the individual protein’s safety and efficacy profile,
`but frequently levels of chemical impurities >5%, or confor-
`mational impurities >1% at the end of the labeled shelf-life
`might be considered unacceptable. Given that typical concen-
`trations of protein in therapeutic formulations are near
`10 micromolar, this suggests that levels of degradation prod-
`ucts typically must be held below 100 nanomolar over the
`course of two year storage. Thus, an average rate of degrada-
`
`1902
`
`August 2007 Vol. 53, No. 8
`
`AIChE Journal
`
`Ex. 2005-0001
`
`

`
`tion of 1 nanomolar/week may indicate an unacceptable level
`of product lability.
`As part of the approval process for protein therapeutics, the
`US Food and Drug Administration requires that protein drug
`stability be demonstrated in real-time, under conditions mim-
`icking the proposed labeled storage conditions, i.e., in the pro-
`posed container/closure system, at the proposed protein con-
`centration in a final formulation and under specified tempera-
`ture conditions. This requires that enormous resources be
`dedicated years before a product can be sold, and represents a
`bottleneck for the entire therapeutic protein development pro-
`cess. For products with anticipated annual sales often of more
`than $500M, delays in development of suitably stable formu-
`lations may, thus, represent lost sales of $1M or more per day.
`Clearly, a goal is to make sure that the formulations that enter
`real-time stability testing have a high-probability of success.
`For many kinds of protein degradation, acceptable levels of
`degradation products at the end of shelf life are very low. This
`creates a quandary. We wish to be reasonably sure at the onset
`of a real-time stability study that at the study’s completion 18
`or 24 months later we will have acceptably low-levels of deg-
`radation products. However, in part, because of analytical lim-
`itations, it is difficult or impossible to conduct a relatively
`short (e.g., one week) experiment under proposed storage con-
`ditions that can be extrapolated to an 18 or 24 month storage
`life. Thus, in order to predict which set of excipients are likely
`to provide an adequate shelf-life, accelerated degradation
`experiments must be conducted. In these experiments, formu-
`lations are subjected to an additional ‘‘stress’’, such as ele-
`vated temperature, freeze-thawing, the presence of air-water
`interfaces or high-or low-ionic strengths, and the kinetics of
`protein degradation measured. Combinations of excipients
`that protect against degradation under ‘‘stressed’’ conditions
`are then assumed to be most likely to confer stability under
`more benign storage conditions. Usually, the result of such
`studies is a formulation that provides the greatest relative sta-
`bility. However, there is no assurance that this level of stabili-
`zation will be sufficient for the shelf-life.
`By definition, accelerated degradation studies are conducted
`under conditions that are different from anticipated actual
`storage conditions. How predictive are these studies of protein
`behavior at actual storage conditions? fortunately, the answer
`(at least for relatively simple accelerated stability studies) is
`often ‘‘not very’’. Discrepancies between the predictions of
`simple accelerated studies and actual behavior might not be
`surprising, given the complicated structure of proteins, and the
`likely presence of multiple degradation pathways, but they
`serve to emphasize the need for better models and mechanistic
`understanding of protein degradation. Significant progress has
`been made in the ‘‘rational design’’ of protein formulations,4,8,
`but there remain lacunae in the mechanistic understanding of
`the protein degradation pathways, and their responses to accel-
`erated stability protocols. Some selected examples of chal-
`lenges and progress in the protein formulation arena follow.
`
`Thermally accelerated protein degradation
`
`Elevated temperature is perhaps the most obvious ‘‘stress’’
`condition that might be applied to accelerate protein degrada-
`tion. Intuitively, one might expect that a simple Arrhenius
`analysis might allow data obtained on protein degradation at
`
`elevated temperatures (e.g., rate constants for protein aggrega-
`tion measured at 50–80 8C) to be extrapolated to typical refri-
`gerated storage conditions. However, often it is found empiri-
`cally that predictions made using such an approach are poor.
