Rational Design of Stable
`Protein Formulations
`Theory and Practice
`
`Edited by
`J ohn F. Carpenter
`and
`Mark C. Manning
`University of Colorado Health Sciences Center
`Denver, Colorado
`
`Springer Science+Business Media, LLC
`
`Celltrion Exhibit 1032
`Page 1
`
`

`

`Library of Congress Cataloging-in-Publication Data
`
`Rational design of stable protein formulations: theory and
`practice/edited by John F. Carpenter, Mark C. Manning.
`p.
`cm. -
`(Pharmaceutical biotechnology; v. 13)
`lncludes bibliographical references and index.
`ISBN 978-1-4613-5131-3
`ISBN 978-1-4615-0557-0 (eBook)
`DOI 10.1007/978-1-4615-0557-0
`l. Protein drugs-Stability. 2. Protein engineering. 3.
`Drugs-Design. 1. Carpenter, John F.
`II. Manning, Mark C.
`Series.
`RS43l.P75 R38 2002
`615' .l9-dc21
`
`III.
`
`2001057997
`
`ISBN 978-1-4613-5131-3
`
`© 2002 Springer Science+Business Media New York
`Originally published by Kluwer Academic I Plenum Publishers, New York in 2002
`Softcover reprint of the hardcover 1 st edition 2002
`
`AII rights reserved
`
`No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form
`or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise,
`without written permission from the Publisher
`
`Celltrion Exhibit 1032
`Page 2
`
`

`

`7
`
`Surfactant-Protein Interactions
`
`Theodore W. Randolph1,2 and LaToya S. Jones 1,3
`
`INTRODUCTION
`
`To retain biological activity, proteins generally must be maintained in a specific,
`three-dimensional conformation. This conformation is only marginally stable, and
`thus relatively minor perturbing forces can disrupt protein structure, causing loss
`of biological activity, as well as formation of non-native protein aggregates. Such
`perturbations are commonly encountered as proteins are produced, stored, trans(cid:173)
`ported, and delivered to patients. For example, it is well known that during
`common industrial processes such as filtering (Maa and Hsu, 1998), storage
`(Mcleod et aI., 2000), agitation (Thurow and Geisen, 1984; Maa and Hsu, 1997)
`freeze/thawing (Eckhardt, Oeswein et aI., 1991; Nema and Avis, 1993; Izutsu
`et aI., 1994), lyophilization (Carpenter and Chang, 1996; Carpenter et aI., 1997),
`nebulization (Ip et aI., 1995) and spray-drying (Broadhead et aI., 1994; Mumen(cid:173)
`thaler et aI., 1994; Maa et aI., 1998; Adler and Lee, 1999; Millqvist-Fureby,
`Malmsten et aI., 1999; Tzannis and Prestrelski, 1999) proteins can suffer damage
`to their native conformation. Further, delivery of protein pharmaceuticals to
`patients may also provoke losses of conformational integrity via unfavorable
`interactions of proteins with surfaces (e.g., inner walls of catheter tubing or
`syringes (Tzannis et aI., 1996».
`
`• Center for Pharmaceutical Biotechnology.
`Theodore W Randolph and LaToya S. Jones
`• Department of Chemical Engineering, University of Colorado,
`Theodore W Randolph
`• Department of Pharmaceutical Sciences, School of
`LaToya S. Jones
`Boulder, CO 80503.
`Pharmacy, University of Colorado Health Sciences Center, Denver, CO 80262.
`Rational Design of Stable Protein Formulations. edited by Carpenter and Manning. Kluwer Academic / Plenum
`Publishers. New York, 2002.
`
`159
`
`J. F. Carpenter et al. (eds.), Rational Design of Stable Protein Formulations
`© Kluwer Academic/Plenum Publishers, New York 2002
`
`Celltrion Exhibit 1032
`Page 3
`
`

