Agitation-induced aggregation and subvisible particulate formation in model proteins

European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Contents lists available at ScienceDirect

European Journal of Pharmaceutics and Biopharmaceutics journal homepage: www.elsevier.com/locate/ejpb

Research paper

Agitation-induced aggregation and subvisible particulate formation in model proteins Murali Jayaraman, Patrick M. Buck, Arun Alphonse Ignatius, Kevin R. King, Wei Wang ⇑ Pharmaceutical Research and Development, Pfizer Inc, Chesterfield, MO, USA

a r t i c l e

i n f o

Article history: Received 1 September 2013 Accepted in revised form 17 January 2014 Available online 23 January 2014 Keywords: Protein Agitation Particle Aggregation Stability Unfolding

a b s t r a c t The kinetics of agitation-induced subvisible particle formation was investigated for a few model proteins – human serum albumin (HSA), hen egg white lysozyme (HEWL), and a monoclonal antibody (IgG2). Experiments were carried out for the first time under relatively low protein concentration and low agitation speed to investigate the details of subvisible particle formation at the initial phase of aggregation (<2%) process. Upon agitation, both soluble higher molecular mass species (HMMS) and subvisible particles (SbVPs) formed at different rates, and via different mechanisms. Agitation enhanced exposure of hydrophobic sites in HSA but did not cause detectable structural changes in HEWL and IgG2. SbVPs from HSA partially dissociates in a neutral pH buffer (SEC mobile phase) but does not upon dilution in the same formulation buffer. Opposite results were obtained for SbVPs from IgG2 and HEWL. Neither the relative hydrophobic surface area nor the Tm of the model proteins seems to be an indicator of tendency for agitation-mediated SbVP formation. Taken together, our data suggests that agitation-induced SbVP formation can occur through different mechanisms and can vary, depending on the protein and solution conditions. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction Aggregation of therapeutic proteins is one of the major issues in protein drug product development for the biopharmaceutical industries [1]. Protein aggregation is a process of conversion of native soluble monomeric proteins into either soluble or insoluble multimeric assemblies and can easily occur under many different process conditions, such as freeze thaw, purification, formulation, filling, lyophilization, and shipping [2–6]. The concern is that aggregated proteins generally have altered pharmacological activities and might have a greater potential to be immunogenic, a serious side effect for protein products [7]. In addition, soluble protein aggregates were shown to have cytotoxic effects [8–10] and insoluble aggregates may be undesirable esthetically for drug products.

Abbreviations: SEC, size exclusion chromatography; DLS, dynamic light scattering; MOE, Molecular Operating Environment; MFI, Micro-Flow Imaging; CC15, COUNT-CAL 15 micron; HSA, human serum albumin; HEWL, hen egg white lysozyme; IgG2, immunoglobulin type 2; DSF, differential scanning fluorimetry; DSC, differential scanning calorimetry; SAS, solvent accessible surface; Tm, melting temperature; HMMS, higher molecular mass species. ⇑ Corresponding author. Biotherapeutics – Pharmaceutical Sciences, Pfizer Inc, 700 Chesterfield Parkway west, Chesterfield, MO 63017, USA. Tel.: +1 636 247 2111; fax: +1 6363 247 0001. E-mail address: wei.2.wang@pfizer.com (W. Wang). http://dx.doi.org/10.1016/j.ejpb.2014.01.004 0939-6411/Ó 2014 Elsevier B.V. All rights reserved.

Therefore, prevention or mitigation of protein aggregation and SbVP formation is an essential task to ensure safety, efficacy, or quality of protein drug products [11]. Agitation is a common stress a protein therapeutic molecule encounters during protein production and shipping processes. It has been hypothesized that native proteins may undergo structural perturbations at the air–water interfaces resulting in partial exposure of the hydrophobic amino acids, which are normally buried in the interior in the native state [12]. It should be noted that these structural perturbations could be very subtle and the existing tools may or may not be able to detect them. The partially exposed hydrophobic amino acid side chains may increase their chances of hydrophobic interactions with one another to form multimeric aggregates. These aggregates could continue to fuse into larger aggregates and eventually insoluble protein particles, as revealed by SEC, dynamic light scattering (DLS) and light obscuration (HIAC) [13–15]. Although use of a low level of a non-ionic surfactant can effectively minimize agitation-induced aggregation, effective control of protein aggregation is still far from satisfaction, as the aggregate formation and conversion of soluble aggregates into subvisible particles are still not well understood. Therefore, understanding the mechanistic basis of protein aggregation and/or subvisible particle formation would be important in developing strategies for aggregation miminization/prevention [16,17].

300

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Detailed characterization of the aggregation process is thus needed in terms of aggregation rate, size distribution, aggregate structure, and reversibility [18–20]. In this work, the formation of aggregates and subvisible particles was investigated under agitation with three model proteins – HSA, HEWL and IgG2. We also examined possible linkages of particle formation with any structural changes, thermal stabilities (Tm and Th), and surface hydrophobicities of these model proteins.

2. Materials and methods 2.1. Materials Human serum albumin (Sigma, catalogue number: A1887) and hen egg white lysozyme (Sigma; catalogue number: L7651) were purchased from Sigma–Aldrich. An in-house IgG2 mAb was used for this study, which was purified as an injectable pharmaceutical candidate through a combination of chromatographic steps. All proteins were prepared at 1 mg/mL in 20 mM sodium acetate buffer, pH 5.5. Six milliliters of samples was filled into 10 mL Type I glass vials (Flint, USP) and sparged with nitrogen to remove the head space oxygen. The sample vials were closed using rubber stoppers (West) and sealed.

2.2. Solvent Accessible Surface Area (SASA) calculation Hydrophobic patches and surface area calculations were computed using MOE (Molecular Operating Environment) by the Chemical Computing Group [21]. The lipophilic surface potential is atom type dependent and calculated using the parameters of Wildman and Crippen [22]. Fig. S1 shows the surface distribution of hydrophobic sites (green) and secondary structural features (red and yellow) in three different proteins.