`A reason for the observed discrepancy between predictions of
`degradation rates, based on simple Arrhenius analysis of ther-
`mally-accelerated stability studies and actual behavior, lies in
`the coupling of thermodynamic equilibria between various
`protein conformations, each with a characteristic reactivity for
`a given degradation pathway and the kinetics of reactions on
`that pathway. Some protein degradation pathways, notably
`those leading to aggregation, occur through partially unfolded
`intermediates or through reactive subpopulations of the pro-
`tein’s native state ensemble. Because of the marginal confor-
`mational stability of proteins, relatively small changes in tem-
`perature can significantly perturb conformational stability,
`which in turn alters the population of aggregation-prone pro-
`tein molecules. For example, Roberts9 has shown that the rate
`of aggregation of recombinant bovine granulocyte colony
`stimulating factor as a function of temperature, shows strik-
`ingly non-Arrhenius behavior and a simple prediction of
`shelf-life based on a simple Arrhenius extrapolation of data
`taken above room-temperature to 5 8C storage conditions,
`would lead to an overestimation of shelf-life by orders of
`magnitude. However, when the temperature dependency of
`recombinant bovine granulocyte colony stimulating factor
`conformational stability, and its effect on the observed rate
`constants were taken into account, the underlying rate con-
`stants for aggregation showed Arrhenius behavior.9
`A commonly used approach to screen excipients for protein
`formulations is to use differential scanning calorimetry to
`measure the apparent ‘‘melting temperature’’ Tm, or tempera-
`ture at which the protein olds in a given formulation. Formula-
`tions that yield elevated values are those in which the protein
`is presumed to be the most stable under real-time storage con-
`ditions.10 However, there are some excipients (notably non-
`ionic surfactants and some preservatives) that lower values,
`but have little detrimental effect or act to increase stability at
`lower temperature storage conditions. For example,
`in the
`presence of the preservative benzyl alcohol, the apparent of
`recombinant human interleukin-1 receptor antagonist
`is
`depressed by about 8 8C, and the protein aggregates rapidly at
`37 8C. However, little effect of benzyl alcohol is seen on the
`aggregation rate 25 8C.11 In part, the contradiction can be
`explained on the basis of the temperature dependency of
`hydrophobic interactions, which are strengthened at higher
`temperatures. At higher temperatures, increased hydrophobic
`interactions favor binding of preservative to the exposed
`hydrophobic regions of olded protein molecules, which,
`according to the Wyman linkage function,12 should result in a
`greater population of olded species, and, hence, a lower Tm
`than in the absence of preservative.11
`Protein aggregation frequently appears to result from multi-
`step and/or multipathway reactions.13–15 Because the activa-
`tion energies for each step in the reaction pathway may be dif-
`ferent, the rate-limiting step for the degradation process may
`change with temperature. In the case of proteins that exhibit
`multiple degradation pathways,
`the dominant degradation
`product that is formed during storage at refrigerated condi-
`tions may be different than that formed at temperatures used
`for accelerated stability studies. For example, when stored at
`
`AIChE Journal
`
`August 2007 Vol. 53, No. 8
`
`Published on behalf of the AIChE
`
`DOI 10.1002/aic
`
`1903
`
`Ex. 2005-0002
`
`

`
`interleukin-1 receptor
`temperatures near room-temperature,
`antagonist forms irreversible, soluble aggregates with nearly
`native secondary structure that are crosslinked through disul-
`fide bonds.16 In contrast, under accelerated degradation condi-
`tions at 55 8C, the protein forms aggregates that are not cross-
`linked through disulfides, but that contain significant non-
`native b-sheet structures.17 Clearly, in this case extrapolations
`of protein aggregation kinetics based on high-temperature
`studies would not be expected to be predictive of low-temper-
`ature storage behavior.