`

`160
`
`Theodore W. Randolph and LaToya S. Jones
`
`The mechanisms of degradation of protein structure and activity are often
`categorized in two broad classes, chemical and physical. Chemical degradation
`refers to those modifications involving covalent bonds, such as deamidation,
`oxidation and disulfide bond shuffling. Physical degradation includes unfolding
`of the protein, undesired adsorption of the protein to surfaces, and aggregation.
`The two categories are not completely independent of one another. For example,
`protein oxidation may result in a greater proclivity to aggregate, and the rate of
`non-native disulfide bond formation may be higher in aggregated proteins.
`Surface-active agents, or surfactants, are often added to protein solutions to
`prevent physical damage during purification, filtration, transportation, freeze(cid:173)
`drying, spray-drying and storage. Surfactants are amphiphilic, containing a polar
`head group and a non-polar tail. This dual nature causes surfactants to adapt spe(cid:173)
`cific orientations at interfaces and in aqueous solutions. It is this characteristic
`that lies at the root of the mechanisms by which surfactants affect the physical
`stability of proteins.
`A well-known example is the anionic surfactant sodium dodecyl sulfate, or
`SDS. The sulfate anion is the hydrophilic head group of SDS, while the long
`aliphatic dodecyl chain forms the tail group. Ionic surfactants such as SDS have
`been known since the late 1930's as effective protein denaturants (Anson, 1939),
`and are commonly used for this purpose, e.g., as a pre-treatment for proteins in
`polyacrylamide gel electrophoresis (SDS-PAGE). In contrast, surfactants used as
`stabilizing agents in protein formulations are typically non-ionic (Loughheed et
`aI., 1983; Twardowski et aI., 1983; Chawla et aI., 1985). This chapter will focus
`on non-ionic surfactants; protein interactions with ionic surfactants have been
`reviewed elsewhere (Jones, 1996). An example non-ionic surfactant is poly(cid:173)
`oxyethylene sorbitan monolaurate (Tween 20®), shown in Figure 1. In this
`molecule, the hydrophilic polyoxyethylene units form the head group, while the
`hydrophobic monolaurate group is the tail. Tween 20 is often added to formula(cid:173)
`tions due to its ability to protect proteins from surface-induced denaturation
`(Chang et aI., 1996; Jones et aI., 1997; Bam et aI., 1998; Kreilgaard et aI., 1998;
`Maa et aI., 1998).
`There are a number of mechanisms by which surfactants can prevent or
`promote damage to proteins. Some of these mechanisms are generic to all
`excipients, and can be explained in the solution thermodynamic framework of the
`Wyman linkage theory (Wyman and Gill, 1990) and the preferential exclusion
`mechanisms developed by Timasheff and colleagues (Arakawa and Timasheff,
`1982, 1983, 1984a,b, 1985a,b,c; Arakawa et aI., 1990; Timasheff, 1998). Others
`derive from the amphiphilicity of surfactants and the resulting effect of micro(cid:173)
`scopic ordering of surfactant molecules at interfaces, which in tum affects the
`kinetics and thermodynamics of protein interfaces. In this chapter, we discuss a
`number of these mechanisms and their implications for protein stability.
`
`Celltrion Exhibit 1032
`Page 4
`
`

`

`Surfactant-Protein Interactions
`
`161
`
`Polyethylene glycol ether
`Triton X-IOO, x=9-10 (average)
`Triton X-114, x=7-8 (average)
`
`w+x+y+z=20
`Polysorbate
`Tween 20, R=C IlH23C02
`Tween 80, R=C 17H33C02
`
`Figure 1. Example non-ionic surfactants.
`
`PROTEINS AND SURFACTANTS AT SURFACES
`
`Because of their dual hydrophobic/hydrophilic nature, surfactants in solu(cid:173)
`tion tend to orient themselves so that the exposure of the hydrophobic portion of
`the surfactant to the aqueous solution is minimized. Thus, in systems containing
`air/water interfaces, surfactants will tend to accumulate at these interfaces,
`forming a surface layer of surfactant oriented in such a fashion that only their
`hydrophilic ends are exposed to water. Such orientation and surface adsorption
`can also occur at solid/water interfaces such as those found in vials, syringes, and
`other containers. Protein molecules also exhibit surface activity (for a review see
`(Magdassi, 1996), and references therein) and as such will also tend to adsorb to
`and orient at these interfaces.
`From classical thermodynamics, the excess surface internal energy dU{ of
`a surface with area A at a temperature T is related to the excess surface entropy
`
`Celltrion Exhibit 1032
`Page 5
`
`