2.5. Light obscuration method (HIAC) Agitation stressed samples were subjected to light obscuration measurement to determine the particle count and size distribution. A HIAC/Royco Liquid Particle Counter Model 9703 plus was used with a sample volume of 3.5 mL for measurements. Four measurements were carried out per sample and the data from the last three measurements were averaged and reported [23]. The data acquisition was done using PharmSpec (HACH Ultra Analytics, Grant’s Pass, OR) software. Laser source was activated at least 30 min prior to measurements and the HIAC system was rinsed and cleaned well with particle free water to obtain baseline. Instrument was calibrated using 15 lm polystyrene standard – CC15 (COUNT-CAL Precision standards from Thermo Scientific). In all the measurements, the undiluted stressed samples were subjected to direct measurements, since total number of particles from 1 mg/mL solution was generally below the instrument’s threshold detection limit (616,000 particles). Samples were degassed under vacuum for 10 min and gently swirled before the analysis. Final particle counts in different size ranges were converted into equivalent molecular mass by applying the average density of the protein and spherical shape to the particle [4]. 2.6. Micro-Flow Imaging (MFI) method Micro-Flow Imaging (MFI) System, model DPA 4200 (Brightwell Technologies, Inc.) was used to capture the images of subvisible particles. Agitation-stressed samples were degassed under vacuum for 10 min and then subjected to MFI analysis. The system was prerinsed using particle-free water to get clear baseline before and after running each samples. For each MFI measurement, 1 mL of degassed sample was used. Samples were loaded into MFI using barrier pipette tip (Neptune) after gently swirling the solution. Images were collected and processed using MFI view analysis suite software, version 1.0. 2.7. Size exclusion chromatography (SEC)

2.3. Agitation studies Samples of the model proteins – human serum albumin (HSA), hen egg white lysozyme (HEWL) and IgG2 were prepared at 1 mg/mL in 20 mM acetate buffer, pH 5.5 solutions. Agitation was carried out in an upright position using an orbital shaker at 50 rpm at room temperature. Experiments were carried out at a relatively low protein concentration and low agitation speed in order to create marginal stress to limit the aggregation (<2%) and stay away from rapid visible particle formation to clearly understand the mechanisms of agitation-induced particle formation in protein solutions. Agitated samples in duplicates at various time points were collected, visually inspected in an USP inspection box for parenterals, and analyzed immediately by a variety of analytical methods (see below for details). Samples, incubated under the same condition without orbital agitation, were used as controls.

2.4. Dynamic light scattering DLS measurements were carried out using DynaPro plate reader (Wyatt Technology). Agitation-stressed samples (120 lL) were directly transferred to fresh wells of a 96-well microplate to record the scattering data. For each sample, 20 data acquisitions were made with an acquisition time of 30 s and the data obtained were averaged. Dynamics V6 software (Wyatt Technology) was used to process the data and to analyze the autocorrelation functions to obtain the hydrodynamic radii (RH).

Aliquots of agitation-stressed samples (100 lL) were centrifuged at 10,000 rpm for 10 min in an Eppendorf bench top centrifuge. Thirty microliters of the supernatant was analyzed on an Agilent 1200 SEC-HPLC system with a Tosoh column 3000 SW xl (Tosoh Bioscience LLC 7.8 mm ID – 30 cm, 5 lm) column, pre-equilibrated with the mobile phase of 200 mM sodium phosphate, 50 mM NaCl buffer. The column temperature was maintained at 4 °C. The separation was carried out isocratically over 40 min at a flow rate of 0.5 mL/min and monitored at 280 nm. Reference protein solutions were also subjected to each SEC measurement along with the test samples and reproducibility of the SEC runs were verified using the integral area of the reference solution. The percentage monomer was calculated relative to the peak area of the reference solution using the Empower data analysis software. The SEC method has a variation of ±1% under normal operating conditions. 2.8. Fluorescence spectroscopy The samples were mixed with Sypro Orange, (S5692 from Sigma–Aldrich; 5000X concentrate in DMSO) at a volume ratio of 10:1 in a 150 lL quartz cuvette. Fluorescence was measured using an excitation wavelength of 495 nm and emission spectra were collected from 520 nm to 650 nm. All the fluorescence measurements were made using Carry eclipse fluorescence spectrophotometer from Varian. Addition of excessive amount of Sypro Orange did not lead to a significant fluorescence intensity change in samples. The fluorescence emission spectra obtained were corrected for blank solution containing only Sypro orange and the fluorescence

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

intensities at 577 nm were plotted against different agitation time interval. 2.9. Differential scanning calorimetry DSC measurements were carried out using VP capillary DSC instrument (GE). Protein samples of 1 mg/mL concentration with 400 lL volume were loaded into 96 well DSC plates and the corresponding buffer was used a reference. The samples were scanned from 20 to 110 °C at a scan rate of 1 °C/min. Thermograms were corrected for buffer baseline and also normalized for protein concentration. Data were analyzed using Origin 7.0 software (OriginLab Corporation) and thermodynamic parameters (Tm and DH) were obtained by fitting thermograms using a non-2-state unfolding algorithm. 2.10. Differential scanning fluorimetry DSF measurements were carried out using 96-well sample plates on a CFX96 Bio-Rad RT PCR instrument. The Sypro Orange dye (see above) was diluted 40 fold in an appropriate sample buffer and 8 lL of diluted dye solution was mixed with 100 lL of sample solution. Fluorescence emission at 585 nm was collected using HEX filter with a fixed excitation wavelength at 480 nm. The temperature was increased from 20 to 95 °C at a heating rate of 0.33 °C/min. The midpoint of the unfolding transition curve (Th) was obtained using a first derivative of the data within the Biorad-RT PCR software. 3. Results 3.1. Agitation-induced protein aggregation and particle formation An agitation time course was carried out at room temperature using three model proteins – HSA, HEWL and IgG2. All the protein solutions remain clear and essentially free from visible particulates throughout the study period. On the other hand, agitation did induce a loss of monomer and presence of soluble higher molecular mass species (HMMS) as monitored by SEC-HPLC for all three model proteins (Fig. 1). In comparison, the total loss of monomers for the three proteins was estimated to be 1.8 ± 0.31% for HSA, 1.5 ± 0.28% for IgG2 and 3.3 ± 0.32% for HEWL after 48 h agitation. There was no detectable change in HMMS levels in HSA, HEWL and IgG2 under the agitation condition (Fig. S2). It is obvious that the presence of HMMS does not account for the total loss of monomers

