Pharm Res (2016) 33:716–728
`DOI 10.1007/s11095-015-1821-0
`
`RESEARCH PAPER
`
`Monoclonal Antibodies Follow Distinct Aggregation Pathways
`D u r i n g P r o d u c t i o n - R e l e v a n t A c i d i c I n c u b a t i o n
`and Neutralization
`
`Thomas Skamris 1 & Xinsheng Tian 1 & Matthias Thorolfsson 2 & Hanne Sophie Karkov 2 & Hanne B. Rasmussen 2 & Annette E. Langkilde 1 &
`Bente Vestergaard 1
`
`Received: 25 June 2015 / Accepted: 29 October 2015 /Published online: 12 November 2015
`# Springer Science+Business Media New York 2015
`
`ABSTRACT
`Purpose Aggregation aspects of therapeutic monoclonal an-
`tibodies (mAbs) are of common concern to the pharmaceutical
`industry. Low pH treatment is applied during affinity purifi-
`cation and to inactivate endogenous retroviruses, directing
`interest to the mechanisms of acid-induced antibody
`aggregation.
`Methods We characterized the oligomerization kinetics at
`pH 3.3, as well as the reversibility upon neutralization, of
`three model mAbs with identical variable regions, representa-
`tive of IgG1, IgG2 and IgG4 respectively. We applied size-
`exclusion high performance liquid chromatography and or-
`thogonal analytical methods, including small-angle X-ray
`scattering and dynamic light scattering and supplemented
`the experimental data with crystal structure-based spatial ag-
`gregation propensity (SAP) calculations.
`Results We revealed distinct solution behaviors between the
`three mAb models: At acidic pH IgG1 retained monomeric,
`whereas IgG2 and IgG4 exhibited two-phase oligomerization
`processes. After neutralization, IgG2 oligomers partially
`reverted to the monomeric state, while on the contrary,
`IgG4 oligomers tended to aggregate. Subclass-specific aggre-
`gation-prone motifs on the Fc fragments were identified,
`
`Thomas Skamris and Xinsheng Tian contributed equally to this work.
`
`Electronic supplementary material The online version of this article
`(doi:10.1007/s11095-015-1821-0) contains supplementary material, which is
`available to authorized users.
`
`* Bente Vestergaard
`bente.vestergaard@sund.ku.dk
`
`1 Department of Drug Design and Pharmacology, University of
`Copenhagen, Universitetsparken 2, 2100 Copenhagen, Denmark
`2 Global Research Unit, Novo Nordisk A/S, Novo Nordisk Park 1
`2760 Måløv, Denmark
`
`which may lead to two distinct pathways of reversible and
`irreversible aggregation, respectively.
`Conclusions We conclude that subtle variations in mAb se-
`quence greatly affect responses towards low-pH incubation
`and subsequent neutralization, and demonstrate how orthog-
`onal biophysical methods distinguish between reversible and
`irreversible mAb aggregation pathways at early stages of acid-
`ic treatment.
`
`KEY WORDS biopharmaceutics . formulation . monoclonal
`antibody . protein stability . small-angle X-ray scattering (SAXS)
`
`ABBREVIATIONS
`AUC
`Area under the curve
`AUP
`Area under the peak
`DLS
`Dynamic light scattering
`Fab
`Antigen-binding fragment
`Fc
`Crystallizable fragment
`HMWS High molecular weight species
`HPLC
`High-performance liquid chromatography
`I0
`Forward scattering intensity
`Ig
`Immunoglobulin
`mAb
`Monoclonal antibody
`MALS Multi-angle static light scattering
`MW
`Molecular weight
`P(r)
`Pair distance distribution function
`PBS
`Phosphate buffered saline
`PDB
`Protein data bank
`pI
`Isoelectric point
`q
`Length of the scattering vector
`Rg
`Radius of gyration
`Rh
`Hydrodynamic radius
`SAP
`Spatial aggregation propensity
`SASA
`Solvent accessible surface area
`SAXS
`Small-angle X-ray scattering
`SEC
`Size-exclusion chromatography
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00001
`
`

`

`Subclass Specific Acid-Induced Antibody Aggregation
`
`717
`
`Tm
`UV
`
`Melting point
`Ultraviolet
`
`INTRODUCTION
`
`Monoclonal antibodies (mAbs) are widely used for therapeutic
`invention against a variety of diseases, and are the highest
`selling class of biologics in recent years (1). As of May 2015,
`51 mAbs had been approved in either Europe or the United
`States, and 307 mAbs were in various stages of clinical devel-
`opment worldwide (2). A major challenge for biopharmaceu-
`tical industries is the potential aggregation of mAbs during
`manufacturing, formulation, storage and shipment, which
`can impact quality, efficacy and safety of therapeutic antibody
`products (3). Antibody aggregation can be induced by various
`stress factors including heat, pH, ionic strength, light, shaking
`and freezing. Low pH treatment is commonly employed dur-
`ing mAb production, either for virus inactivation or affinity
`purification, and low pH treatment has been suggested as a
`deliberate chemical stressor to mimic in-process aggregate
`formation in the development of aggregate-removal technol-
`ogy (4). Thus, understanding the aggregation mechanisms of
`mAbs during pH shifts is important for rational design of
`manufacturing processes.
