RESEARCH ARTICLE – Pharmaceutical Biotechnology

Competitive Adsorption of Monoclonal Antibodies and Nonionic Surfactants at Solid Hydrophobic Surfaces 3,4 ´ SEBASTIAN J. KAPP,1 IBEN LARSSON,2 MARCO VAN DE WEERT,1 MARITE´ CARDENAS, LENE JORGENSEN1 1

Department of Pharmacy, University of Copenhagen, Copenhagen 2100, Denmark Amgros I/S, Copenhagen 2400, Denmark 3 Department of Chemistry, University of Copenhagen, Copenhagen 2100, Denmark 4 Department of Biochemical Sciences, Malmo¨ University, Malmo¨ 20506, Sweden 2

Received 19 August 2014; revised 23 October 2014; accepted 23 October 2014 Published online 1 December 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24265 ABSTRACT: Two monoclonal antibodies from the IgG subclasses one and two were compared in their adsorption behavior with hydrophobic surfaces upon dilution to 10 mg/mL with 0.9% NaCl. These conditions simulate handling of the compounds at hospital pharmacies and surfaces encountered after preparation, such as infusion bags and i.v. lines. Total internal reflection fluorescence and quartz crystal microbalance with dissipation monitoring were used to follow and quantify this. Furthermore, the influence of the nonionic surfactant polysorbate 80 (PS80) on the adsorption process of these two antibodies was investigated. Despite belonging to two different IgG subclasses, both antibodies displayed comparable adsorption behavior. Both antibodies readily adsorbed in the absence of PS80, whereas adsorption was reduced in the presence of 30 mg/L surfactant. The sequence of exposure of the surfactant and protein to the surface was found to have a major influence on the extent of protein adsorption. Although only a fraction of adsorbed protein could be removed by rinsing with C 2014 Wiley Periodicals, Inc. 30 mg/L surfactant solution, adsorption was entirely prevented when surfaces were pre-exposed to PS80.  and the American Pharmacists Association J Pharm Sci 104:593–601, 2015 Keywords: adsorption; protein formulation; physical stability; surfactants; protein; monoclonal antibodies; QCM-D; TIRF

INTRODUCTION Monoclonal antibodies (mAbs) are widely used in the treatment of various diseases including inflammatory diseases and cancer, thus accounting for an expanding economic earning in the pharmaceutical field. For example, the group of cancer antibodies, which includes Panitumumab (PAN; Vectibix ) and Rituximab (MabThera ), yielded total annual sales of US$ 23.47 billion in 2012.1 However, despite intensive research on mAbs, there still remain challenges in terms of obtaining a sufficiently stable product that can be marketed. Antibodies highly depend on an appropriate threedimensional fold to achieve target binding. If this ordered structure is lost, efficacy is consequently lost. Protein adsorption to solid surfaces is one process by which proteins can unfold,2–5 and this can lead to subsequent aggregation.6 Furthermore, protein adsorption onto particulates7 and generation of subvisible protein aggregates because of exposure to a variety of surfaces have been associated with development of immunogenicity toward therapeutic proteins.8–10 Nonionic surfactants have been shown to prevent or significantly reduce this detrimental physical degradation11–15 and thus are frequently used in pharmaceutical formulations. A range of mAbs intended for clinical use are formulated with nonionic surfactants to increase their stability, for example, Rituximab, Trastuzumab, Bevacizumab, and Cetuximab. A study R

R

Correspondence to: Lene Jorgensen (Telephone: +45-3533-6369; Fax: +453533-6030; E-mail: [email protected]) This article contains supplementary material available from the authors upon request or via the Internet at http://onlinelibrary.wiley.com/. Journal of Pharmaceutical Sciences, Vol. 104, 593–601 (2015)  C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association

performed by Garidel et al.16 showed that the stabilizing properties of polysorbate on mAbs are not because of the formation of protein–surfactant complexes in solution. Hence, the stabilizing effect may be because of a molecular chaperone effect of polysorbate,17 preferential hydration of the protein as described for polyethyleneglycol,18,19 or reduced interfacial affinity of the protein because of blocking properties of the surfactant.20,21 The latter would result in reduced protein adsorption and unfolding on the surface. As surfactants are added at very low concentration, preferential exclusion effects are expected to be minimal,19 and thus surface–surfactant interactions represent one likely explanation for the reduced surface adsorption of proteins.16 The aim of the present study was to gain further insight into the blocking properties of polysorbate 80 (PS80) toward the adsorption of mAbs onto solid hydrophobic interface. Specifically, it is the intention to relate this data of model surfaces (water contact angle from 85◦ to 90◦ ) to polymers such as polyethylene, polypropylene, and polyvinylchloride used in the production of infusion bags (water contact angles from 87◦ to 104◦ ).22–25 To this purpose, the adsorption kinetics of two model mAbs in the presence and absence of the commonly used surfactant PS80 were studied. Both mAbs in this study, mAb-1 and PAN, are formulated without surfactant. PAN has the advantage of being commercially available and may thus be used as a potential reference material in further studies. The use of a reference antibody allows for comparability of future studies under a variety of conditions. Two different methods were used to follow the adsorption kinetics, quartz crystal microbalance with dissipation monitoring (QCM-D) and total internal reflection fluorescence (TIRF). Whereas TIRF measurements can be made specific for a single adsorbing species by only following the specific fluorescence of that compound, in our case the mAb, QCM-D Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

593

594

RESEARCH ARTICLE – Pharmaceutical Biotechnology

measurements report on the adsorption process as a whole, and yield insight into the viscoelastic properties of the adsorbed layer.26,27

EXPERIMENTAL Materials and Methods Sodium chloride, sulfuric acid (98%), hydrochloric acid (25%), ammonia solution (25%), and hydrogen peroxide (30%) were obtained from Merck (Darmstadt, Germany). Nacetyltryptophan, Tween 80 (PS80), 1-propanethiol, and (3, 3, 3-trifluoropropyl)chloromethylsilane were obtained from Sigma–Aldrich (Steinheim, Germany). All chemicals are of analytical grade and were used without further purification. Water purified to a resistivity of 18.2 Mcm−1 was used in all preparations. mAb1, an IgG1 (pI 8.44) formulated with 205 mM sucrose in a 25-mM histidine buffer at pH 6, was graciously donated by Medimmune (Cambridge, UK), PAN, an IgG2, was bought as the product Vectibix (expiry date: May 2015; Amgen, Thousand Oaks, CA). PAN, with a pI of 6.63,28 is formulated with 100 mM sodium chloride in a 83-mM sodium acetate buffer at pH 5.8.29 The CMC of PS80 is reported to be 14–15 mg/L in water,30 and this value is not expected to be affected by the presence of mAbs16 and only marginally lowered by salt.31–33 The terminology used in this article will be relative to the CMC of PS80 in water: 1/2xCMC (7 mg/L), 2xCMC (30 mg/L). R