301

for all three proteins. This is likely due to formation of aggregates with molecular weights larger than the separation range of the SEC column, and/or formation of subvisible particulates. Indeed, analysis of the agitated samples by HIAC clearly shows a growth in the number of subvisible particles with different size ranges from 1.5 to 80 lm for all three proteins (Fig. 2A–C). The number of particles appears to grow in two phases – a rapid initial phase (up to 9 h) and a slow growth/condensation phase occurring from 9 to 48 h. The most abundant particles are the smallest ones of 1.5 lm size and the relative abundance dropped gradually with increasing particle size for all three proteins. The relative amounts of proteins in these particles of different sizes were also estimated, assuming spherical shaped particles and average density of proteins [4] as shown in Fig. 2D–F. HSA had a broader particle distribution with less than 0.1 lg of proteins as particles of P1.5 and 625 lm at the early stage of agitation. With time, these small particles eventually condense to form larger particles (P25–80 lm). Similarly, the agitated HEWL showed less than 0.02 lg of proteins as particles of P1.5 and 620 lm for up to 4 h of agitation. With time, two particle mass distributions were observed with size ranges between 1.5 and 20 lm, and between 25 and 50 lm, with the later being dominant. Although the trend in particle formation in IgG2 is somewhat similar to HSA samples, the mass of protein converted into subvisible particles was above 0.2 lg with a particle size range of P1.5 and 650 lm. The median of distribution was 35 ± 10 lm, and did not grow larger with agitation in contrast to HSA and HEWL. The total amount of proteins in these particles formed under agitation varied significantly – about 1.0, 0.25 and 2.5 lg, respectively, for HSA, HEWL and IgG2 after 9 h. A variety of particle shapes were found using micro-flow imaging method. Representative and random-selected particles in the range of P10 and 650 lm are shown in Fig. 3. While HSA formed opaque amorphous protein particles with irregular shapes, HEWL produced translucent branched fibrous particles, and IgG2 formed a mixture of translucent fibrous type and opaque amorphous particles. There were no visible particles (>100 lm) observed by visual inspection in all the protein samples agitated for 48 h. The aggregation process during agitation was also monitored by DLS. Fig. 4 shows comparative results of aggregation in HSA (a), HEWL (b) and IgG2 (c). The monomeric HSA, HEWL and IgG2 had hydrodynamic radii of 3.9, 1.5 and 4.5 nm respectively. Although an aggregate with a size of 15 nm is present in the starting HSA sample, both HEWL and IgG2 have no detectable levels of aggregates. Upon agitation, the mean aggregate size increased with time. There is a stepwise increase in hydrodynamic radius (RH) of oligomeric HSA (Fig. 4A) but such oligomers were not detectable for both HEWL and IgG2. Rather, these two proteins formed directly into aggregates of large sizes (44 and 128 nm) (Fig. 4B and C). Therefore, the sequential event for formation of soluble aggregates is apparently different for these three model proteins. The size of the first detectable HSA aggregate is around 15 nm, while those for HEWL and IgG2 are 44 nm and 128 nm respectively, corresponding to roughly 30, 13,200 and 7800 molecules assuming spherical shapes. 3.2. Reversibility of protein aggregates/particles

Fig. 1. Agitation-induced aggregation kinetics of HSA (j), HEWL (s) and IgG2 (4) monitored by SEC method.

To understand the stability of protein particles formed under agitation stress condition, the reversibility of these particles was examined upon dilution in SEC mobile phase buffer using HIAC particle counting method (Fig. 5A–C). Dilution of the 48-h agitated samples by 5 fold reduced the number of HSA particles by 38% in 5 h. In contrast, HEWL and IgG2 samples showed 2 and 3 fold increase in the number of particles under the same conditions. Dilution experiments carried out using formulation buffer (acetate pH 5.5) showed a 2–3 fold increase in the number of HSA particles

302

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Fig. 2. Agitation-induced formation of subvisible particles, measured using HIAC, is shown in left panel and the corresponding mass distribution is shown in right panel for HSA (A and D), HEWL (B and E) and IgG2 (C and F). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. Representative images of subvisible particles formed under agitation captured using a MFI system from Brightwell. The scale bar represents 25 lm.

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

303

observed in HSA samples, suggesting dye binding to hydrophobic patches while HEWL and IgG2 samples showed minimal fluorescence (1.8 ± 0.5 a.u.). It is possible that dye binding may be proportional to the size of the hydrophobic patch. Upon agitation, the fluorescence increased significantly with time for HSA, suggesting an increase in surface hydrophobicity of HSA monomers and/or enhanced binding of oligomeric intermediates. In contrast, no significant change was observed for the other two proteins. Taken together, Sypro orange fluorescence results show that the agitation-induced formation of protein particles does not involve significant exposure of additional hydrophobic dye binding sites at least for HEWL and IgG2. 3.4. Thermal stability of model proteins before and after agitation

Fig. 4. Agitation-induced aggregation of HSA, HEWL and IgG2, revealed by DLS. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

upon dilution but approximately 50% decrease in the number of HEWL or IgG2 particles (Fig. 5). These results suggest that protein particles formed during agitation may remain the same, dissociate, or grow under certain solution conditions. 3.3. Surface hydrophobic properties of model proteins upon agitation The surface hydrophobic properties calculated using MOE method are shown in Fig. S1 and Table 2. The total SAS hydrophobic surface areas are IgG2 > HSA > HEWL, while HSA has the largest hydrophobic patch, followed by IgG2 and HEWL. The surface hydrophobic properties of these model proteins upon agitation were measured by SYPRO orange fluorescence. Fig. 6 shows the relative fluorescence intensity of protein samples with time during agitation. At time zero, significant fluorescence (85 ± 8 a.u.) was

The thermal stability of the three model proteins before and after agitation was compared by both DSC and DSF (Fig. 7). The DSC curve of HSA before agitation showed three unfolding transitions with Tm1 at 56.6 °C, Tm2 at 70.1 °C and Tm3 at 80.4 °C respectively. The corresponding transitions of the 7-h agitated sample were observed at 56.7 °C, 70.0 °C and 82.3 °C, respectively. The agitation stress did not alter the Tm and enthalpy (DH) values of the first two transitions, however, the Tm3 was slightly increased with a significantly lower DH. The DSF curve of the same HSA sample exhibited three unfolding transitions (Th – hydrophobic melting temperature) at 58.7 °C, 68.9 °C and 77.0 °C, respectively. The 7-h agitated sample exhibited similar values but the fluorescence intensity was significantly increased (22%). Control experiments carried out with free dye exhibited very minimal fluorescence change at the melting temperature in the DSF measurements. The DSC thermogram of HEWL showed a single melting transition at 76.6 °C. There was no difference in Tm and DH values between the unagitated and 7-h agitated HEWL samples. Similarly, a single unfolding transition was observed at 72.3 °C by DSF measurement, which is a few degrees lower than that determined by DSC. In addition, the relative fluorescence intensities at the onset unfolding temperature were about 7% higher for the agitated HEWL sample compared to the unagitated sample. The DSC thermogram of IgG2 molecule exhibited three transitions at 72.9 °C, 82.1 °C and 87.1 °C, respectively. Agitation for 7 h showed a negligible difference in enthalpy and melting temperatures. DSF data of IgG2 showed similar number of transitions but the Th values are 2–6 °C lower than the DSC data. Similar to HEWL, the agitated IgG2 sample exhibited a higher fluorescence intensity at the onset unfolding temperature. The significant increase in fluorescence intensity for HSA (Table 1) suggests an increase in exposed hydrophobic surfaces as a result of agitation stress. Among the three model proteins, HSA had the lowest Tm value (56.6 °C) (Table 2), suggesting that it is the least thermally stable protein, at least for one domain. The Tm1 value of HSA obtained by DSC is 2.2 °C lower than the hydrophobic melting temperature by DSF (Fig. 8). In comparison, the highest Tm of HSA and all the Tm’s of the other two model proteins are a few degree higher than those by DSF. These differences in the thermal unfolding pattern are possibly due to the variations in hydrophobic exposure and unfolding of different domains. 4. Discussion Physical instability of different proteins may result in formation of soluble and insoluble aggregates in liquid protein formulations. However, the exact mechanism of agitation-induced aggregation at the air–water interface remains unclear. To better understand this process, we have employed multiple techniques on three model proteins, HSA, HEWL and IgG2. These proteins have diverse