`

`162
`
`Theodore W. Randolph and LaToya S. Jones
`
`Sf and the chemical potential and number of surface excess moles of each
`adsorbed species:
`
`dU{J = TdS{J + adA + I, Ilidn~l
`
`c
`
`i~2
`
`(1)
`
`Here the subscript 1 refers to a dividing surface chosen so that there is no excess
`adsorption of species 1, the solvent (water) (Gibbs, 1961), and a is the surface
`tension. The equilibrium criterion, oS = 0, requires that a be constant across the
`surface. Thus, if a protein adsorbs to an interface, at equilibrium the surface
`tension forces must be continuous and constant across the whole interface, includ(cid:173)
`ing across the protein. The stability criterion at equilibrium requires that:
`
`( da)
`dA Tn>
`
`0
`
`(2)
`
`If the surface tension of the interface is greater than the internal tension in the
`protein, then in order to meet these two conditions, the surface area of the protein
`must increase until the two tensions are equal, i.e., the protein must unfold. In
`some cases, nearly complete loss of native activity is lost upon adsorbing to
`the interface (Rothen, 1947; Verger et aI., 1973). The Gibbs adsorption equation
`relates the surface tension to the concentration of adsorbed species at an
`interface:
`
`c
`-da = sfdT+ I,C.ldlli
`i=2
`
`(3)
`
`where C,l and sfdT are, respectively, the excess surface adsorption and excess
`surface entropy of component i, both relative to a dividing surface with no surface
`excess of solvent (1), and Ili is the chemical potential of species i. Adsorption of
`protein to the interface thus lowers the interfacial tension, making unfolding less
`likely as adsorption progresses.
`If the process of surface adsorption and unfolding of protein were to stop
`after the formation of an eqUilibrium monolayer, the amount of adsorbed protein
`would be so small as to be generally of no consequence. Indeed, a significant
`amount of work has been dedicated to development of methods with sensitivities
`high enough to characterize the minute amount of protein adsorbed to the inter(cid:173)
`face (Tupy et aI., 1998; Vermeer and Norde, 2000). However, depending on the
`degree of surface hydrophobicity and characteristics of the protein in question,
`additional processes can occur in the adsorbed films, leading to behavior that is
`
`Celltrion Exhibit 1032
`Page 6
`
`

`

`Surfactant-Protein Interactions
`
`163
`
`no longer described by the Gibbs' adsorption equation. Among the possible
`processes are gas-to-liquid surface phase transitions, surface precipitation, and
`the formation of surface sublayers. In many cases, the adsorbed molecules can
`be rapidly released from the surface and surface sublayers, and may exchange
`with bulk protein molecules. Alternatively, because of slow refolding kinetics,
`proteins can become irreversibly adsorbed at the interface, or have a rate
`of exchange somewhere between these extremes (Dickinson, 1999; Norde and
`Giacomelli, 1999). These processes of adsorption and release of structurally(cid:173)
`perturbed protein molecules into the bulk solution have been implicated as one
`of the causes of protein aggregation and denaturation.
`When discussing protein adsorption at interfaces, globular proteins are
`typically characterized as being either "hard" or "soft," having a low or high
`degree of flexibility, respectively, and by their degrees of hydrophobicity. Soft,
`hydrophobic proteins, attaining monolayer coverage of the air/water interface in
`the matter of minutes, are generally more surface reactive at hydrophobic sur(cid:173)
`faces than hard, hydrophilic proteins, attaining coverage of the same surfaces in
`a matter of hours (Tripp et aI., 1995). The driving force for protein adsorption is
`the decrease in the entropy of the water molecules that are ordered around the
`hydrophobic protein domains when the protein is in the bulk solution. Thus, the
`role of the relative degree of hydrophobicity on protein surface adsorption is
`rather straightforward: given two proteins only differing in their hydrophobici(cid:173)
`ties, the more hydrophobic protein will have a greater number of productive
`interactions with the surface and will form the monolayer more quickly.
`Middelberg et al. (2000) proposed that the difference in adsorption kinetics of
`Lac21 and Lac28 peptides is because the monomeric Lac21 has more hydropho(cid:173)
`bic residues exposed than its tetratmeric counterpart, Lac28; thus, Lac21 more
`readily forms a monolayer at an octane-water interface. Protein flexibility
`is important in protein spreading that occurs at the interface. A flexible protein
`can expose additional non-polar residues, leading to an increased strength in
`binding to the surface. Finally, the protein flexibility dictates the number of
`proteins that can adsorb at the interface, and their spreading rate (N orde and
`Giacomelli, 1999).
`Protein adsorption to a hydrophobic surface does not necessarily lead to a
`complete loss of "native" structure: some proteins actually gain structure. For
`example, melittin, a honeybee venom peptide, increases its a-helical content
`slightly when adsorbed to hydrophobic quartz. In contrast, adsorption of a
`tetramer of the same peptide is thought to require a loosening of the a-helical
`content. The orientation of the adsorbed helices in the peptide is parallel to the
`quartz plane, with the hydrophobic moieties facing the plane (Smith and Clark,
`1992). Caessens et al. (1999) also report an increase in helical content on the
`adsorption of the predominately random coil p-casein and p-casein peptides at
`the teflon/water interface.
`
`Celltrion Exhibit 1032
`Page 7
`
`