R

Ultraviolet–visible A nanodrop UV/Vis spectrophotometer (Thermo Scientific, Waltham, Massachusetts) was used to determine protein concentrations with an extinction coefficient of 1.4 mL mg−1 cm−1 for the antibodies. The extinction coefficient of Nacetyltryptophan was determined to be 20.5 mL mg−1 cm−1 . Total Internal Reflection Fluorescence Surface Modification Silanized quartz surfaces were used for the TIRF experiments. Quartz slides from TIRF Technologies (Cary, North Carolina) were cleaned by the procedure adapted from Kern and Puotinen34 where the surfaces are first immersed in a solution prepared from 25% NH3 , 30% H2 O2 , and H2 O (1:1:5, by volume) at 80◦ C for 5 min, rinsed with water, immersed in 30% HCl, 30% H2 O2 , and H2 O (1:1:5, by volume) at 80◦ C for 5 min, and subsequently rinsed with water and ethanol. The clean quartz surfaces are modified with (3, 3, 3-trifluoropropyl)chloromethylsilane using vapor deposition as previously described35 under argon atmosphere.36 To assure proper and homogenous modification, the contact angles of a sessile drop of ultrapure water on the silanized surfaces were determined. Contact angles were found to be 89 ± 1◦ , measured ¨ G2 contact at three different sites on the surface with a Kruss ¨ GmbH, Hamburg, Germany). angle measuring system (Kruss Each experiment was performed with a newly modified, unused surface. Experimental Run The quartz slide forms the central piece of the TIRF flow cell (TIRF Technologies), which is restricted by a gasket and backblock fitted with the inlet and outlet tubing. The flow cell was fitted into a Spex Fluorolog 3–22 (Jobin Yvon Horiba, Longjumeau Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

Cedex, France). Following a modified protocol from Pinholt et al.,36 a stable baseline was established at a constant flow rate of 4.17 :L/s. Subsequently, a constant wavelength analysis of a concentration range of nonadsorbing external standards of N-acetyl-tryptophan37 was performed, followed by the protein sample (10 mg/mL). Each sample was loaded and oscillated with a flow rate of 4.17 :L/s, whereas the flow rate was increased to 16.67 :L/s in the dissociation phase. The latter served a double purpose: (1) faster elimination of the bulk from the cell and (2) removing loosely adsorbed protein species. Excitation and emission wavelength for the time-resolved experiments were set to 295 and 350 nm, respectively. Integration time was fixed to 0.1 s, slit widths to 5 nm. Quantification of the adsorbed amount, , was performed according to the procedure by Roth and Lenhoff.37 Quartz Crystal Microbalance with Dissipation Monitoring Surface Modification Gold sensors (Biolin Scientific, Gothenburg, Sweden) were cleaned according to the manufacturer’s protocol.38 Modification was achieved by submersion into a 5-mM propanethiol solution in absolute ethanol and incubation for at least 12 h. The sensors were rinsed, kept submerged in absolute ethanol, and were used the same day. Contact angle measurements were performed prior to the QCM-D experiment. Typical values are higher than 85◦ . Regeneration of the sensors was achieved by oxidation of the thiol with piranha solution.39 Sulfates are reported to have low affinity to gold and can be removed with polar solvents.39 All further steps were performed according to the manufacturer’s protocol.38 Experimental Run The sensors were mounted in a Qsense E4 (Biolin Scientific) and equilibrated at a flow rate of 100 :L/min in ultrapure water until drift of frequency was less than 1 Hz/h. The experiment was initiated by recording a stable baseline in water, then in 0.9% NaCl, followed by loading of the protein solution for 15 min under constant flow, and concluded with a 0.9% NaCl solution rinse. In some experiments, an additional rinse with PS80 at 2xCMC in 0.9% NaCl was performed. Data were analyzed in QTools 4 (Biolin Scientific). Density of the sodium chloride solution for viscoelastic modeling was estimated to be 1004 kg/m3 by linear interpolation from reference.40 Density of the adsorbed antibody layer was set to 1100 kg/m3 assuming 60% water content.41,42 Overtones 5, 7, and 9 were further fitted to the Sauerbrey approach43 giving similar results for the total wet adsorbed amount.

RESULTS The effect of PS80 on the adsorption behavior of the two immunoglobulins of subclass G (IgGs) in the study was performed in three distinct stages: (1) adsorption of the protein with a consecutive rinsing step with 2xCMC PS80, (2) adsorption of mixed protein and PS80 samples, and (3) adsorption of PS80 followed by protein. These are detailed in the following paragraphs. Adsorption of Antibodies to Hydrophobic Surfaces To investigate the influence of PS80 on the adsorption behavior of the antibodies, a reference point was established using both DOI 10.1002/jps.24265

RESEARCH ARTICLE – Pharmaceutical Biotechnology

Figure 1. Typical traces for protein adsorption (a) adsorbed mass () as calculated from the Sauerbrey fit to the seventh overtone (QCM-D) for 10 mg/mL PAN sample. After equilibration in ultra pure water (t = 0 s), the sensors were exposed to running solution (1), protein solution and incubated for 15 min (2), rinsed with 0.9% NaCl solution (3), PS80 at 2xCMC (4), and finally 0.9% NaCl solution (5). (b) Adsorbed mass () as calculated from TIRF data for 10 mg/mL mAb-1 sample. After an initial baseline was recorded in buffer, protein was injected and incubated for 15 min (2), followed by a rinsing step with 0.9% NaCl (3) in order to remove loosely adsorbed and bulk protein, and finally a rinse with PS80 (4) in order to remove additional adsorbed protein.