304

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Fig. 5. Reversibility of HSA (A and D), HEWL (B and E) and IgG2 (C and F) particles formed during agitation on 5 fold dilution either in 200 mM sodium phosphate buffer (pH 7.0) containing 50 mM NaCl (A–C) or in 10 mM sodium acetate formulation buffer (pH 5.5) (D–F).

structural and physical properties (Molecular weight, pI, surface hydrophobicity and thermal stability). Moreover, these model proteins have been previously investigated for their aggregation behavior under agitation and other stress conditions [24–26].

4.1. Mass balance during agitation All the model proteins lost a significant amount of monomer with formation of soluble aggregates (HSA and HEWL only) and

305

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309 Table 1 Average fluorescence intensities from DSF melting curves of the unagitated and 7-h agitated samples of HSA, HEWL and IgG2. Protein

Agitation time (h)

Average (three samples) ± SD

HSA

0 7

4470 ± 112 5493 ± 260

HEWL

0 7

2023 ± 4 2018 ± 3

IgG2

0 7

2071 ± 9 2069 ± 14

subvisible particles upon agitation. When the percentage of remaining protein monomers is compared with the sum of soluble aggregates by SEC-HPLC and insoluble proteins converted into subvisible particles detected by HIAC, it is clear that the total number does not add up to 100%. The higher percentage of monomer loss than what can be accounted for by SEC and HIAC could be attributed to several possibilities. First, particles of submicron range (<1.5 lm in size) are not measurable by the HIAC method (Table 2). Second, particles were assumed to be spherical in calculating the amount of protein in particles, but the real volume fraction may vary widely [4]. More structural and morphological information of aggregates is required to calculate the true mass. The last possibility is the underestimation of the number of subvisible particles by HIAC due to failure to capture translucent protein particles. It has been reported earlier that HIAC is less sensitive to such particles in the 2–10 micron size range and may possibly lead to underestimation relative to MFI [27]. 4.2. Agitation-induced aggregation process It has been generally believed that agitation induces protein aggregation due to unfolding or partial unfolding of the protein at the air–water interface [14,25,28,29]. The influence of agitation on the aggregation tendency may vary, depending on the molecular weight, domain structural stability and spatial distribution of hydrophobic sites in proteins [30,31]. Our results demonstrate that all three model proteins formed subvisible particles upon agitation. Is there any commonality in aggregation pathway which involves a stepwise conversion of monomers and oligomers into subvisible particles in different proteins? Our SEC and DLS results suggest that the mechanism of subvisible particle formation may vary depending upon the protein under a similar stress condition. HSA produced oligomeric intermediates (15 nm in size) with a stepwise increase in size as monitored by DLS and showed a clear increasing trend in aggregate level as monitored by SEC. In comparison, HEWL produced oligomeric aggregates in the size range of 44 nm or higher and subsequently produced protein aggregates in the submicron range (300 nm), whereas, the IgG2 molecule produced >128 nm aggregates immediately after agitation. One, however, has to

understand that these size estimations as measured by DLS may not be very accurate, as the DLS method cannot distinguish well protein monomers from dimers and trimers in solution and is semi-quantitative in estimating higher molecular weight aggregates [32,33]. Nevertheless, the lack of smaller oligomeric aggregates (<44 nm) in agitated HEWL and IgG2 samples as monitored by DLS is supported by minimal or no detection of oligomeric aggregates as monitored by SEC. The differences in particle size distribution among the three model proteins in the initial 9-h agitation period suggest a variation in the fast phase of assembly process (Fig. 2). It is clear from the mass data that a significant amount of particles in the smaller range of 1.5–10 lm were formed initially in HEWL, whereas HSA and IgG2 formed larger ones in the range of 10–25 lm upon agitation. The presence of smaller particle size distribution (610 lm) in the HEWL sample at the early agitation time period (4 h) could be possibly due to the lower molecular mass of the monomer, since assembly of the same number of protein monomers would generate a particle of a smaller size. The size difference might contribute to the higher rate of formation of subvisible particles estimated for HSA and IgG than that for HEWL (Fig. 9). Although much research work has been published on subvisible particle formation under different stress conditions, the exact mechanism of conversion of submicron-sized aggregates/particles into larger particles is not clearly understood. Results from this study and other earlier reports suggest that subvisible particle formation is a complex process and it may take multiple routes with or without soluble aggregates, depending upon the type of protein and the nature of stress conditions applied [4,12,19,34]. It is very likely that particle formation occurs through conversion of oligomeric aggregates into lower micron-sized particles followed by their condensation into larger ones [13,19]. A condensation process is likely responsible for the clear growth of smaller subvisible particles into larger ones (P25 and 680 lm) with time for HSA and HEWL, while such a process seems to be less significant for IgG2. The agitation-induced formation of subvisible particles shows apparent fast and slow phases with time. If unfolding or partial unfolding of proteins at air/water interfaces initiates the aggregation process, the rate and extent of protein aggregation may be dictated by the rate of protein accumulation at the interface and the total area of the interface. We speculate that the slow phase may reflect a period when the air–water interfaces are saturated with protein aggregates, which would prevent fast formation of new aggregates. 4.3. Protein surface hydrophobicity and aggregation The hydrophobicity of a protein can be characterized in terms of overall sequence hydrophobicity and solvent accessible surface (SAS) hydrophobicity. Protein surface hydrophobicity has been suggested to play an important role in protein aggregation [35].