`

`164
`
`Theodore W. Randolph and LaToya S. Jones
`
`Surfactants will also adsorb to the interface. The Gibbs adsorption isotherm
`again predicts that this interaction will lower the surface tension. Thus, if sur(cid:173)
`factants co-adsorb to an interface together with proteins, there will be a smaller
`driving force for surface adsorption. Furthermore, competition between the
`protein and the surfactant for the interface may reduce equilibrium protein
`adsorption. For example, Tween 20 addition displaces beta-lactoglobulin films
`from air-water interfaces (Roth et aI., 2000). This effect is not universal: the ionic
`surfactant sodium dodecyl sulfate forms a complex with high surface activity that
`exhibits enhanced surface adsorption (Green et aI., 2000). Finally, because the
`surface tension is lowered after surfactant is adsorbed, less damage to the protein
`that does adsorbed may occur. For example, a loss of a-helix and an increase in
`p-tum structures occur when bovine serum albumin adsorbs to polystyrene par(cid:173)
`ticles. This effect is decreased when the surface is more crowded (Norde and
`Giacomelli, 2000).
`Protein stabilization by nonionic surfactants can often be observed by for(cid:173)
`mulating with micromolar concentrations of surfactant. This is due to the high
`surface-activity of this class of excipients, which renders a higher effective con(cid:173)
`centration of surfactant molecules at interfaces than in the bulk solution. When
`the concentration of the surfactant is much lower than its critical micelle con(cid:173)
`centration (CMC), surfactant molecules lie flat at the air/water and hydrophobic
`solid/water interfaces (Figure 2A) (Porter, 1994). As the concentration is
`increased, more molecules adsorb to these interfaces, such that the surface con(cid:173)
`centration remains linearly proportional to the bulk concentration (Tanford,
`1973). This crowding forces the surfactant molecules to order themselves such
`that the hydrophilic groups are oriented towards the bulk water and the hydro(cid:173)
`carbon chains are pointed towards the air or hydrophobic solid (Porter, 1994;
`Fainerman et aI., 2000) (Figure 2B). At sufficiently high surfactant concentra(cid:173)
`tions (i.e., at or above the CMC), there is an oriented monolayer of surfactant
`molecules and maximum surfactant absorption, at the interface (Figures 2C &
`D). The surface saturation is responsible for the sharp slope change (to essen(cid:173)
`tially zero) observed in experimental plots of surface properties (surface tension,
`osmotic pressure) versus surfactant concentration (Porter, 1994) (Figure 3). It
`should be noted that the linear variations in the surface properties shown in Figure
`3 are indicative that the surface activity of a surfactant is due to the hydropho(cid:173)
`bic effect: the variation would be cooperative if hydrocarbon self-affinity was the
`appropriate explanation (Tanford, 1973). Surfactant micelles are formed in the
`bulk phase when the concentration is of the surfactant is above the CMC (Figure
`2D). Thus, a range of surfactant concentrations could inhibit protein denaturation
`at an interface; however, the necessity of CMC levels of surfactant to completely
`inhibit protein damage would be a strong indication that the damage is caused by
`adsorption of protein at the interface and that inhibiting protein adsorption at the
`interface plays a role in preventing protein aggregation.
`
`Celltrion Exhibit 1032
`Page 8
`
`