QCM-D and TIRF for the protein adsorption in the absence of any surfactant. In both cases, the experiment started by flushing the cells with 0.9% NaCl for baseline recording, followed by loading and incubation of the protein sample for 15 min in the absence of PS80 (Fig. 1, arrow 2). The following rinsing step with 0.9% NaCl solution removed bulk mAb along with loosely attached mAb (Fig. 1, arrow 3). The remaining adsorbed protein film was stable over the investigated time frame. Quartz crystal microbalance with dissipation monitoring measures different overtones of the resonance frequency and dissipation values, which relate to adsorbed mass and rigidity of the adsorbed layer, respectively. The various QCM-D overtones did not present a large spread (2.5% for mAb-1 and 0.7% for PAN), indicating a rigid-like behavior of the film.27 This is also supported by the low dissipation upon rinsing after protein adsorption [D7(mAb-1; t = 60 min]: 1.06 × 10−6 ± 0.02 × 10−6 ; D7(PAN; t = 60 min): 0.972 × 10−6 ± 0.002 × 10−6 ) that resulted in a dissipation-to-frequency ratio of less than 1/50 (Supplementary Figs. 1 and 2). However, the Sauerbrey method tends to underestimate adsorbed amounts when films are viscoelastic (dissipation values >0).27,44,45 Hence, the QCM-D data were also modeled using the Voigt approach for viscoelastic DOI 10.1002/jps.24265

595

Figure 2. Adsorbed amount for 10 mg/mL PAN (a) and mAb-1 (b) after rinsing extensively with 0.9% NaCl solution as determined by QCM-D (modeled by viscoelastic and Sauerbrey approach as fitted to seventh overtone) and TIRF.

modeling. This yields significantly higher (∼40%) adsorption values for Voigt modeling as compared with Sauerbrey (Fig. 2). Interestingly, the TIRF analysis gave consistently lower adsorbed amounts than the QCM-D analysis regardless of the method used to fit the QCM-D data or the type of antibody. However, this difference is not statistically significant when comparing the Sauerbrey fit and TIRF data (Fig. 2). The adsorbed layer of PAN was found to be 12.2 ± 0.4 nm, whereas for mAb-1, it was 12.5 ± 0.4 nm, as calculated from Voigt modeling of QCM-D data. The TIRF setup also allows for recording fluorescence emission scans of the adsorbed proteins. Changes in the micro environment of the intrinsic fluorophore because of structural changes could be observed as a shift in the emission maximum. However, comparison of adsorbed mAb and cuvette measurements showed no difference (data not shown). Surfactant Rinsing Removes a Fraction of the Adsorbed Antibody Layers The obtained TIRF data only relies on the intrinsic fluorescence of the mAbs. Therefore, the drop in fluorescence signal upon rinsing with 2xCMC PS80 in 0.9% NaCl solution in Figure 1b (arrow 4) must be because of the partial removal of Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

596

RESEARCH ARTICLE – Pharmaceutical Biotechnology

protein. In contrast, upon introduction of the same surfactantcontaining rinsing solution in the QCM-D, a mass increase was registered (Fig. 1a, arrow 4). This increase cannot be because of changes in the coupling to the bulk solution, as the salt concentration was kept constant and the PS80 concentration was rather low. Instead, this apparent mass increase is probably because of surfactant adsorption at the interface. A consecutive rinsing step with 0.9% NaCl revealed that the adsorbed mass returned to values close to those before surfactant addition (Fig. 1a, compare steady states after arrows 3 and 5). Taking together the QCM-D and TIRF data, exposure of a PS8O solution to a preadsorbed antibody layer thus leads to surfactant adsorption leading to partial replacement of proteins at the solid–liquid interface. Surfactant–Antibody Mixtures and Their Effect on the Adsorbed Layer To evaluate the effect of PS80 on the antibody adsorption behavior, premixed surfactant PS80/antibody samples were exposed to hydrophobically modified surfaces at increasing surfactant content (0, 7, and 30 mg/L, corresponding to 0xCMC, 1/2xCMC, and 2xCMC, respectively) (Fig. 3). The reduction in proteinadsorbed amount was pronounced already at a PS80 concentration of 1/2xCMC for mAb-1. The impact of the surfactant addition is comparable for both antibodies, showing a decreased protein adsorption as the PS80 concentration increases. The reduction of adsorbed protein in the mixed samples with 2xCMC is about 70% (from 530 to 130 ng/cm2 ) for PAN, whereas it is almost 90% (from 780 to 70 ng/cm2 ) for mAb-1 samples (judged by sample mean).

Figure 4. Adsorbed amount as a function of time for mixed mAb/PS80 sample at 1/2xCMC run as measured by QCM-D (a) and TIRF (b). Upon protein loading, a spiking event is observed that is present in both methods at short time scale (indicated by arrow).

Early Event in Mixed Sample Adsorption The injection of a mixed PS80/mAb sample lead to a spiking event in the adsorption at the very early stage of the sampleloading phase (Fig. 4). This spike in the adsorption kinetics occurred both in the TIRF and QCM-D traces and occurred on a time scale of about 60 s of initial injection. The spike did not disappear when the system, excluding the flow cell, was presaturated with surfactant. This indicates this event is not a methodological artifact induced by surfactant depletion because of adsorption to tubing. Polysorbate Present Before and During the Protein Adsorption Phase Pre-exposing the surface to a 1/2xCMC PS80 solution resulted in complete prevention of protein adsorption, as measured by TIRF and exemplified in Figure 5 for PAN.

DISCUSSION Adsorption of Antibodies to surface/Comparison of Data from QCM-D and TIRF

Figure 3. Adsorbed amount for mixtures of the surfactant PS80 and 10 mg/mL PAN (a) and mAb-1 (b) as a function of PS80 concentration as measured by TIRF. Individual measurements are shown. Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

The methods used to determine the adsorbed mass are QCMD and TIRF. QCM-D provides two general options for data analysis: Sauerbrey and viscoelastic modeling. Although the Sauerbrey approach tends to underestimate the mass of films with viscoelastic properties,27,44,45 it enables a simplified direct (qualitative and quantitative) comparison of data, as no further parameter than the Sauerbrey constant (17.7 ng cm−2 Hz−1 at DOI 10.1002/jps.24265

RESEARCH ARTICLE – Pharmaceutical Biotechnology

Figure 5. Total internal reflection fluorescence data of 10 mg/mL PAN sample introduced to the cell after preincubation of the cell with 1/ xCMC PS80. No adsorption is detectable upon rinsing with 0.9% 2 NaCl.