Table 2 Summary of molecular properties and agitation-related findings for the three model proteins.

a b c

Protein

pI

Tm

HSA

4.7

56.6, 70.1, 80.4

HEWL

11

76.6

IgG2

8.6

72.9, 82.1, 87.1

SbVP formationb, %/h (nM/h)

Total SAS hydrophobicity (Å2)

Normalized hydrophobicityc (Å2)

Largest hydrophobic patch (Å2)

Soluble oligomer

Stepwise formation of submicron aggregate

Reversibility at pH 7/pH 5.5

0.16 ± 0.03

0.0145 ± 0.005 (80.5)

17,060

0.1137

292.3

+

+

Yes/No

0.02 ± 0.002

0.0044 ± 0.0006 (177)





No/Yes





No/Yes

HMMS formationa, %/h

0.006 ± 0.001

0.018 ± 0.003 (46.6)

3279

0.022

31,310

0.208

Rates were obtained from first 9 h of agitation. Rates were obtained from first 9 h of agitation for particles of 1.5–80 lm. Calculated based on the size of the largest protein.

68.48 132.3

306

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Fig. 6. Sypro Orange fluorescence intensity change with agitation time in HSA (j), HEWL (s) and IgG2 (4).

Similarly, protein subvisible particles are able to form and grow by exposed hydrophobic surfaces [20,36]. Our estimated, size-normalized total hydrophobic SAS for the three model proteins, HSA, HEWL and IgG2 are 0.114 Å2, 0.022 Å2 and 0.208 Å2, respectively, based on their actual hydrophobic SAS of 1.71E4 Å2, 3.28E3 Å2 and 3.13E4 Å2 (Table 2). It is obvious that IgG2 has the largest SAS surface area. Based on these values, the aggregation or particle formation rate would be highest for IgG2, if surface hydrophobicity controls the aggregation process. In agreement with this, the amount of protein estimated in the form of subvisible particles is highest for IgG2, (0.2%). On the other hand, the corresponding amounts for HEWL and HSA are similar (both 0.1%), although their surface hydrophobicity is significantly different. Therefore, the relationship between surface hydrophobicity and aggregation/particle formation needs further evaluation. The Sypro-Orange fluorescence results indicate that HSA in native state (fatty acid free version) showed significant dye binding relative to HEWL and IgG2. Obviously, the highest dye-binding result does not match the protein size-normalized surface hydrophobicity estimated by computer modeling (Table 2). These results may suggest possible contribution of other factors to protein-dye binding, such as the size of SAS hydrophobic sites [37]. It was reported that HSA in native state has a number of high affinity dye binding hydrophobic cavities [37,38], Therefore, we also compared the relative sizes of the individual surface hydrophobic sites in all three proteins and HSA does have the largest hydrophobic site (292.29 Å2). It is thus possible that a threshold area of a hydrophobic site is needed for efficient dye binding in proteins and HSA meets this requirement. HSA did generate the highest percentage of soluble aggregates upon agitation. It remains to be determined whether the formation of soluble aggregates upon agitation has any relationship with the size of the hydrophobic sites. It is likely that forces other than surface hydrophobicity also play a significant role in aggregation/particle formation. One such force would be electrostatic interactions. These three proteins have pI’s HSA (4.7), HEWL (11.1) and IgG2 (8.6). At the formulation pH used in this study (pH 5.5), the net pH difference from the pI for the three model proteins is +0.8, 5.5, and 3.1, respectively. The experimental solution pH is apparently closer to the pI of HSA, and we speculate that chance of attractive charge-charge interactions in HSA is higher than the other two proteins, which could contribute formation of soluble aggregates [39,40]. Agitation induced apparent exposure of additional hydrophobic surfaces in HSA, as evidenced by the enhanced Sypro Orange

Fig. 7. Comparison of DSC thermograms (left Y-axis) with DSF melting curves (right Y-axis) of HSA (A), HEWL (B) and IgG2 (C). DSC thermogram of native (—) and 7-h agitated samples (-h-h-). DSF melting curves of native (ssss) and 7-h agitated samples (dddd).

fluorescence intensity. The increased availability of hydrophobic sites could be due to either agitation-induced structural perturbation with additional exposure of hydrophobic surfaces in protein monomers, and/or availability of more hydrophobic sites of the freshly-formed oligomeric aggregates. Since no significant change was observed in intrinsic fluorescence (data not shown), it is likely that the enhanced fluorescence is due to contribution of the latter. In addition, since the rate of aggregation or particle formation in HSA did not seem to change after agitation, suggesting that the availability of additional hydrophobic sites may be contributed

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

Fig. 8. Comparison of melting temperatures from DSF method (Th) and DSC method (Tm). HSA (j), HEWL (d) and IgG2 (N).

mainly from the soluble oligomers. The DSF data with increased fluorescence intensity at the melting temperature in stressed samples also suggest exposure of buried hydrophobic sites upon unfolding [41,42]. As discussed above, HSA generated the largest amount of soluble aggregates while IgG2 generated the highest amount of protein particles upon agitation among the three model proteins. These results suggest no relationship between the amount of stable oligomeric intermediate species and number of particles produced. 4.4. Thermal stability and protein aggregation The three Tm values of HSA were observed at 56.7 °C, 70.1 °C and 80.4 °C by DSC. HSA molecule is composed of three homologous domains, which are typically labeled as I, II and III. These domains are subdivided into two sub-domains (A and B), which share common structural elements [43]. Typically two to three thermal transitions were observed for fatty acid free form of HSA and our present results are in agreement with the reported values [44]. A single thermal unfolding transition was observed for HEWL at 76 °C, which is also consistent with the reported value [45]. The thermograms of IgG2 sample showed three transitions at 72.9 °C, 82.1 °C and 87.1 °C, corresponding to CH2 and Fab/CH3 domains as reported [46]. The Tm1 of HSA is 8–10 °C lower than Tm of HEWL and Tm1 of IgG2, and HSA generated the highest amount of oligomeric aggregates upon agitation, although only a moderate amount of protein aggregates was converted into particles. It is possible that the thermal stability of a protein might play a certain