`

`Surfactant-Protein Interactions
`
`165
`
`A
`
`B
`
`C
`
`~ c::P ~ ~!
`c:::P
`water
`~~~~ TITITIffiJffiJTI
`- hyd~(d)i~70i1:a / 7 7 7 7
`
`~
`
`/ 7 7 7 7
`
`TITITITITITIOO
`/ 7 7 7 7
`
`f£
`·····_·· .. ······ .. ··air··==· .. ·· .. ······ .... ··~··~~·~········ .. · .... ··· .... g·ggOO·ggg· ... ·· .... ··· .... ~
`
`water ~ ./
`
`Figure 2. Simplified models of the interfacial behavior of a nonionic surfactant at several concen(cid:173)
`trations in water. Circles-polar head groups (e.g., polyoxyethelenes). Rectangles-hydrophobic tails
`(e.g., hydrocarbon chains). Models above the dotted line is for the hydrophobic solid (ice)/water inter(cid:173)
`face and those above are for the air/water interface. (A) Surfactant concentration is well below the
`CMC (B) Surfactant concentration is greater than in A, but still below the CMC. (C) Surfactant con(cid:173)
`centration is at the CMC. (D) The surfactant concentration is above the CMC. (This model is adapted
`from Figures 4.4 and 4.8 of Porter (Porter, 1994)).
`
`Surface tension
`Osmotic pressure
`
`concentration
`
`Figure 3. Representation of changes two properties in determining the CMC of a surfactant.
`
`Decreased protein adsorption at interfaces (e.g., airlliquid, ice/liquid) in the
`presence of nonionic surfactants such as Tween 20 can be attributed to the surface
`activity ofnonionic surfactants (Chang et aI., 1996; Kreilgaard et aI., 1998; Miller
`et aI., 2000a,b) and, in some cases, direct interactions between the surfactant and
`protein molecules (Dickinson, 1998; Bam et aI., 1998; Miller et aI., 2000a,b). In
`
`Celltrion Exhibit 1032
`Page 9
`
`

`

`166
`
`Theodore W. Randolph and LaToya S. Jones
`
`mixed protein/surfactant systems in which the surfactant binds to hydrophobic
`regions of the protein, the protein is less surface-active than it would be in a solu(cid:173)
`tion devoid of the nonionic surfactant. This explains the increase in surface
`tension relative to that of the pure protein solution that can be observed at
`extremely low surfactant concentrations. Adsorption of surfactant and protein
`molecules in the mixed system is competitive. Nonionic surfactants usually bind
`tighter than proteins or protein-surfactant complexes at interfaces (Dickinson,
`1998). Thus, above a critical concentration of the surfactant, protein adsorption
`becomes negligible and the adsorption isotherms for mixed surfactant/protein
`systems can be roughly identical that of a pure surfactant solution, as observed
`by Miller et aI. (2000a) for the HSAICIODMPO when the CIODMPO concentra(cid:173)
`tion exceeds 10-7 moVcm3 •
`
`PROTEIN-SURFACTANT INTERACTIONS IN SOLUTION
`
`In addition to altering the interaction of proteins with surfaces, non-ionic
`surfactants can also interact directly with proteins in solution. For example,
`Tween 20 acts as a chemical chaperone, aiding in the refolding of proteins (Bam
`et aI., 1998; Kreilgaard et aI., 1999). In vivo, proteins fold while at average con(cid:173)
`centrations of approximately 35 mg/ml (Hartl, 1996): in vitro, non-native protein
`molecules at this concentration (e.g., due to freeze concentration) usually aggre(cid:173)
`gate. Inside cells, protein folding is ·aided by naturally occurring molecular chap(cid:173)
`erones. Unlike folding catalysts that have steric information to guide the protein
`folding, molecular chaperones act by non-covalently binding to partially-folded
`proteins, usually via hydrophobic interactions, to prevent misfolding or aggrega(cid:173)
`tion while the protein is attempting to adopt its native conformation (Gatenby
`and Ellis, 1990; Hartl, 1996). Chaperone-assisted protein refolding can prevent
`the protein from falling into kinetic traps or simply allow more time for the
`protein to refold (Gatenby and Ellis, 1990). The hydrophobic effect, the driving
`force of protein folding and surface activity of Tween 20, is also implicated in
`the interactions between the exposed hydrophobic regions of the partially folded
`protein and the hydrocarbon tail of the surfactant. These mixed surfactant/protein
`complexes and protein folding are dynamic processes, eventually the protein
`attains a conformation in which the hydrophobic groups are not as surface
`exposed. Unless there are hydrophobic patches in the native protein conforma(cid:173)
`tion, the surfactant molecules will not necessarily bind to the protein, as in the
`case of molecular chaperones. In vitro, surfactants such as Tweens (Bam et aI.,
`1996), polyethylene glycol (Cleland and Wang, 1990; Cleland et aI., 1992;
`Cleland and Randolph, 1992; Cleland 1993; Cleland and Wang, 1993), Triton
`
`Celltrion Exhibit 1032
`Page 10
`
`