`It is well established that mAb aggregation can follow dif-
`ferent pathways, depending on the experimental conditions
`and the solution properties of individual mAbs. Andersen
`et al. (5) reported that under thermal stress, where only the
`CH2 domain is partly unfolded, IgG1 aggregation kinetics
`show two coupled phases. In the first phase the monomers
`rapidly assemble into oligomers, while the coagulation of these
`oligomers governs the second phase, and the aggregation rate
`slows down. Arosio et al. (6) further demonstrated that the
`aggregation kinetics vary depending on the nature of
`destabilizing conditions: At ambient temperatures and non-
`acidic pH values, mAbs form stable oligomers, which can be
`partly reverted to monomers by decreasing the ionic strength.
`However, when the temperature is raised to 37°C and at
`lower pH values, the mAbs irreversibly form larger aggre-
`gates. In addition, different aggregation kinetics were ob-
`served for IgG1 and IgG2.
`Indeed several studies reveal differences in stability of IgG
`subclasses with identical variable regions. The subclass specific
`aggregation propensity has been ranked IgG1 < IgG2 <
`IgG4 at a broad pH range under thermal stress (7–9). As a
`multidomain protein, a mAb can undergo multiple thermal
`unfolding transitions. Differential scanning calorimetry (DSC)
`revealed that the CH2 domain generally has the lowest ther-
`mal stability of the IgG molecule, and plays a major role in
`determining the rate and extent of Fc aggregation (10,11). In
`addition, the absence of CH2 glycans can dramatically de-
`crease the stability of the Fc region as well as the intact
`
`antibody (10). Previously, we compared the CH2 domain
`melting transitions of three IgG subclasses by differential scan-
`ning fluorescence (DSF). Below pH 4 and in the presence of
`100 mM NaCl the order of CH2 stability ranks as IgG1 >
`IgG4 > IgG2. Accordingly, IgG1 exhibits the highest stability
`at acidic pH. In addition, based on small-angle X-ray scatter-
`ing (SAXS) data (7), we observed that IgG1 has intermolecu-
`lar repulsive forces at acidic pH and intermediate ionic
`strength, which likely contributes to its improved stability.
`Chennamsetty et al. (12) identified 14 aggregation prone mo-
`tifs within the IgG1 constant regions, most of which are locat-
`ed at the lower hinge region and CH2 domains. It was dem-
`onstrated that mutations in some of these motifs could en-
`hance the stability of IgG1 (13). Altogether, the stability of
`mAbs with respect to aggregation can be affected by both
`experimental conditions, as well as multiple intrinsic factors
`including the net charge, the specific nature of the Fab do-
`main, surface hydrophobicity and the level of denaturation.
`Here, we investigate how model mAbs, representing three
`different IgG subclasses, differ in their aggregation kinetics
`and in the nature of oligomers that are formed at acidic pH.
`Applying orthogonal biophysical analytical methods, we in-
`vestigate the development over time after acidic incubation,
`and follow how the initial oligomers further aggregate into
`higher molecular weight species (HMWS). In addition, we
`monitor the reversibility of aggregation upon neutralization.
`We show how two fundamentally different aggregation path-
`ways dominate the aggregation kinetics, and how the investi-
`gated model mAbs reveal subclass specific differences.
`
`MATERIALS AND METHODS
`
`Materials
`
`Three humanized monoclonal antibodies (IgG1, IgG2 and
`IgG4) with identical light chains and identical variable region
`of heavy chains were produced as previously described (7).