307

Fig. 9. Aggregation measured by SEC (open symbols) and HIAC (closed symbols) – HSA (h, j), HEWL (s, d) and IgG2 (4, N). The rate of particle formation calculated by fitting the subvisible particle mass data of HSA, 0.0145 ± 0.005% h1; HEWL, 0.0044 ± 0.0006% h1 and IgG2, 0.018 ± 0.003% h1 respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

role in controlling their relative tendency of oligomeric aggregate formation upon agitation. Agitation was found to have minimal effects on the thermal stability of these proteins. The DSC results revealed no change in Tm and DH values between the starting and 7-h agitated HEWL and IgG2 samples. Strong intermolecular structural hierarchy and stability were indeed reported in HEWL [45,47]. However, there was a significant reduction in the DH value of the Tm3 of HSA upon agitation. This result suggests notable unfolding of the protein under agitation [48], and likelihood of involvement of the partially unfolded and structurally perturbed monomers in the aggregation process [49]. As stated above, the relative thermal stability of a protein could contribute to the tendency of protein aggregation upon agitation. On the other hand, the role of thermal stability and hydrophobic surface characteristics of a protein in protein aggregation has been challenged [50]. Sypro Orange is widely used to determine the degree of surface exposed hydrophobic region in a protein using steady state fluorescence or differential scanning fluorimetry (DSF) [51]. Thermally-induced protein unfolding typically results in exposure of hydrophobic sites exhibiting an acceleration in dye binding by DSF, followed by a deceleration phase after the maximal unfolding temperature due to aggregation or disintegration of hydrophobic patches [52]. The Tm1 value observed for HSA by DSC is 2.2 °C lower than that by DSF, whereas, the Tm value was 4.3 and

Fig. 10. Schematic view of aggregation pathway leading to subvisible particles under agitation in proteins. HSA forms oligomeric aggregate intermediates before formation of submicron-sized aggregates (Path A), while HEWL and IgG2 forms directly submicron-sized aggregates (Path B). Further condensation/growth of these aggregates lead to formation of larger subvisible particles. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

308

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309

5.4 °C higher for HEWL and IgG by DSC than those by DSF. It has been reported earlier that a lower Tm by DSF (than DSC) could be due to an earlier occurrence of exposure of hydrophobic sites than domain unfolding [53]. This would explain the trend observed for HEWL and IgG2 in this study. However, in HSA, the thermal unfolding precedes the hydrophobic exposure event. The opposite sequential events could be possibly explained by several factors including surface exposed hydrophobic sites on the protein and their aggregation at elevated temperatures. As discussed, HSA has the largest size of hydrophobic sites for effective dye binding and thermal protein unfolding may not lead to exposure of additional large-sized hydrophobic sites. Further, thermal unfolding could lead to formation of transient but stable oligomeric states (HSA) or unstable and aggregation-prone intermediate states (HEWL and IgG2) [54]. 4.5. Reversibility of protein particles Reversible aggregation has been reported typically for aggregates formed by relatively weak non-covalent interactions [55]. Formation of reversible protein subvisible particles is possible if soluble oligomers are formed prior to particle formation [56]. In this study, it was found that approximately 38% of HSA subvisible particles were dissociated after 24 h incubation in the SEC-HPLC mobile phase (pH 7) at room temperature. At pH 5.5, the reversibility was not observed, and in fact, more subvisible particles were observed. The reduction in the number of particles in HSA samples at pH 7 suggests that these protein particles formed are not stable enough and may dissociate/associate depending on the solution conditions. A higher solution pH can certainly facilitate repulsive protein–protein interactions in HSA and possibly, lead to formation of stable oligomeric intermediates as reported in the literature [56]. Increase in the number of HSA particles in the formulation buffer suggests possible involvement of HSA self stabilization at a higher concentration. In contrast, the number of HEWL and IgG2 particles actually increased significantly upon dilution into the neutral pH buffer, suggesting that the diluted solution conditions favor aggregation and particle formation. Particles diluted in formulation buffer exhibited an opposite trend compared to neutral pH buffer, suggesting reversible nature of particles. It is well known that manipulation of the solution conditions (pH, ionic strength, etc.) can either favor dissociation or growth of the soluble aggregates obtained under various stress conditions [20,57]. Our study demonstrates reversibility of protein particles, and this may help to understand the entire protein aggregation process. Reversibility of agitation-induced particle formation was also observed for an IgG1 antibody [58]. 4.6. Proposed aggregation model Based on what we found in this study and previous reports, we propose a model (Fig. 10) that describes the protein aggregation during the agitation process [13]. Agitation leads to partial folding of proteins at the air/water interface. Self association of the partially unfolded monomers results in formation of either oligomeric intermediates as observed for HSA (Path A, Fig. 10) or larger, submicron-range aggregates for HEWL and IgG2 (Path B). Under Path B, proteins (HEWL and IgG2) have minimal exposure of hydrophobic surfaces and negligible amount of soluble oligomers that would not favor reversible formation of the oligomeric intermediate states. Further condensation/growth of the HSA oligomeric intermediates leads to formation of larger, submicron-range aggregates, which continue to grow into micron-sized subvisible particles. Structural characterization of the proteins by intrinsic fluorescence and circular dichroism (data not shown) revealed no evidence of a structural change during agitation, suggesting that majority of the