f = 5 MHz)46 is required for fitting. The Sauerbrey equation may be used if the film is sufficiently rigid (D/F ratio of about 1/30).47 Small dissipation values, like those found in our study, may still lead to significant differences between adsorbed amount obtained by Sauerbrey and viscoelastic modeling (Fig. 2). The adsorbed mass for protein adsorption in the absence of surfactant as measured by TIRF was consistently lower than the Voigt modeling of the QCM-D data (Fig. 2). This is expected because of the fundamental difference of the two methods: TIRF as an optical method yields the “dry” adsorbed protein mass, whereas QCM-D provides a “wet” adsorbed mass.41 The protein fraction, hydrodynamic effects in heterogeneous films, entrapped water, as well as hydration water contribute to the QCM-D signals.48 Thus, in theory, the amount of associated water could be estimated from differences in adsorbed amount determined by QCM-D and TIRF or any other optical method.41,42,49 However, hydrodynamic effects in heterogeneous films further complicate the interpretation of QCM-D data and the total adsorbed amount will depend on protein hydration, orientation on the surface, and surface arrangement. For example, V¨or¨os reported at least five times smaller amount of a human IgG adsorbed to Teflon measured by optical waveguide lightmode spectroscopy41 as compared with QCM-D. These results were obtained by a Sauerbrey fit to the seventh overtone. This suggests an estimated water content of 80% (based on comparison of graphs presented in the paper).41 However, this value may be overestimated as the measurement was performed in the presence of the protein solution, and hence there may be bulk coupling contributing to the measured QCM-D values. From literature data of polyclonal antibodies binding to antigens prebound on the surface, a conservative estimate of 60% water content has been drawn based on a geometrical estimate.42 We point out that the mode of binding to these antigens is fundamentally different and the numbers may therefore not be directly comparable. It should further be mentioned that the apparent water content observed in the presented results is lower than expected (from Fig. 2, about 25% for mAb-1 and 45% for PAN). However, because of the large standard deviation in the TIRF results a quantification of the water contribution in the QCM-D measurements should be used with caution. For the two antibodies studied, no significant difference in adsorbed amount was found when measured by TIRF or QCMD (Fig. 2). Taking into consideration that IgGs are structurally DOI 10.1002/jps.24265

597

closely related, with the majority of the molecule being defined by the constant region, it seems appropriate to assume that their hydration shell in solution is also comparable. This can change at the interface, depending on the orientation of the molecule, spreading, hydration forces, and overall layer formation. Wiseman and Frank50 analyzed the adsorption of mAbs adsorbed into defined orientations by preadsorbing protein G or its antigen streptavidin. They concluded that protein G leads to a more side-on than Fab-up orientation and binding to the antigen to a Fab-down orientation. According to their study a D value of approximately 1 × 10−6 may indicate an end-on (Fab-down) orientation, whereas a D value of approximately 0.5 × 10−6 likely indicates a more side-on orientation.50 Typical D values obtained in our study are close to 1 × 10−6 that therefore indicates that the antibodies possibly adopt an extended end-on orientation on the surface. Indeed, the layer thickness modeled by QCM-D for PAN (12.2 nm) and mAb-1 (12.5 nm) are closer to the long-axis dimensions (16.4 nm) of PAN reported in solution than its short axis of 2.5 nm.51 Oom et al.52 published vicoelastically modeled QCM-D data for mAb adsorption (50 mg/mL) onto polystyrene (11.8 nm) and Teflon (15.8 nm) surfaces, which is in good agreement with the obtained results in our study. Similarly, a layer thickness of 12.5 nm was reported for an mAb at a solution concentration of 5 mg/mL adsorbed onto a hydrophobic surface by neutron reflectometry.53 The mAb solution concentration is an important variable for antibody orientation to consider as thickness values suggesting a flat-on or a mixed orientation were found at low concentrations on polystyrene (3.0 nm), teflon surfaces (3.7 nm),52 self-assembled 1-dodecanethiol monolayers as determined by QCM-D50 and possibly a mixture for octadecyltrichlorosilane (9.9 nm) as determined by neutron reflectivity.53 Interpretation of low-concentration investigations measured by QCM-D has to be considered carefully, as the method is averaging over the sensor surface. Hence, an incomplete surface coverage may result in an anomalously low thickness. However, the surface coverage for an IgG remained constant over a solution concentration interval of 0.01–1 mg/mL, which indicates full surface coverage.50 The deviations between size in solution and adsorbed species may therefore be attributed to unfolding on the surface, defects in the adsorbed layer yielding a reduced apparent thickness and/or different orientations of the adsorbed species. mAbs have indeed been shown to partly unfold upon adsorption to hydrophobic surfaces,3 likely involving unfolding of the Fab fragment, whereas the Fc retains its native fold.54 Buffer conditions affect protein adsorption,55,56 which further limits comparability between studies in the literature. Although the studies performed by V¨or¨os41 and H¨oo¨ k et al.42 were conducted at pH 7.4 in 10 mM HEPES buffer, the current data were obtained in 0.9% NaCl (154 mM) at pH 6. These conditions were chosen as preparation of mAb formulations for patients in hospitals regularly includes dilution into 0.9% NaCl. Buijs et al.6 theoretically estimated adsorbed amounts of an extended antibody with contracted Fab fragments to  = 550 ng/cm2 . This corresponds approximately to the mean value found for PAN determined by TIRF ( = 525 ng/cm2 ). The mean TIRF result for mAb-1 is higher with  = 780 ng/cm2 . These two values are not significantly different and thus a full quantitative evaluation cannot be made. The surface properties between QCM-D and TIRF experiments may vary because Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