`

`Surfactant-Protein Interactions
`
`167
`
`X-lOO (Donate et aI., 1998), and lubrol (Donate et aI., 1998), have been impli(cid:173)
`cated in aiding protein refolding by acting as chemical chaperones.
`Surfactants, such as Tween 20, can also affect the thermodynamic confor(cid:173)
`mational stability of a protein. As discussed in other chapters in this book, and
`in detail in references from Timasheff and co-workers (Lee and Timasheff, 1981;
`Arakawa and Timasheff, 1982; Arakawa and Timasheff, 1985; Timasheff, 1998),
`thermodynamic stability is increased if a ligand exhibits greater binding to the
`native state of a protein than to a non-native state. However, with many excipi(cid:173)
`ents (e.g., sucrose) excluded volume effects produce a non-specific negative
`binding to the native state, and a concomitantly larger negative binding to
`expanded, non-native conformations. This differential negative binding also
`results in a stabilization of the native state. At the low concentrations of surfac(cid:173)
`tant (ca. 100 micromolar) typically used in formulations of therapeutic proteins,
`thermodynamic effects due to excluded volume can usually be neglected. More
`important is specific binding of surfactants to either the native or unfolded states
`of a protein. Randolph and colleagues report that some proteins and nonionic sur(cid:173)
`factants, including Tween 20, form mixed protein: detergent complexes (Bam et
`aI., 1995, 1996, 1998; Jones, et aI., 1999). In the presence of Tween 40 at a 4: 1
`surfactant: protein molar ratio, native recombinant human growth hormone is sig(cid:173)
`nificantly stabilized: the denaturation midpoint in guanidine hydrochloride solu(cid:173)
`tions increases from 4.6M guanidine hydrochloride in the absence of Tween 40
`to 5.9M in the presence of Tween 40 (Bam et aI., 1996). Likewise, 4: 1 Tween
`40 increases the ~G of unfolding of recombinant human growth hormone by
`4.1 kcal/mol (Bam et aI., 1996), and 10: 1 Tween 40 increases its melting point
`slightly from 88.8 to 89.4 0 C (Bam et aI., 1998). Stabilization of recombinant
`human growth hormone by surfactants results in reduced aggregation in agitated
`solutions (Bam et aI., 1998). With bovine serum albumin in the presence of sur(cid:173)
`factant there are decreased amounts of thermally-induced protein aggregates,
`relative to surfactant-free controls (Arakawa and Kita, 2000). Conformational sta(cid:173)
`bilization of proteins by non-ionic surfactants is not universal; stability of IgG is
`unaffected by low concentrations of Tween 20 (Vermeer and Norde, 2000), and
`recombinant human interferon-y shows lower free energies of unfolding in the
`presence of Tween 20 (Webb et aI., 2000).
`
`SURFACTANT EFFECTS ON PROTEIN ASSEMBLY STATE
`
`The hydrophobic portion of non-ionic surfactants can bind to hydrophobic
`patches on proteins. This naturally causes the surfactant to order itself so that
`more hydrophilic groups are solvent exposed, resulting in a "hydrophobicity
`
`Celltrion Exhibit 1032
`Page 11
`
`