`Pressure is a variable that may provide a useful alternative
`to temperature for accelerated stability studies. Analogous to
`using Arrhenius plots to determine activation energies, semilo-
`garithmic plots of reaction rate constants vs. pressure may be
`used to determine the activation volume for a reaction. Acti-
`vation volumes may then be used to predict rate constants at
`other pressures of interest. Webb et al.18 measured activation
`volumes for aggregation of interferon-g, and found that the
`volume change required for aggregation 41 mL/mol, was
`unfolding of interferon-g (209 mL/mol), suggesting that the
`reactive species involved in the aggregation of interferon-g is
`a partially unfolded,19 rather than a completely olded spe-
`cies.18 Additional measurements of folding equilibria and
`aggregation kinetics made as a function of temperature and
`surface tension also show that the transition state for aggrega-
`tion of interferon-g is partially, rather than fully unfolded
`(Figure 1). In the case of interleukin-1 receptor antagonist,
`activation volumes for aggregation were nearly identical to
`those required to old the protein, suggesting that a nearly com-
`pletely olded species was required for aggregation. Interest-
`ingly, the degradation products were disulfide bonded dimers
`similar to those seen after long-term storage at atmospheric
`pressure and temperatures near room-temperature.17
`
`only about 20% of the volume change required for complete
`
`Figure 1. Reaction coordinate for aggregation of interferon-
`at 32 8C.
`Interferon-g, a protein whose native state is a homo-
`dimer, unfolds and aggregates rapidly upon dissocia-
`tion into monomers. When the transition state is
`formed from the native state,
`the protein’s partial
`molar volume (magenta) decreases by 41 mL/mol, the
`partial molar solvent-exposed surface area increases
`(black) by 3.5 nm2/molecule, and the associated activa-
`tion energy Ea (blue) is 130 kJ/mol. In contrast, com-
`plete dissociation and unfolding of the native state
`decrease of 209 mL/mol, a partial molar solvent-
`dimer
`is associated with a partial molar volume
`exposed surface area increase of 12.7 nm2/molecule, a
`free energy change of 27.2 kJ/mol, and an enthalpy
`change DH of 460 kJ/mol.1
`
`Mechanisms of Protein Aggregation
`
`Proteins are highly susceptible to the formation of non-
`native aggregates and precipitates.20,21 Irreversible, non-native
`protein aggregation is a ubiquitous concern for biopharma-
`ceuticals and biotechnology products,22 because the biological
`activity of a protein in an aggregate is usually greatly reduced.
`More importantly, non-native protein aggregates can cause
`adverse reactions in patients, ranging from immune response
`to anaphylactic shock and even death.23–25 Adverse responses
`to aggregates of a given protein cannot be predicted, nor can
`the maximal level of aggregates that can be safely tolerated
`be determined without costly and time-consuming clinical tri-
`als.4 Unfortunately, the link between immunogenicity and pro-
`tein aggregates is often not discovered until side effects are
`noted, following either long-term administration or increases
`in the patient population size after
`the drug has been
`approved. Thus, it is essential during product development to
`design proteins and protein formulations that minimize protein
`aggregation.
`Protein aggregates generally exhibit secondary structures
`that are rich in b-sheet structures, and that are dramatically
`perturbed from the protein’s native secondary structure.26 Pro-
`tein aggregation rates depend strongly on protein conforma-
`tion,27 and even relatively small perturbations in protein struc-
`ture can be sufficient to form transition-state species on the
`aggregation pathway.18,19 The kinetics of protein aggregation
`are controlled by both the concentration and the reactivity of
`these partially unfolded, transient intermediate species. If we
`assume that the free energy change associated with partial
`unfolding to form an aggregation-competent transition state
`(DG*) is of the same order of magnitude as that for complete
`unfolding (20–60 kJ/mol), on average fewer than 1/10,000 of
`the protein molecules exist in the transition state, or about 3
`nM at typical formulation conditions. This creates an experi-
`mental challenge, because the concentrations of these transient
`species are too low for direct measurement.
`Although the properties of proteins in the aggregation-
`competent transition state cannot be accessed spectroscopi-
`cally, some insight into how formulation conditions affect
`aggregation rates can be gained by making two assumptions.