598

RESEARCH ARTICLE – Pharmaceutical Biotechnology

of different substrates and modifying agents used affecting the total adsorbed amount. The QCM-D sensors are gold-based and have no intrinsic charge, whereas residual silanol groups on the TIRF slides carry negative charge.57–59 There may also be large differences in roughness on these substrates. However, charge has been shown not to affect hydrophobicity,60 and both surfaces display comparable contact angles after modification. Thus, hydrophobic interactions seem to be the main driving force in the adsorption process.6,61 On the contrary, surface charge may also influence adsorption.6 Negatively charged polystyrene beads of higher-charge density yielded a higher absorbed amount for IgGs than beads of lower-charge density.61 Analogously, it could be hypothesized that higher-charge on the protein, globally or in charge clusters, may result in higher adsorbed amount. However, increased charge of the protein increases lateral repulsion. The pI of PAN is 6.63, whereas that for mAb-1, it is 8.44 and thus PAN should be less charged at pH 6, possibly resulting in increased adsorbed amount. Indeed, experimental data suggest that proteins adsorb maximally around their pI, whereas moving away from pI the resulting protein charge increases internal and lateral repulsion on the surface, thereby reducing the adsorbed amount.6,62,63 Surfactant Rinsing Removes a Fraction of Adsorbed mAbs The adsorbed amount measured by TIRF decreases upon rinsing with PS80, indicating removal of bound protein. To ensure that this reduction was not an artifact caused by quenching of the tryptophan fluorescence by PS80, a successive rinsing step with 0.9% NaCl solution was performed (Supplementary Fig. 3). Furthermore, cuvette measurements with N-actetyltryptophan and mAbs showed that no quenching occurs upon adding PS80 (data not shown). This leads to the assumption that at least two different adsorbed protein species are present, where one is more loosely attached and can be removed by PS80 exposure, whereas the other is irreversibly bound and inert to rinsing with surfactant (Fig. 1b, arrow 4). The quantification of adsorbed amount indicates that  is close to a corresponding monolayer.59 Partial removal of proteins adsorbed to hydrophobic surfaces upon exposure to surfactants has been reported previously.53,64–66 Neutron reflectometry data suggest that at hydrophobic surfaces a monolayer is favored,53 but neutron reflectivity may not detect the loosely adsorbed proteins because of low contrast of this layer. Taking the random adsorption model67 into consideration, proteins may initially adsorb in a more side-on orientation, and as the surface approaches saturation, more mAbs adsorb in an end-on orientation.50 Furthermore, it could be argued that end-on oriented mAbs may not be as tightly bound, as less of the mAbs surface is bound to the surface, which would make it easier for PS80 to outcompete the mAbs from the surface. Competition of mAbs and PS80 and the Effect on Protein Surface Concentration Sequential adsorption of mAb followed by a rinse with PS80 revealed a competitive behavior (Fig. 1). Therefore, mixed samples at different protein to surfactant ratios were investigated in order to gain further insight into the behavior of the two IgG subclasses studied. Protein adsorption was clearly reduced in the presence of 2xCMC PS80 (Fig. 3), whereas a significant reduction at 1/2xCMC (p < 0.05) was only observed for mAb-1, though the trend seems to be the same for both mAbs. This Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

finding is expected as IgG1 and IgG2 are almost identical in molecular weight and only differ in the number of disulfide bonds.68 Comparable trends were observed for the adsorption of human serum albumin onto methylated silica in the presence of PS2069 and recombinant factor VIII.70 These results are in agreement with the hypothesis of the surfactant monomer being more surface active than the protein and thus the outcompeting species.71 Two major factors have been proposed to explain this influence of polysorbate: (1) the surfactant occupies hydrophobic sites (on the solid surface and potentially the protein layer) because of lower interfacial energy, and (2) the hydrophilic head group hinders protein adsorption by steric repulsion.72 The repulsion argument is supported by a study where polyethylene glycol (PEG) chains were grafted on silica, preventing adsorption of bovine serum albumin.73 Mollmann et al.71 have described the full displacement of insulin by PS80 below CMC adsorbed on the same type of surface used in this study. They proposed that the observed displacement follows the orogenic displacement model, which explains displacement of proteins by adsorption and nucleation of surfactant molecules in defects of the adsorbed protein layer. Subsequent growth of these surfactant clusters displaces protein by increasing the surface pressure until the protein layer collapses.74 This process is reflected in a sigmoidal desorption over time curve suggesting a nucleation dependency. For both mAbs in this study, no such effect could be observed in rinsing experiments at 2xCMC at the investigated time frame up to 60 min (Fig. 1). Attempts to prolong total sample run times did not succeed because of the entrapment of air bubbles in the TIRF cell. A possible explanation for the higher mechanical resistance of the antibody film as compared with insulin could be a more elastic behavior of the adsorbed layer.75 Furthermore, the much larger size of antibodies increases the contact points at the surface and the neighboring protein molecules, thus strengthening the protein film. Early Event in Competition Studies When mixed samples of protein and PS80 were loaded on the surface, both methods showed an initial, short-lived large adsorption (Fig. 4). A similar observation has been made in mixed samples of lysozyme, but was not further commented.76 The event happens at comparable time scale in the two surfacesensitive methods, even though different flow rates are applied (Fig. 4). An intuitive approach to explain this is by the different diffusion coefficients of PS80 (1.8 × 10−6 cm2 /s for the monomer, 1. 8 × 10−7 cm2 /s for the micelle)77 and mAb (1.99 × 10−11 cm2 /s).53 However, these values suggest that the surface would be saturated with the surfactant before the antibody even arrives. The time scale of the event (in terms of height and width of the spike) remained constant regardless of surfactant concentrations being below or above CMC, further suggesting that they cannot be solely attributed to differences in diffusion coefficients (Supplementary Fig. 5). The presence of this spiking event is likely not because of PS80 depletion from the bulk as a result of adsorption to tubing as the height and width should then be reduced upon increasing PS80 concentration. As PS80 adsorption is only partly reversible upon rinsing with buffer,76 the depletion hypothesis was tested by first saturating all sample tubing, excluding the flow cell, with PS80. The mixed protein sample was loaded after rinsing the tubing DOI 10.1002/jps.24265

RESEARCH ARTICLE – Pharmaceutical Biotechnology

with 0.9% NaCl solution. It can be assumed that PS80 depletion from a sample loaded via a previously saturated tube will be significantly reduced or even absent. However, the resulting adsorption profile corresponded to that of the non-pre-exposed mixed samples, thus providing direct evidence contradicting the depletion hypothesis. An alternative explanation is the rearrangement of the adsorbed protein on the surface upon exposure to surfactant. Polysorbate Present at All Times Because of the high resistance of adsorbed PS80 to extensive buffer rinsing (Supplementary Fig. 4), it is expected that no protein adsorption is detectable when 1/2xCMC PS80 is present in the running buffer. Over the course of the experimental stages (establishment of baseline, calibration with nonadsorbing standards), the hydrophobic surface is slowly saturated with surfactant. Hence, when the protein is finally introduced to the flow cell, it is sterically prevented from adsorption, similar to the adsorption-preventing property of PEG-modified surfaces.73 It can thus be expected that at very high surfactant concentrations (multiple times CMC), the surface is immediately saturated with PS80, preventing all protein adsorption. Indeed, it has been reported that PS80 concentrations of 0.1% can prevent surface-induced aggregation of protein.78 It has been suggested that no specific surfactant-to-protein ratio is needed to stabilize mAbs, as polysorbates do not directly interact with the protein.16,79 However, combining the results presented in this work with literature, it appears that there is a threshold concentration between above CMC that leads to complete blocking of any protein adsorption to hydrophobic surfaces.78