protein population is in near native state during agitation, while a small fraction initiates the aggregation process [59]. In summary, this study confirms that agitation can induce particle formation in three model proteins. However, the particle-formation process may be different depending on the protein. In some cases, protein subvisible particle formation may go through a transient oligomeric state, and in other instances, formation of oligomeric intermediates may not be an obligatory step. Morphology and reversibility of protein subvisible particles may depend on both the protein and the actual experimental conditions. The overall surface hydrophobicity of a protein is not necessarily an indicator for relative tendency of agitation-induced aggregation. Similarly, the relative thermal stability (Tm) of a protein may not correlate with its aggregation tendency during agitation [60]. The aggregation process is likely dependent on a number of factors – the number and size of the individual hydrophobic site, thermal stability, charge distribution, and the nature of protein– protein interactions. The reversibility of HSA particles and its possible self stabilization were demonstrated, which could be related to the surface activity of the protein [13,14]. Since the aggregation tendency and surface activity are strongly influenced by the solution conditions, other solution-related factors should be evaluated to get a comprehensive view of the agitation-induced particle formation. Further research work is in progress to understand the role of these variables. Acknowledgements The authors thank Drs. Bilikallahalli Muralidhara and Satish Singh for sharing their view of the project direction and Maria Toler for her technical assistance in particle characterization. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ejpb.2014.01.004. References [1] W. Wang, S. Nema, D. Teagarden, Protein aggregation–pathways and influencing factors, Int. J. Pharm. 390 (2010) 89–99. [2] A. Hawe, J.C. Kasper, W. Friess, W. Jiskoot, Structural properties of monoclonal antibody aggregates induced by freeze-thawing and thermal stress, Eur. J. Pharm. Sci. 38 (2009) 79–87. [3] A. Nayak, J. Colandene, V. Bradford, M. Perkins, Characterization of subvisible particle formation during the filling pump operation of a monoclonal antibody solution, J. Pharm. Sci. 100 (2011) 4198–4204. [4] J.G. Barnard, S. Singh, T.W. Randolph, J.F. Carpenter, Subvisible particle counting provides a sensitive method of detecting and quantifying aggregation of monoclonal antibody caused by freeze-thawing: insights into the roles of particles in the protein aggregation pathway, J. Pharm. Sci. 100 (2011) 492–503. [5] P.K. Gupta, E. Porembski, N.A. Williams, Approaches to reducing subvisible particle counts in lyophilized parenteral formulations, J. Pharm. Sci. Technol. 48 (1994) 30–37. [6] T.M. Scherer, S. Leung, L. Owyang, S.J. Shire, Issues and challenges of subvisible and submicron particulate analysis in protein solutions, AAPS J. 14 (2012) 236– 243. [7] W. Wang, S.K. Singh, N. Li, M.R. Toler, K.R. King, S. Nema, Immunogenicity of protein aggregates – concerns and realities, Int. J. Pharm. 431 (2012) 1–11. [8] L. Curatolo, B. Valsasina, C. Caccia, G.L. Raimondi, G. Orsini, A. Bianchetti, Recombinant human IL-2 is cytotoxic to oligodendrocytes after in vitro self aggregation, Cytokine 9 (1997) 734–739. [9] S. Poon, N.R. Birkett, S.B. Fowler, B.F. Luisi, C.M. Dobson, J. Zurdo, Amyloidogenicity and aggregate cytotoxicity of human glucagon-like peptide-1 (hGLP-1), Protein Pept. Lett. 16 (2009) 1548–1556. [10] R. Kayed, E. Head, J.L. Thompson, T.M. McIntire, S.C. Milton, C.W. Cotman, C.G. Glabe, Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis, Science 300 (2003) 486–489. [11] J.F. Carpenter, T.W. Randolph, W. Jiskoot, D.J. Crommelin, C.R. Middaugh, G. Winter, Y.X. Fan, S. Kirshner, D. Verthelyi, S. Kozlowski, K.A. Clouse, P.G. Swann, A. Rosenberg, B. Cherney, Overlooking subvisible particles in therapeutic protein products: gaps that may compromise product quality, J. Pharm. Sci. 98 (2009) 1201–1205.

M. Jayaraman et al. / European Journal of Pharmaceutics and Biopharmaceutics 87 (2014) 299–309 [12] J.F. Carpenter, M.C. Manning, T.W. Randolph, Long-term storage of proteins, in: J.E. Coligan (Ed.), Current Protocols in Protein Science, Wiley, New York, 2002, pp. 4.6.1–4.6.6. [13] H.C. Mahler, R. Muller, W. Friess, A. Delille, S. Matheus, Induction and analysis of aggregates in a liquid IgG1-antibody formulation, Eur. J. Pharm. Biopharm. 59 (2005) 407–417. [14] S. Kiese, A. Papppenberger, W. Friess, H.C. Mahler, Shaken, not stirred: mechanical stress testing of an IgG1 antibody, J. Pharm. Sci. 97 (2008) 4347– 4366. [15] A. Eppler, M. Weigandt, A. Hanefeld, H. Bunjes, Relevant shaking stress conditions for antibody preformulation development, Eur. J. Pharm. Biopharm. 74 (2010) 139–147. [16] J.F. Carpenter, B.S. Kendrick, B.S. Chang, M.C. Manning, T.W. Randolph, Inhibition of stress-induced aggregation of protein therapeutics, Methods Enzymol. 309 (1999) 236–255. [17] W. Wang, Protein aggregation and its inhibition in biopharmaceutics, Int. J. Pharm. 289 (2005) 1–30. [18] B. Demeule, R. Gurny, T. Arvinte, Detection and characterization of protein aggregates by fluorescence microscopy, Int. J. Pharm. 329 (2007) 37–45. [19] H.C. Mahler, W. Friess, U. Grauschopf, S. Kiese, Protein aggregation: pathways, induction factors and analysis, J. Pharm. Sci. 98 (2009) 2909–2934. [20] M.K. Joubert, Q. Luo, Y. Nashed-Samuel, J. Wypych, L.O. Narhi, Classification and characterization of therapeutic antibody aggregates, J. Biol. Chem. 286 (2011) 25118–25133. [21] Molecular Operating Environment (MOE), Chemical Computing Group Inc., 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2011. [22] S.A. Wildman, G.M. Crippen, Prediction of physiochemical parameters by atomic contributions, J. Chem. Inf. Comput. Sci. 39 (1999) 868–873. [23] S.K. Singh, M.R. Toler, Monitoring of subvisible particles in therapeutic proteins, Methods Mol. Biol. 899 (2012) 379–401. [24] L. Hu, C. Olsen, N. Maddux, S.B. Joshi, D.B. Volkin, C.R. Middaugh, Investigation of protein conformational stability employing a multimodal spectrometer, Anal. Chem. 83 (2011) 9399–9405. [25] R.M. Fesinmeyer, S. Hogan, A. Saluja, S.R. Brych, E. Kras, L.O. Narhi, D.N. Brems, Y.R. Gokarn, Effect of ions on agitation- and temperature-induced aggregation reactions of antibodies, Pharm. Res. 26 (2009) 903–913. [26] A. Saluja, R.M. Fesinmeyer, S. Hogan, D.N. Brems, Y.R. Gokarn, Diffusion and sedimentation interaction parameters for measuring the second virial coefficient and their utility as predictors of protein aggregation, Biophys. J. 99 (2010) 2657–2665. [27] D.K. Sharma, D. King, P. Oma, C. Merchant, Micro-flow imaging: flow microscopy applied to sub-visible particulate analysis in protein formulations, AAPS J. 12 (2010) 455–464. [28] J. Zhang, E.M. Topp, Protein G, protein A and protein A-derived peptides inhibit the agitation induced aggregation of IgG, Mol. Pharm. 9 (2012) 622–628. [29] R. Thirumangalathu, S. Krishnan, M.S. Ricci, D.N. Brems, T.W. Randolph, J.F. Carpenter, Silicone oil- and agitation-induced aggregation of a monoclonal antibody in aqueous solution, J. Pharm. Sci. 98 (2009) 3167–3181. [30] S.A. Abbas, V.K. Sharma, T.W. Patapoff, D.S. Kalonia, Opposite effects of polyols on antibody aggregation: thermal versus mechanical stresses, Pharm. Res. 29 (2012) 683–694. [31] N. Chennamsetty, V. Voynov, V. Kayser, B. Helk, B.L. Trout, Design of therapeutic proteins with enhanced stability, Proc. Natl. Acad. Sci. USA 106 (2009) 11937–11942. [32] J.S. Philo, Is any measurement method optimal for all aggregate sizes and types?, AAPS J 8 (2006) E564–571. [33] K. Ahrer, A. Buchacher, G. Iberer, A. Jungbauer, Detection of aggregate formation during production of human immunoglobulin G by means of light scattering, J. Chromatogr. A 1043 (2004) 41–46. [34] E.Y. Chi, S. Krishnan, T.W. Randolph, J.F. Carpenter, Physical stability of proteins in aqueous solution: mechanism and driving forces in nonnative protein aggregation, Pharm. Res. 20 (2003) 1325–1336. [35] L. Zhang, D. Lu, Z. Liu, How native proteins aggregate in solution: a dynamic Monte Carlo simulation, Biophys. Chem. 133 (2008) 71–80. [36] J. Park, K. Nagapudi, C. Vergara, R. Ramachander, J.S. Laurence, S. Krishnan, Effect of pH and excipients on structure, dynamics, and long-term stability of a model IgG1 monoclonal antibody upon freeze-drying, Pharm. Res. 30 (2013) 968–984.