`The first assumption is that DG* and DGumf, are positively
`correlated. Thus, measurements of excipient effects on DGumf,
`which can be made using various calorimetric and spectro-
`scopic techniques.28 should allow at least a qualitative predic-
`tion of excipient effects on DG*, and excipients that stabilize
`the native state against unfolding and increase DGumf, should
`also reduce the equilibrium concentrations of partially
`unfolded aggregation-competent species. A second assump-
`tion is that the dominant protein-protein interactions between
`native proteins are similar to those between protein molecules
`in the transition state. Protein-protein interactions can be
`quantified by measurement of the osmotic second virial coeffi-
`cient B22.14 Large, positive B22 values reflect net repulsive
`interactions between protein molecules. Formulation strat-
`egies may, thus, be designed so as to reduce protein aggrega-
`tion by adding using solution conditions (e.g., pH), and exci-
`pients that increase DGumf and/or B22 values.14 Examples of
`protein formulations that have been stabilized by addition of
`agents that increased DGumf include stabilization of acidic
`fibroblast growth factor,29 and recombinant keratinocyte
`
`1904
`
`DOI 10.1002/aic
`
`Published on behalf of the AIChE
`
`August 2007 Vol. 53, No. 8
`
`AIChE Journal
`
`Ex. 2005-0003
`
`

`
`growth factor30 by polyanionic excipients, stabilization of
`recombinant human growth hormone and recombinant human
`nerve growth factor by addition of zinc,31 and stabilization of
`recombinant human interferon-g by addition of sucrose.27
`Commercial formulations of recombinant human granulocyte
`colony stimulating factor, in contrast, minimize aggregation
`by adopting the strategy of maximizing protein-protein repul-
`sive interactions by formulating at low pH, where B22 values
`are large and positive.14
`
`Formulations at High-Protein
`Concentrations
`
`Many of the first generation of recombinant protein ther-
`erythropoietin and interferon-g,
`apeutics,
`such as
`are
`extremely potent molecules
`that
`required only minute
`amounts of protein per dose. For example, erythropoietin is
`administered in a dosage form containing about 60 mg/mL
`erythropoietin. In contrast, newer, antibody-based products
`are less potent, and, hence, require much larger doses. For
`1
`example, the monoclonal antibody Herceptin
`is sold in a
`vial containing 440 mg protein. The requirement for nearly
`10,000-fold increases in protein dosages, combined with
`practical limitations on the volume (<1.5 mL) that can be
`delivered in a single subcutaneous dose has led to the need
`to develop formulations that are highly concentrated in
`protein.
`Development of these formulations poses a number of seri-
`ous obstacles to commercialization.32,33 Although protein con-
`centrations rarely exceed 10 mM, even in highly concentrated
`formulations, due to the relatively large molecular weight of
`proteins,
`this may represent a substantial volume fraction
`(10–15%) of the formulation. Solution nonidealities caused
`by protein-protein interactions in these solutions may result
`in undesirably high-solution viscosities,32 opalescence,34 and
`increased rates of aggregation.35 High-viscosities are problem-
`atic because they can make manufacturing operations, such as
`filtration impractical, or limit the ability to deliver doses via
`narrow-bore syringe needles. Opalescence, although not nec-
`essarily harmful in itself, compromises the ability to detect
`product aggregation or particulate contamination within a vial,
`and creates difficulties during clinical trials, because of the
`lack of availability of opalescent placebo solutions. Most com-
`mon analytical techniques used to examine protein-protein
`interactions have been developed for use with much lower
`protein concentrations, and so current understanding of the
`interactions that cause high-viscosities or opalescence concen-
`tration is limited in part by the lack of appropriate analytical
`technologies.