CONCLUSION It has been shown in this work that PAN and mAb-1 adsorb readily to hydrophobic surfaces under conditions frequently encountered in hospital pharmacies. In the absence of PS80, protein layers are formed, which are resistant to solution and partly to surfactant rinse. However, the extent of protein adsorption is not expected to impact drug application, as relevant doses for mAb therapy are much larger. The sequence of exposure has been found to be crucial for the extent of protein adsorption: although mixed mAb and PS80 samples still showed adsorption, no protein adsorption could be detected for surfaces pre-exposed to PS80. This may be of practical importance, when adsorption to infusion bags and iv lines is critical. For example, in early clinical trials in humans where concentrations are low to test safety, presaturation of the surfaces with PS80 may be beneficial. Protein adsorption to interfaces is a complex topic and variations in methodology, experimental parameters, and masked identities of mAbs complicate the generation of general conclusions. Therefore, we suggest PAN to be included as a reference antibody in future studies. It carries the advantages of being a highly defined commercially available and affordable product, also allowing for simple buffer exchange because of the low molecular excipients in the formulation.

ACKNOWLEDGMENTS MedImmune is thanked for providing mAb-1. Amgros I/S and the Drug Research Academy (DRA) are thanked for financial DOI 10.1002/jps.24265

599

support of the project. DRA is further acknowledged for financing the Nanodrop. “Apotekerfonden af 1991” is acknowledged for financing the fluorescence spectrophotometer used in this study. Jinying Liu is thanked for her assistance in obtaining some TIRF measurements.

REFERENCES 1. La Merie Publishing. 2013. Blockbuster biologics 2012. La Merie Publishing. Accessed May 5, 2014, at: www.pipelinereview.com. 2. Elwing H, Nilsson B, Svensson KE, Askendahl A, Nilsson UR, Lundstr¨om I. 1988. Conformational changes of a model protein (complement factor 3) adsorbed on hydrophilic and hydrophobic solid surfaces. J Colloid Interface Sci 125:139–145. 3. Vermeer AWP, Bremer MGEG, Norde W. 1998. Structural changes of IgG induced by heat treatment and by adsorption onto a hydrophobic Teflon surface studied by circular dichroism spectroscopy. Biochim Biophys Acta 1425:1–12. 4. Zoungrana T, Findenegg GH, Norde W. 1997. Structure, stability, and activity of adsorbed enzymes. J Colloid Interface Sci 190:437– 448. 5. Zoungrana T, Norde W. 1997. Thermal stability and enzymatic activity of alpha-chymotrypsin adsorbed on polystyrene surfaces. Colloids Surf B Biointerfaces 9:157–167. 6. Buijs J, Lichtenbelt JWT, Norde W, Lyklema J. 1995. Adsorption of monoclonal IgGs and their F(ab’)2 fragments onto polymeric surfaces. Colloids Surf B Biointerfaces 5:11–23. 7. Shomali M, Freitag A, Engert J, Siedler M, Kaymakcalan Z, Winter G, Carpenter JF, Randolph TW. 2014. Antibody responses in mice to particles formed from adsorption of a murine monoclonal antibody onto glass microparticles. J Pharm Sci 103:78–89. 8. Van Beers MMC, Gilli F, Schellekens H, Randolph TW, Jiskoot W. 2012. Immunogenicity of recombinant human interferon beta interacting with particles of glass, metal, and polystyrene. J Pharm Sci 101:187–199. 9. Rosenberg AS. 2006. Effects of protein aggregates: An immunologic perspective. AAPS J 8:E501–E507. 10. Jiskoot W, Randolph TW, Volkin DB, Middaugh CR, Sch¨oneich C, Winter G, Friess W, Crommelin DJA, Carpenter JF. 2012. Protein instability and immunogenicity: Roadblocks to clinical application of injectable protein delivery systems for sustained release. J Pharm Sci 101:946–954. 11. Bam NB, Cleland JL, Yang J, Manning MC, Carpenter JF, Kelley RF, Randolph TW. 1998. Tween protects recombinant human growth hormone against agitation-induced damage via hydrophobic interactions. J Pharm Sci 87:1554–1559. 12. Patapoff TW, Esue O. 2009. Polysorbate 20 prevents the precipitation of a monoclonal antibody during shear. Pharm Dev Technol 14:659–664. 13. Kreilgaard L, Jones LS, Randolph TW, Frokjaer S, Flink JM, Manning MC, Carpenter JF. 1998. Effect of tween 20 on freeze-thawingand agitation-induced aggregation of recombinant human factor XIII. J Pharm Sci 87:1593–1603. 14. Wang W, Wang YJ, Wang DQ. 2008. Dual effects of tween 80 on protein stability. Int J Pharm 347:31–38. 15. Chou DK, Krishnamurthy R, Randolph TW, Carpenter JF, Manning MC. 2005. Effects of tween 20 and tween 80 on the stability of albutropin during agitation. J Pharm Sci 94:1368–1381. 16. Garidel P, Hoffmann C, Blume A. 2009. A thermodynamic analysis of the binding interaction between polysorbate 20 and 80 with human serum albumins and immunoglobulins: A contribution to understand colloidal protein stabilisation. Biophys Chem 143:70–78. 17. Bam NB, Cleland JL, Randolph TW. 1996. Molten globule intermediate of recombinant human growth hormone: Stabilization with surfactants. Biotechnol Prog 12:801–809. 18. Lee LLY, Lee JC. 1987. Thermal stability of proteins in the presence of poly(ethylene glycols). Biochemistry 26:7813–7819. Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