`

`168
`
`Theodore W. Randolph and LaToya S. Jones
`
`reversal". This "hydrophobicity reversal" means that the protein-surfactant
`complex is more hydrophilic that either the surfactant or protein alone, and effec(cid:173)
`tively increases the solubility of the complex. This, in turn can reduce the pro(cid:173)
`pensity of the protein to form higher-order aggregates. For example, bovine
`mitochondrial cytochrome bel is dimeric in solutions at low ionic strength and
`low surfactant levels. At Tween 20 concentrations above 5 mg/mg protein, a
`homogeneous, monomeric, reversible and enzymatically active protein-surfactant
`complex is formed (Musatov and Robinson, 1994). Likewise, in freeze-thaw
`studies of recombinant factor XIII, addition of Tween 20 at concentrations near
`the CMC blocked the progression of aggregates from a relatively low molecular
`weight, soluble fraction to insoluble aggregates. Figure 4 shows levels of
`
`100
`t::
`]i 75
`e Cl.
`(J) 50
`>
`~
`t::
`~ 0
`
`25
`
`0
`
`o
`
`OJ
`
`en 30
`~
`~ 20
`OJ co
`(J)
`:0 10
`:>
`~
`*-
`
`0
`
`.~:~.~.:':':':':.i::-.. :...:..:. ... = .. :.:.:.:. ... = .. ~
`
`en
`2100
`co
`OJ
`~ 75
`OJ
`~ 50
`:0
`~ 25
`en
`~ 0
`
`o
`
`200
`150
`100
`50
`Tween 20 concentration [IlM]
`
`250
`
`Figure 4. Recovery of native rFXIII (A) and formation of soluble (B) and insoluble aggregates (C)
`following 10 freeze-thaw cycles of 1 mglml ( ___ ), 5 mglml ( ..... 0 .. · .. ) and 10 mg/ml rFXIII (-T-)
`as a function of Tween 20. Results are plotted as mean values +/- standard deviation for duplicate
`samples (reproduced from Kreilgaard et aI., 1998).
`
`Celltrion Exhibit 1032
`Page 12
`
`

`

`Surfactant-Protein Interactions
`
`169
`
`aggregation for factor XIII after freeze thaw cycling in the presence and absence
`of Tween 20. Note that Tween addition did not completely block aggregation,
`but was very effective at preventing the formation of insoluble aggregates
`(Kreilgaard et aI., 1998).
`Non-ionic surfactants can also have the opposite effect on protein assembly
`state. In cases where a non-ionic surfactant destabilizes the conformation of
`a protein, this effect may compete against the solubilizing effect of surfactant
`binding and hydrophobicity reversal. For example, Bax, a monomeric protein that
`regulates apoptosis, readily forms dimers in the presence of Tween 20. However,
`these dimers are apparently non-native, as they do not expose the characteristic
`N-terminal Bax epitope (Hsu and Youle, 1998). In the case of the hydrophobic
`lipase from Humicola lanuginose, Tween 20 addition caused the formation of
`large, insoluble non-native aggregates (Kreilgaard et aI., 1999).
`
`SURFACTANT EFFECTS ON PROTEINS DURING FREEZING,
`FREEZE-DRYING AND RECONSTITUTION
`
`The processes of freezing, drying, and reconstitution of protein solutions
`present a number of stresses that may denature proteins. Many of these stresses
`are associated with surfaces: new ice-water and ice-glassy solid interfaces are
`formed during freezing, drying replaces ice-glass interfaces with air-glass inter(cid:173)
`faces, and reconstitution exposes the glassy solid surfaces to aqueous solution.
`In each of these steps protein adsorption to surfaces is potentially damaging.
`The ice-water interface has been implicated as a source of damage to proteins
`(Strambini and Gabellieri, 1996), as has the solid-air interface (Hsu et aI., 1995).
`Addition of nonionic surfactants can reduce this damage, presumably by com(cid:173)
`peting with the protein for the ice-water interface (Chang, 1996). For example,
`addition of Tween 80 to solutions of recombinant hemoglobin reduced aggrega(cid:173)
`tion seen during freeze thaw studies (Kerwin et aI., 1998). Interestingly, Tween
`80 did not offer protection against methemoglobin formation or hemoglobin
`aggregation during long-term frozen storage.
`Non-ionic surfactants also have been shown to affect the recovery of native
`protein from lyophilized formulations. Sarciaux et al. (1999) showed that the
`addition of Tween 80 to the formulation solution or the reconstitution medium
`for lyophilized formulations resulted in reduced levels of aggregates. Likewise,
`Zhang et al. (1995, 1996) have demonstrated that, following long-term storage,
`surfactants in the reconstitution medium can affect protein recovery. We have
`recently shown (Webb et aI., 2000) that addition of Tween 20 to the reconstitu(cid:173)
`tion medium for lyophilized preparations of recombinant human interferon-y
`results in decreased levels of aggregates. The mechanism for such reduced
`
`Celltrion Exhibit 1032
`Page 13
`
`