`In recent unpublished studies, we have shown that sim-
`ple Carnahan-Starling hard-sphere models36 of protein ac-
`tivity coefficients accurately predict the protein-concentra-
`tion dependence of apparent rate constants for dimerization
`of recombinant human interleukin-1 receptor antagonist. In
`contrast, the effect of protein concentration on solution vis-
`cosities for the same protein are poorly predicted from
`hard-sphere models. A detailed understanding of the pro-
`tein-protein interactions that cause viscosity to vary from
`those predicted form hard-sphere models is not available at
`this time.
`
`Heterogeneous Nucleation during
`Processing and in Final
`Product Containers
`
`Even under solution conditions where protein physical sta-
`bility appears to be optimized—so as to minimize protein
`aggregation in the bulk solution—there can be formation of
`visible and subvisible protein particles that may constitute
`only a minute fraction of the total protein population. The
`presence of even a small number of protein particles can
`render a product clinically unacceptable. Particle formation
`can occur routinely during processing steps, such as pumping
`of protein solution during vial/syringe filling. In other cases,
`particle formation may appear to be random. For example,
`particles will be seen in a small fraction of vials or prefilled
`syringes in a given product lot. Unfortunately, these particles
`formed during vial filling operations appear downstream of
`sterile filtration steps and practically cannot be removed by fil-
`tration.
`We hypothesize that protein particle formation can arise
`from heterogeneous nucleation of protein aggregation on
`the surfaces of microparticles of foreign materials. These
`particulate contaminants can include metals shed from vial
`filling pumps, tungsten microparticles produced during man-
`ufacture of glass syringes, and glass microparticles shed
`from vials as a result of high-temperature depyrogenation
`procedures. Although such particles and the protein aggre-
`gates that we hypothesize result from them are ubiquitous,
`virtually no systematic characterization of the problem, and
`the mechanisms governing it have been addressed in the
`literature. Aggregation at microparticle surfaces has been
`studied only for a limited number of surfaces and proteins,
`and under an even more limited range of solution condi-
`tions, with few of the tested conditions being relevant for
`parenteral formulations of therapeutics. Furthermore, only
`two published studies have focused on therapeutic pro-
`teins,37,38 and none have focused on monoclonal antibodies,
`which are the largest class of therapeutic products currently
`being tested clinically.
`It should be noted that for a commercial pharmaceutical
`product, usually it is not economically practical to eliminate
`the risk of heterogeneous nucleation by re-engineering the sur-
`face properties of containers, pumps or tubing to completely
`eliminate shedding of particles, or by re-engineering the pro-
`tein to reduce interactions with a surface. In fact in a recent
`review chapter Akers and Nail state, ‘‘Regardless of the qual-
`ity of glass, the reputation of the manufacturer, the method of
`manufacture, or the method of cleaning, glass particulates are
`unavoidable.’’39 Thus, development of a safe, effective formu-
`lation of a protein therapeutic depends on determining solu-
`tion conditions that prevent the interactions of proteins with
`microparticulate contaminants that nucleate formation of pro-
`tein particles, while still maintaining protein stability in bulk
`solution. However, current commercial formulation develop-
`ment8 does not include testing for heterogeneous nucleation
`during processes, such as vial filling. This problem is not triv-
`ial. For example, relatively high-concentrations of nonionic
`surfactant may reduce protein binding to contaminants, but
`could also foster unacceptably rapid aggregation of the protein
`in bulk solution.
`
`AIChE Journal
`
`August 2007 Vol. 53, No. 8
`
`Published on behalf of the AIChE
`
`DOI 10.1002/aic
`
`1905
`
`Ex. 2005-0004
`
`

`
`Numerous studies have investigated the effects of surface
`chemistry on protein adsorption (see, for example40, and refer-
`ences therein). A major motivation of these earlier studies has
`been trying to understand the roles of protein interactions with
`surfaces associated with implantable medical devices. Conse-
`quently, the solution conditions for these investigations have
`generally been chosen to mimic physiological conditions, e.g.,
`phosphate-buffered saline, and the surfaces tested have often
`been those characteristic of implantable devices or natural sur-
`faces found in vivo, such as bone. In contrast, for formulations
`of therapeutic proteins, solution conditions are typically cho-
`sen that are not physiological, but rather optimized to provide
`long-term storage stability to the protein. For example,
`Amgen’s recombinant human granulocyte colony stimulating
`factor product is formulated in HCl, pH 4.0, a solution condi-
`tion that provides two-year shelf-life for the protein (in con-
`trast, the protein forms aggregates within a week if stored in
`phosphate-buffered saline at pH 7). Thus, the effects of solu-
`tion conditions (especially, the role of pharmaceutical exci-
`pients required for control of tonicity, antimicrobial activity,
`or protein stability in bulk solutions) on protein interactions
`with foreign microparticles have received limited attention in
`the literature.