600

RESEARCH ARTICLE – Pharmaceutical Biotechnology

19. Arakawa T, Timasheff SN. 1985. Mechanism of polyethylene glycol interaction with proteins. Biochemistry 24:6756–6762. 20. Dixit N, Maloney K, Kalonia D. 2013. Protein-silicone oil interactions: Comparative effect of nonionic surfactants on the interfacial behavior of a fusion protein. Pharm Res 30:1848–1859. 21. Lee HJ, McAuley A, Schilke KF, McGuire J. 2011. Molecular origins of surfactant-mediated stabilization of protein drugs. Adv Drug Deliv Rev 63:1160–1171. 22. Westerdahl CAL, Hall JR, Schramm EC, Levi DW. 1974. Gas plasma effects on polymer surfaces. J Colloid Interface Sci 47:610–620. 23. Owens DK, Wendt RC. 1969. Estimation of the surface free energy of polymers. J Appl Polym Sci 13:1741–1747. 24. Occhiello E, Morra M, Morini G, Garbassi F, Johnson D. 1991. On oxygen plasma-treated polypropylene interfaces with air, water, and epoxy resins. II. Epoxy resins. J Appl Polym Sci 42:2045–2052. 25. Strobel M, Corn S, Lyons CS, Korba GA. 1985. Surface modification of polypropylene with CF4, CF3H, CF3Cl, and CF3Br plasmas. J Polym Sci Polym Chem Ed 23:1125–1135. 26. Vikinge TP, Hansson KM, Sandstr¨om P, Liedberg B, Lindahl TL, Lundstr¨om I, Tengvall P, H¨oo¨ k F. 2000. Comparison of surface plasmon resonance and quartz crystal microbalance in the study of whole blood and plasma coagulation. Biosensors Bioelectron 15:605–613. 27. Rodahl M, Hook F, Fredriksson C, Keller A, Krozer A, Brzezinski P, Voinova M, Kasemo B. 1997. Simultaneous frequency and dissipation factor QCM measurements of biomolecular adsorption and cell adhesion. Faraday Discuss 107:229–246. 28. Rehder D, Bondarenko P, Chelius D, McAuley A, Matsumura M. 2010. Stable polypeptide formulations. Patent US20100158908 A1. 29. EMA. 2007. Scientific discussion—Vectibix, WC500047707. http://www.ema.europa.eu/docs/en GB/document library/EPAR Scientific Discussion/human/000741/WC500047707.pdf 30. Wan LSC, Lee PFS. 1974. CMC of polysorbates. J Pharm Sci 63:136–137. 31. Schott H, Han SK. 1976. Effect of inorganic additives on solutions of nonionic surfactants IV: Krafft points. J Pharm Sci 65:979–981. 32. Armstrong JK, Leharne SA, Stuart BH, Snowden MJ, Chowdhry BZ. 2001. Phase transition properties of poly(ethylene oxide) in aqueous solutions of sodium chloride. Langmuir 17:4482–4485. 33. Kerwin BA. 2008. Polysorbates 20 and 80 used in the formulation of protein biotherapeutics: Structure and degradation pathways. J Pharm Sci 97:2924–2935. 34. Kern W, Puotinen DARC. 1970. Cleaning solutions based on hydrogen peroxide for use in silicon semiconductor technology. RCA Rev 31:187–206. 35. Mollmann SH, Bukrinsky JT, Frokjaer S, Elofsson U. 2005. Adsorption of human insulin and AspB28 insulin on a PTFE-like surface. J Colloid Interface Sci 286:28–35. 36. Pinholt C, Kapp SJ, Bukrinsky JT, Hostrup S, Frokjaer S, Norde W, Jorgensen L. 2013. Influence of acylation on the adsorption of GLP-2 to hydrophobic surfaces. Int J Pharm 440:63–71. 37. Roth CM, Lenhoff AM. 1995. Electrostatic and van der Waals contributions to protein adsorption: Comparison of theory and experiment. Langmuir 11:3500–3509. 38. BiolinScientific/qsense. 2008. User’s guides—Cleaning and immobilization protocols. Stockholm, Sweden: Qsense. 39. Love JC, Estroff LA, Kriebel JK, Nuzzo RG, Whitesides GM. 2005. Self-assembled monolayers of thiolates on metals as a form of nanotechnology. Chem Rev 105:1103–1170. 40. Rogers PSZ, Pitzer KS. 1982. Volumetric properties of aqueous sodium chloride solutions. J Phys Chem Ref Data 11:15–81. 41. V¨or¨os J. 2004. The density and refractive index of adsorbing protein layers. Biophys J 87:553–561. 42. H¨oo¨ k F, V¨oo¨ s J, Rodahl M, Kurrat R, B¨oni P, Ramsden JJ, Textor M, Spencer ND, Tengvall P, Gold J, Kasemo B. 2002. A comparative study of protein adsorption on titanium oxide surfaces using in situ ellipsometry, optical waveguide lightmode spectroscopy, and quartz crystal microbalance/dissipation. Colloids Surf B Biointerfaces 24:155– 170. Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015