`The physicochemical properties along with chemical modifi-
`cations were previously characterized and details can be found
`in the original work (7). The frozen aliquots of antibodies,
`stored in PBS at -80°C were thawed upon use and buffer
`exchanged into 5 mM histidine at pH 6.5 using IllustraTM
`NAPTM 5 Columns (GE Healthcare, Uppsala, Sweden).
`The protein concentration was determined based on A280nm
`and then adjusted to 18 mg/mL (protein stock).
`
`Sample Preparation
`
`The acidic samples for the studies of acid-induced antibody
`aggregation were freshly prepared by mixing the protein stock
`with equal volumes of acidic buffer (100 mM Na-Citrate,
`200 mM NaCl, pH 3.1), which results in an effective
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00002
`
`

`

`718
`
`Skamris et al.
`
`concentration of 9 mg/mL and an effective pH of 3.3. The
`samples were incubated at room temperature (25°C) and the
`development was measured by different analytical methods
`over time. The reversibility of antibody oligomer/aggregate
`formation was investigated after neutralizing the acidic sam-
`ples to pH 6.5 by adding 13% v/v of 1 M Tris buffer at pH 9.0.
`The antibody control of each IgG subclass without low pH
`treatment was prepared with the same protein concentration
`and buffer condition as the neutralized samples. Figure 1 de-
`scribes the selected time points for all measurements. In addi-
`tion, SAXS analysis further investigated the effect of sucrose
`on antibody stability using 100 mM Na-Citrate (pH 3.1) con-
`taining 0.5 M sucrose as acidic buffer.
`
`Size-Exclusion Chromatography
`
`At time points 0.5, 2, 3.5, 5, 8, 12 and 24 h (Fig. 1), the acidic
`samples were measured by size-exclusion high performance
`liquid chromatography (SE-HPLC) on an Agilent 1200
`(Agilent Technologies) using a TSK gel G3000SWXL column
`(Tosoh Corporation). At time points 2, 5 and 24 h, the acidic
`samples were neutralized and then measured twice by SE-
`HPLC after incubation at room temperature for 10 and
`90 min, respectively. The mobile phase consisted of 50 mM
`Na-Citrate (pH 3.3) containing 100 mM NaCl for acidic sam-
`ples and standard PBS buffer (Invitrogen) for control and neu-
`tralized samples. The flow rate was 0.8 mL/min and the
`injected amount of antibody was 20 μg. The UV absorbance
`of eluted protein was detected at 280 nm. In addition, an IgG4
`sample at pH 3.3 after 2 h incubation was selected for multiple
`angle light scattering (MALS) measurement on miniDAWN
`TREOS detector (Wyatt Technology). The molecular weights
`(MW) of the eluted protein species were calculated with
`ASTRA 6 software (Wyatt Technology). Finally, the relative
`amounts of monomers, dimers and HMWS were
`
`calculated using UniChromTM software (New Analytical
`Systems Ltd.).
`
`Small-Angle X-ray Scattering (SAXS)
`
`The acidic samples were measured by SAXS at time points 0,
`2, 5, 12 and 24 h (Fig. 1). At each time point, the acidic
`samples were neutralized and measured. Just prior to mea-
`surement, the samples were diluted to obtain a total of three
`concentrations (approximately 1, 3, and 9 mg/mL). SAXS
`data was collected on the European Molecular Biology Lab-
`oratory Beamline P12 at the DORIS III storage ring (DESY,
`Hamburg, Germany). An EMBL/ESRF new generation
`sample changer (14)
`loaded the samples into a 10°C
`capillary flow cell with an exposure time of 0.045 s
`for each frame. 20 frames were collected for each mea-
`surement. The scattering intensity was recorded by a D
`photon counting Pilatus 2 M pixel detector (Dectris) in
`the momentum transfer range of 0.05–3.5 nm−1. The
`momentum transfer range is defined as q=4πsin(θ)/λ,
`where 2θ is the scattering angle and λ is the wavelength
`of the X-ray (λ =0.124 nm). The collected frames were
`checked for radiation damage before averaging and
`buffer subtraction. The ATSAS 2.5.2 software package
`was utilized for data analysis (15). SAXS curves from
`the concentration series were investigated individually
`for indications of particle attraction or repulsion. For
`the characterization of the oligomerization state, data
`from the concentration series were merged in the re-
`gions where the scattering patterns were identical
`to
`eliminate the impact of structure factors. The pair dis-
`tance distribution functions, P(r), were generated from
`the indirect Fourier transformation using GNOM and
`the radius of gyration (Rg) was further derived from
`the P(r) (16).