`In previous studies of protein aggregation in solution, we
`found that the rate of aggregation can be manipulated by alter-
`ing solution conditions to modify protein conformational and
`colloidal stability.14 However, for cases where heterogeneous
`nucleation is operative, it is unclear whether control of these
`factors is sufficient to prevent protein aggregation. For example,
`in our studies of protein recombinant human platelet activating
`factor acetylhydrolyase, we observed significant protein particle
`formation, even in formulations that conferred both conforma-
`tional and colloidal stabilities.37 We traced the cause of particle
`formation to the presence of small numbers of glass micropar-
`ticles that were present in the drug product vials, presumably
`created during commercial depyrogenation procedures.
`
`Conclusions
`
`The remarkable advances in proteomics, development of
`fully humanized monoclonal antibodies and rapid drug candi-
`date screening have led to a large number of proteins that are
`under development as possible therapeutics. Development of
`stable, pharmaceutically acceptable formulations now poses a
`bottleneck that must be addressed if we are to take full
`advantage of these remarkable new drug candidates. The cou-
`pling of conformational equilibria with reaction kinetics,
`under solution conditions that tax existing analytical techni-
`ques will provide chemical engineers and pharmaceutical sci-
`entists with challenges for some time to come.
`
`Literature Cited
`
`1. Manning, M, Patel K, Borchardt R. Stability of protein
`pharmaceuticals. Pharmaceutical Res. 1989;6:903–918.
`2. Krishnan S, Chi EY, Wood SJ, et al. Oxidative dimer
`formation is the critical rate-limiting step for Parkin-
`son’s disease alpha-synuclein fibrillogenesis. Biochem.
`2003;42(3):829–837.
`3. DePaz RA, Barnett CC, Dale DA, Carpenter JF, Gaert-
`ner AL, Randolph TW. The excluding effects of sucrose
`
`on a protein chemical degradation pathway: Methionine
`oxidation in subtilisin. Archives of Biochem and Bio-
`phys. 2000;384(1):123–132.
`4. Chang BS, Hershenson S. Practical approaches to pro-
`tein formulation development. In: Carpenter JF, Man-
`ning MC, eds. Rational Design of Stable Protein Formu-
`lations. Kluwer: New York; 2002:1–25.
`Jaenicke, R., Protein Stability and Protein Folding. Ciba
`Foundation Symposia. 1991;161:206–221.
`6. Gazit E. The ‘‘Correctly folded’’ state of proteins: Is it a
`metastable state. Angewandte Chemie-Intl Ed. 2001;41
`(2):257-þ.
`
`5.
`
`7. Onuchic JN, LutheySchulten Z, Wolynes PG. Theory of
`protein folding: The energy landscape perspective.
`Annual Rev of Phys Chem. 1997;48:545–600.
`8. Carpenter JF, Pikal MJ, Chang BS, Randolph TW.
`Rational design of Stable Lyophilized Protein Formula-
`tions: Some Practical Advice. Pharmaceutical Res.
`1997;14(8):969–975.
`9. Roberts CJ. Kinetics of irreversible protein aggregation:
`Analysis of extended Lumry-Eyring models and impli-
`cations for predicting protein shelf life. J of Phys Chem
`B. 2003;107(5):1194–1207.