309

[37] B. Ahmad, M.Z. Ahmed, S.K. Haq, R.H. Khan, Guanidine hydrochloride denaturation of human serum albumin originates by local unfolding of some stable loops in domain III, Biochim. Biophys. Acta 1750 (2005) 93–102. [38] A. Hawe, M. Sutter, W. Jiskoot, Extrinsic fluorescent dyes as tools for protein characterization, Pharm. Res. 25 (2008) 1487–1499. [39] S. Yadav, J. Liu, S.J. Shire, D.S. Kalonia, Specific interactions in high concentration antibody solutions resulting in high viscosity, J. Pharm. Sci. 99 (2010) 1152–1168. [40] S. Yadav, S.J. Shire, D.S. Kalonia, Viscosity behavior of high-concentration monoclonal antibody solutions: correlation with interaction parameter and electroviscous effects, J. Pharm. Sci. 101 (2012) 998–1011. [41] C. Munch, A. Bertolotti, Exposure of hydrophobic surfaces initiates aggregation of diverse ALS-causing superoxide dismutase-1 mutants, J. Mol. Biol. 399 (2010) 512–525. [42] V. Kayser, N. Chennamsetty, V. Voynov, B. Helk, B.L. Trout, Conformational stability and aggregation of therapeutic monoclonal antibodies studied with ANS and thioflavin T binding, MAbs 3 (2011) 408–411. [43] S. Curry, P. Brick, N.P. Franks, Fatty acid binding to human serum albumin: new insights from crystallographic studies, Biochim. Biophys. Acta 1441 (1999) 131–140. [44] A. Michnik, K. Michalik, A. Kluczewska, Z. Drzazga, Comparative DSC study of human and bovine serum albumin, J. Therm. Anal. Calorim. 84 (2006) 113– 117. [45] S. Arai, M. Hirai, Reversibility and hierarchy of thermal transition of hen eggwhite lysozyme studied by small-angle X-ray scattering, Biophys. J. 76 (1999) 2192–2197. [46] E. Garber, S.J. Demarest, A broad range of Fab stabilities within a host of therapeutic IgGs, Biochem. Biophys. Res. Commun. 355 (2007) 751–757. [47] J.M. Sanchez-Ruiz, Probing free-energy surfaces with differential scanning calorimetry, Annu. Rev. Phys. Chem. 62 (2011) 231–255. [48] B. Farruggia, F. Rodriguez, R. Rigatuso, G. Fidelio, G. Pico, The participation of human serum albumin domains in chemical and thermal unfolding, J. Protein Chem. 20 (2001) 81–89. [49] A.W. Vermeer, W. Norde, The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein, Biophys. J. 78 (2000) 394–404. [50] K. Ahrer, A. Buchacher, G. Iberer, A. Jungbauer, Thermodynamic stability and formation of aggregates of human immunoglobulin G characterised by differential scanning calorimetry and dynamic light scattering, J. Biochem. Biophys. Methods 66 (2006) 73–86. [51] M. Gunther, J. Lipka, A. Malek, D. Gutsch, W. Kreyling, A. Aigner, Polyethylenimines for RNAi-mediated gene targeting in vivo and siRNA delivery to the lung, Eur. J. Pharm. Biopharm. 77 (2011) 438–449. [52] F.H. Niesen, H. Berglund, M. Vedadi, The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability, Nat. Protoc. 2 (2007) 2212–2221. [53] F. He, S. Hogan, R.F. Latypov, L.O. Narhi, V.I. Razinkov, High throughput thermostability screening of monoclonal antibody formulations, J. Pharm. Sci. 99 (2010) 1707–1720. [54] T.R. Jahn, S.E. Radford, Folding versus aggregation: polypeptide conformations on competing pathways, Arch. Biochem. Biophys. 469 (2008) 100–117. [55] L.O. Narhi, J. Schmit, K. Bechtold-Peters, D. Sharma, Classification of protein aggregates, J. Pharm. Sci. 101 (2012) 493–498. [56] J.S. Philo, T. Arakawa, Mechanisms of protein aggregation, Curr. Pharm. Biotechnol. 10 (2009) 348–351. [57] W.G. Lilyestrom, S. Yadav, S.J. Shire, T.M. Scherer, Monoclonal antibody selfassociation, cluster formation, and rheology at high concentrations, J. Phys. Chem. B 117 (2013) 6373–6384. [58] S. Kiese, A. Pappenberger, W. Friess, H.C. Mahler, Equilibrium studies of protein aggregates and homogeneous nucleation in protein formulation, J. Pharm. Sci. 99 (2010) 632–644. [59] E. Sahin, A.O. Grillo, M.D. Perkins, C.J. Roberts, Comparative effects of pH and ionic strength on protein–protein interactions, unfolding, and aggregation for IgG1 antibodies, J. Pharm. Sci. 99 (2010) 4830–4848. [60] D.D. Banks, R.F. Latypov, R.R. Ketchem, J. Woodard, J.L. Scavezze, C.C. Siska, V.I. Razinkov, Native-state solubility and transfer free energy as predictive tools for selecting excipients to include in protein formulation development studies, J. Pharm. Sci. 101 (2012) 2720–2732.