`
`Fig. 1 Schematic overview of the
`samples measured by SEC, SAXS
`and DLS. The horizontal axis
`indicates the duration of acidic
`incubation (pH 3.3). The vertical
`axis indicates the time after
`neutralization (pH 6.5). The
`symbols represent the sampling
`points for each method. At each
`point of SAXS analysis, two
`formulations and three dilutions for
`each IgG were measured.
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00003
`
`

`

`Subclass Specific Acid-Induced Antibody Aggregation
`
`Dynamic Light Scattering (DLS)
`
`All samples were studied in duplicates in a 384-well microplate
`(UV-Star® Microplate, Greiner Bio-One), which was centri-
`fuged for 10 min at 3000 rpm prior to measuring. Data collection
`consisted of 10 acquisitions of 5 s for each well with auto-
`attenuation of laser power using a DynaPro plate reader (Wyatt
`Technology) equipped with an 831 nm laser. The data was
`processed using the DYNAMICS 7 software (Wyatt Technolo-
`gy). The hydrodynamic radii (Rh) were obtained by analyzing the
`autocorrelation functions applying the measured refractive indi-
`ces and viscosities of the corresponding buffers. The acidic sam-
`ples were measured continuously over 20 h. The neutralized
`samples were prepared at time points 2, 5 and 16 h and mea-
`sured in parallel with the acidic samples (Fig. 1).
`
`Calculations of the Electrostatic Potential and Surface
`Hydrophobicity of Fc Fragments
`
`The structural data of Fc fragments for human IgG1 (1HZH),
`IgG2 (4HAG) and IgG4 (4C54) were downloaded from the PDB
`database (the respective PDB entries are shown in parenthesis),
`and the Fab fragments and hinge region of the IgG1 crystal
`structure (1HZH) were removed. 1HZH-Fc and the Fc part of
`the IgG1 molecule in the current study have identical primary
`sequences, which is also the case for 4C54 and the IgG4 mole-
`cule. 4HAG has a M282V mutation in the CH2 domain in
`comparison with the IgG2 molecule investigated in this study.
`The total solvent accessible surface area (SASA) and total hydro-
`phobic surface area on both glycosylated and aglycosylated Fc
`fragments were calculated by MOE software (17) at pH 3.3 and
`6.5. The net charge of these structures in the folded and unfolded
`states was computed as a function of pH using the PDB2PQR
`web server (18). Finally, we calculated the spatial aggregation
`propensity (SAP) to predict the aggregation prone surface regions
`on the Fc fragments of all three IgG subclasses (13). The SAP
`values were calculated in MATLAB R2014a by applying the
`definition and tool developed (13) and a radius of 10 Å. Visual-
`izations of the SAP calculations were performed in PyMOL by
`mapping the SAP values onto the Fc crystal structures. For quan-
`tification of local SAP, the sum of the SAP values was calculated
`for a spherical sample volume with radius of 10 Å around a
`selected center residue in each hydrophobic patch (19).
`
`RESULTS
`
`SE-HPLC Analysis Reveals Fundamentally Different
`Behavior of Model IgG1, IgG2 and IgG4 Molecules
`at Low pH and Upon Neutralization
`
`According to the SE-HPLC profiles (Fig. 2), the three model
`mAbs for IgG subclasses exhibited significantly different
`
`719
`
`stability over time at low pH. The non-stressed control sam-
`ples of all three IgGs were of very high quality with more than
`99% monomers in solution.
`The acidic samples of IgG1 remained homogeneous even
`after 24 h at pH 3.3, thus revealing extraordinary stability and
`impressive resistance towards low-pH treatment. However,
`subtle band broadening was observed in the former part of
`the monomer peak (Fig. 2a), implying that a small amount of
`IgG1 with larger hydrodynamic radius may exist at low pH.