`10. Remmele RL, Jr, Nightlinger NS, Srinivasan S, Gombotz
`WR. Interleukin-1 receptor (IL-1R) liquid formulation
`development using differential scanning calorimetry.
`Pharmaceutical Res. 1998;15(2):200–8.
`11. Roy S, Katayama D, Dong AC, Kerwin BA, Randolph
`TW, Carpenter JF. Temperature dependence of benzyl
`alcohol- and 8-anilinonaphthalene-1-sulfonate-induced
`aggregation of recombinant human interleukin-1 recep-
`tor antagonist. Biochem. 2006;45(12):3898–3911.
`12. Wyma J, Jr, Gill SJ. Binding and Linkage. Functional
`Chemistry of Biological Macromolecules. Mill Valley,
`CA:University Science Books; 1990.
`13. Krishnan S, Chi EY, Webb JN, et al. Aggregation of
`granulocyte colony stimulating factor under physiologi-
`cal conditions: Characterization and thermodynamic in-
`hibition. Biochem. 2002;41(20):6422–6431.
`14. Chi EY, Krishnan S, Kendrick BS, Chang BS, Carpenter
`JF, Randolph TW. Roles of conformational stability and
`colloidal stability in the aggregation of recombinant
`human granulocyte colony-stimulating factor. Protein
`Sci. 2003;12(5):903–913.
`15. Chi EY, Krishnan S, Randolph TW, Carpenter JF. Phys-
`ical stability of proteins in aqueous solution: Mechanism
`and driving forces in nonnative protein aggregation.
`Pharmaceutical Res. 2003;20(9):1325–1336.
`16. Chang BS, Beauvais RM, Arakawa T, et al. Formation
`of an active dimer during storage of interleukin-1 recep-
`tor antagonist in aqueous solution. Biophysical J. 1996;
`71(6):3399–3406.
`17. Seefeldt MB, Kim YS, Tolley KP, Seely J, Carpenter
`JF, Randolph TW. High-pressure studies of aggregation
`of recombinant human interleukin-1 receptor antagonist:
`Thermodynamics, kinetics, and application to acceler-
`ated formulation studies. Protein Sci. 2005;14(9):2258–
`2266.
`18. Webb JN, Webb SD, Cleland JL, Carpenter JF, Ran-
`dolph TW. Partial Molar Volume, Surface Area, and
`Hydration Changes for Equilibrium olding and Forma-
`
`1906
`
`DOI 10.1002/aic
`
`Published on behalf of the AIChE
`
`August 2007 Vol. 53, No. 8
`
`AIChE Journal
`
`Ex. 2005-0005
`
`

`
`tion of Aggregation Transition State: High-Pressure and
`Co-Solute Studies on Recombinant Human Interferon-
`gamma. Proceedings of the National Academy of Scien-
`ces. 2001;98(13):7259–7264.
`19. Kendrick BS, Carpenter JF, Cleland JL, Randolph TW.
`A transient expansion of the native state precedes aggre-
`gation of recombinant human interferon-g. Proceedings
`of the National Academy of the Sciences USA. 1998;95:
`14142–14146.
`20. Cleland JL. Impact of Protein Folding on Biotechnol-
`ogy, in Protein Folding. In Vivo and In Vitro. Cleland
`JL, ed. Washington, DC: American Chemical Society;
`1993:1–21.
`21. Carpenter JF, Kendrick BS, Chang BS, Manning MC,
`Randolph TW. Inhibition of stress-induced aggregation
`of protein therapeutics. Methods in Enzymology. 1999;
`309:236–55.
`22. Cleland JL, Langer R.Formulation and Delivery of Pro-
`teins and Peptides. In: JL Ca, Langer R, eds. ACS Sym-
`posium Series. American Chemical Society: Washing-
`ton, DC; 1994.
`23. Moore WV, Leppert P. Role of aggregated human
`growth-hormone (Hgh) in development of antibodies
`to Hgh. J of Clinical Endocrinology and Metabolism,
`1980;51(4):691–697