¨ 43. Sauerbrey G. 1959. Verwendung von Schwingquarzen zur Wagung ¨ ¨ dunner Schichten und zur Mikrowagung. Z Physik 155:206–222. 44. Voinova MV, Rodahl M, Jonson M, Kasemo B. 1999. Viscoelastic acoustic response of layered polymer films at fluid–solid interfaces: Continuum mechanics approach. Phys Scr 59:391–396. 45. Reed CE, Kanazawa KK, Kaufman JH. 1990. Physical description of a viscoelastically loaded AT-cut quartz resonator. J Appl Phys 68:1993. 46. H¨oo¨ k F, Rodahl M, Brzezinski P, Kasemo B. 1998. Energy dissipation kinetics for protein and antibody−antigen adsorption under shear oscillation on a quartz crystal microbalance. Langmuir 14:729–734. 47. BiolinScientific/qsense. 2014. Accessed November, 2014, at:. Accessed November, 2014, at: http://www.biolinscientific.com/ service/viscoelastic-modeling/. 48. H¨oo¨ k F, Kasemo B, Nylander T, Fant C, Sott K, Elwing H. 2001. Variations in coupled water, viscoelastic properties, and film thickness of a mefp-1 protein film during adsorption and cross-linking: A quartz crystal microbalance with dissipation monitoring, ellipsometry, and surface plasmon resonance study. Anal Chem 73:5796–5804. 49. Malmstr¨om J, Agheli H, Kingshott P, Sutherland DS. 2007. Viscoelastic modeling of highly hydrated laminin layers at homogeneous and nanostructured surfaces: Quantification of protein layer properties using QCM-D and SPR. Langmuir 23:9760–9768. 50. Wiseman ME, Frank CW. 2011. Antibody adsorption and orientation on hydrophobic surfaces. Langmuir 28:1765–1774. 51. Mosbæk C, Konarev P, Svergun D, Rischel C, Vestergaard B. 2012. High concentration formulation studies of an IgG2 antibody using small angle x-ray scattering. Pharm Res 29:2225–2235. 52. Oom A, Poggi M, Wikstr¨om J, Sukumar M. 2012. Surface interactions of monoclonal antibodies characterized by quartz crystal microbalance with dissipation: Impact of hydrophobicity and protein selfinteractions. J Pharm Sci 101:519–529. 53. Couston RG, Skoda MW, Uddin S, van der Walle CF. 2013. Adsorption behavior of a human monoclonal antibody at hydrophilic and hydrophobic surfaces. mAbs 5:126–139. 54. Vermeer AWP, Giacomelli CE, Norde W. 2001. Adsorption of IgG onto hydrophobic Teflon. Differences between the Fab and Fc domains. Biochim Biophys Acta 1526:61–69. 55. Norde W. 1984. Ion participation in protein adsorption at solid surfaces. Colloids Surf 10:21–31. 56. Mathes J, Friess W. 2011. Influence of pH and ionic strength on IgG adsorption to vials. Eur J Pharm Biopharm 78:239–247. 57. Schillen K, Claesson PM, Malmsten M, Linse P, Booth C. 1997. Properties of poly(ethylene oxide)−poly(butylene oxide) diblock copolymers at the interface between hydrophobic surfaces and water. J Phys Chem B 101:4238–4252. 58. Wang Y, Ferrari M. 2000. Surface modification of micromachined silicon filters. J Mater Sci 35:4923–4930. 59. Buijs J, van den Berg PAW, Lichtenbelt JWT, Norde W, Lyklema J. 1996. Adsorption dynamics of IgG and its F(ab’)2 and Fc fragments studied by reflectometry. J Colloid Interface Sci 178:594–605. 60. Rahman M, Thottappillil R, Berg M, Hillborg H. 2002. Comment on ‘effect of surface charge on hydrophobicity levels of insulating materials’. Generat Transm Distrib IEE Proc 149:300–304. 61. Elgersma AV, Zsom RLJ, Norde W, Lyklema J. 1991. The adsorption of different types of monoclonal immunoglobulin on positively and negatively charged polystyrene latices. Colloids Surf 54:89–101. 62. Bagchi P, Birnbaum SM. 1981. Effect of pH on the adsorption of immunoglobulin G on anionic poly(vinyltoluene) model latex particles. J Colloid Interface Sci 83:460–478. 63. Elgersma AV, Zsom RLJ, Lyklema J, Norde W. 1992. Adsorption competition between albumin and monoclonal immuno-gammaglobulins on polystyrene latices. J Colloid Interface Sci 152:410– 428. 64. Elwing H, Welin S, Askendal A, Nilsson U, Lundstr¨om I. 1987. A wettability gradient method for studies of macromolecular interactions at the liquid/solid interface. J Colloid Interface Sci 119:203–210. ´ 65. Cardenas M, Elofsson U, Lindh L. 2007. Salivary mucin MUC5B could be an important component of in vitro pellicles of human saliva: DOI 10.1002/jps.24265

RESEARCH ARTICLE – Pharmaceutical Biotechnology

An in situ ellipsometry and atomic force microscopy study. Biomacromolecules 8:1149–1156. 66. Svendsen IE, Lindh L, Arnebrant T. 2006. Adsorption behaviour and surfactant elution of cationic salivary proteins at solid/liquid interfaces, studied by in situ ellipsometry. Colloids Surf B Biointerfaces 53:157–166. 67. Feder J. 1980. Random sequential adsorption. J Theor Biol 87:237– 254. 68. Zhang W, Czupryn MJ. 2002. Free sulfhydryl in recombinant monoclonal antibodies. Biotechnol Prog 18:509–513. 69. Zhang M, Ferrari M. 1997. Reduction of albumin adsorption onto silicon surfaces by tween 20. Biotechnol Bioeng 56:618– 625. 70. Joshi O, McGuire J, Wang DQ. 2008. Adsorption and function of recombinant factor VIII at solid−water interfaces in the presence of Tween-80. J Pharm Sci 97:4741–4755. 71. Mollmann SH, Elofsson U, Bukrinsky JT, Frokjaer S. 2005. Displacement of adsorbed insulin by tween 80 monitored using total internal reflection fluorescence and ellipsometry. Pharm Res 22:1931– 1941. 72. Jeon SI, Lee JH, Andrade JD, De Gennes PG. 1991. Protein–surface interactions in the presence of polyethylene oxide: I. Simplified theory. J Colloid Interface Sci 142:149–158.

DOI 10.1002/jps.24265

601

73. Perez E, Lahlil K, Rougeau C, Moraillon A, Chazalviel JN, Ozanam F, Gouget-Laemmel AC. 2012. Influence of the molecular design on the antifouling performance of poly(ethylene glycol) monolayers grafted on (111) Si. Langmuir 28:14654–14664. 74. Mackie AR, Gunning AP, Wilde PJ, Morris VJ. 1999. Orogenic displacement of protein from the oil/water interface. Langmuir 16:2242– 2247. 75. Mackie AR, Gunning AP, Pugnaloni LA, Dickinson E, Wilde PJ, Morris VJ. 2003. Growth of surfactant domains in protein films. Langmuir 19:6032–6038. 76. Joshi O, McGuire J. 2009. Adsorption behavior of lysozyme and tween 80 at hydrophilic and hydrophobic silica−water interfaces. Appl Biochem Biotechnol 152:235–248. 77. Amani A, York P, de Waard H, Anwar J. 2011. Molecular dynamics simulation of a polysorbate 80 micelle in water. Soft Matter 7:2900– 2908. 78. Katakam M, Bell LN, Banga AK. 1995. Effect of surfactants on the physical stability of recombinant human growth hormone. J Pharm Sci 84:713–716. 79. Hoffmann C, Blume A, Miller I, Garidel P. 2009. Insights into protein−polysorbate interactions analysed by means of isothermal titration and differential scanning calorimetry. Eur Biophys J 38:557– 568.

Kapp et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:593–601, 2015