`

`170
`
`Theodore W. Randolph and LaToya S. Jones
`
`aggregation was shown to be a surfactant effect on dissolution rates. Addition of
`Tween 20 slowed the dissolution of the lyophilized solid, allowing protein that
`partially unfolded during freeze-drying to refold before aggregating. Interest(cid:173)
`ingly, Tween 20 in aqueous solutions destabilizes recombinant human interferon(cid:173)
`y against urea-induced unfolding, impedes refolding during rapid dilution from
`urea solutions, and actually increases aggregation during agitation.
`
`ENZYMATIC DEGRADATION OF NON-IONIC SURFACTANTS
`
`Although most non-ionic surfactants are thought of as chemically inert com(cid:173)
`ponents of a formulation, specific chemical interactions between proteins can
`occur. For example, some enzymes show hydrolytic activity toward Tweens.
`Smegmatocin, an esterase from Mycobacterium smegmatis, shows a broad
`thermal and pH stability in its activity against Tween 80 (Tomioka, 1983). The
`byproduct of Tween 80 hydrolysis, oleic acid, is toxic to some bacteria. Similar
`bacteriocins, which require Tween their expression, have been ascribed to other
`mycobacteria (Saito et aI., 1983). It is not clear how widespread esterase activ(cid:173)
`ity is against Tweens. However, it is clear that caution should be used when for(cid:173)
`mulating proteins with esterase activities in Tween solutions.
`
`RECOMMENDATIONS FOR PROTEIN FORMULATION
`
`Clearly, it is desirable to minimize the addition of any excipient to a for(cid:173)
`mulation. This rule of thumb is even more pertinent for surfactants, because there
`is ample evidence that high concentrations of surfactants can be destabilizing to
`protein structure. On the other hand, small amounts of surfactant often provide
`benefits in preventing aggregation that greatly outweigh any conformationally
`destabilizing effect. How then should surfactant levels be chosen for optimal for(cid:173)
`mulation? The answer appears to depend on the mechanism(s) by which a par(cid:173)
`ticular protein is protected from damage by surfactant addition. In cases where
`surfactants act to stabilize the native state of a protein by binding to the protein,
`a specific surfactant: protein stoichiometry may need to be maintained in order
`to provide optimal protection. In these cases, changes in the protein concentra(cid:173)
`tion within a formulation will dictate proportional changes in surfactant concen(cid:173)
`tration to maintain a fixed molar ratio. This appears to be the case for recombinant
`human growth hormone, where protection against agitation-induced damage cor(cid:173)
`related with the molar ratio of surfactant to protein rather than to the surfactant's
`CMC (Bam et aI., 1998). In the case of specific binding, the choice of nonionic
`
`Celltrion Exhibit 1032
`Page 14
`
`

`

`Surfactant-Protein Interactions
`
`171
`
`surfactant may be important. In the case of recombinant human growth hormone,
`for example, different binding stoichiometries and degrees of protein stabiliza(cid:173)
`tion were seen for a variety of common surfactants (Bam et aI., 1995). A general
`recommendation for proteins that show specific binding to the native state of the
`protein is to formulate so that the ratio of surfactant to protein is slightly above
`the binding stoichiometry for a particular surfactant. The choice of surfactant may
`be dictated by the degree of stabilization (which should correlate with the degree
`of binding) provided to a protein