`In contrast, after acidification, both IgG2 and IgG4 formed
`substantially larger species, which eluted before the monomer
`peak. In order to rule out the possibility that these earlier
`eluting peaks originated from unfolded and extended mono-
`mers, we characterized the MWs of the protein species from
`the three major peaks on the IgG4 SE-HPLC profile by SEC-
`MALS (Fig. 2a and Supplementary Fig. S1). The MWs of the
`peaks were determined to be 153, 302, 487 kDa, correspond-
`ing to the MW of monomer, dimer, and trimer species, re-
`spectively. Based on these findings, we can conclude that the
`development in earlier eluting species is indeed ascribed to the
`formation of aggregates. The slight increase of the estimated
`trimer MW may be due to the overlap of the trimer peak with
`earlier eluting peaks, which contain even larger protein spe-
`cies. We further calculated the relative percentage of peak
`area for monomer, dimer and HMWS (including all species
`larger than dimer). A table with the determined AUP and
`AUC can be found in Supplementary Table SI.
`Accordingly, both IgG2 and IgG4 exhibited a decrease of
`monomers in solution over time as shown in Fig. 3a. Interest-
`ingly, we observed that the aggregation kinetics differed sig-
`nificantly between IgG2 and IgG4. In the beginning of acidic
`treatment, the HMWS of IgG4 formed more rapidly than for
`IgG2. After 30 min, there were approximately 35% dimers
`and 15% HMWS in the IgG4 sample, while only 20% dimers
`and 3% HMWS in the IgG2 sample. The HMWS content
`further increased until reaching a relatively steady phase of
`aggregate formation after 8 h. Interestingly, after 24 h at pH
`3.3 IgG4 reached a lower percentage of oligomeric states
`where IgG2 had formed substantial amounts of larger oligo-
`mers and probably soluble aggregates (Figs. 2a and 3a). In
`addition, we observed that the amount of dimers decreased
`slowly and linearly for both IgG2 and IgG4 after 4 h (Fig. 3a),
`thus the HMWS might form via dimer-monomer or dimer-
`dimer oligomerization.
`The reversibility of acidified antibodies was investigated
`after 2, 5 and 24 h of acidification. The neutralized samples
`were analyzed after 10 and 90 min, respectively. As shown in
`Fig. 2b, IgG1 exhibited identical SE-HPLC profiles as the
`non-stressed control sample, indicating that the non-native
`molecules formed at low pH could be reverted to the native
`state even after 24 h. In contrast, IgG4 did not seem to be
`recovered after neutralization, despite that the initial percent-
`age of higher oligomeric states at pH 3.3 was lower than IgG2.
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00004
`
`

`

`720
`
`Fig. 2 SE-HPLC results for three
`IgG subclasses after acidification (a)
`and neutralization (b). The SE-
`HPLC profiles were normalized
`according to the total areas under
`the curves. The control samples
`were non-stressed antibodies at
`pH 6.5. The acidic samples were
`neutralized after incubation for 2, 5
`and 24 h. Then, the neutralized
`samples were further measured
`after 10 min and 1.5 h, respectively.
`
`Skamris et al.
`
`Instead, larger oligomers were formed upon neutralization at
`the expense of dimers in solution, while the amount of mono-
`mers did not significantly change at pH 6.5 (Fig. 3b). We
`observed a significant increase of IgG2 monomers in solution
`after neutralization, accompanied by a decrease of both di-
`mers and HMWS. At 2 h after acidic treatment, most IgG2
`was recovered through neutralization and further monomers
`were recovered over time throughout the time frame investi-
`gated (Fig. 3b).
`Altogether, it is evident from these analyses that IgG1 ex-
`hibited low oligomerization propensity at low pH, while the
`readily acid-induced IgG2 oligomerization was partially re-
`versed through neutralization. IgG4 oligomerized less vividly
`than IgG2, but in an irreversible fashion.
`
`SAXS Reveals Subtle Structural Changes
`and pH-Dependent Repulsive and Attractive
`Intermolecular Effects
`
`We applied SAXS to further analyze the acidic and neutral-
`ized samples immediately after acidification or neutralization
`and at several subsequent time points. We further monitored
`the effect of dilution in order to elaborate on the nature of
`interactive forces between the oligomers. Also, the effect of
`sucrose addition was investigated.
`As shown in Fig. 4, the scattering intensities of SAXS
`curves at low scattering angles (the low q region) for both
`IgG2 and IgG4 significantly increased over time at acidic
`pH, indicating the formation of larger particles. These sam-
`ples were measured directly without removal of potential in-
`soluble aggregates by centrifugation. The Guinier-regions of
`the SAXS curves were also resolvable (Supplementary
`Fig. S2), meaning that we can effectively monitor the largest
`distances present in the scatterers in solution, within the
`
`scattering angles monitored. Hence, we concluded that the
`samples consisted of well-defined oligomers rather than inho-
`mogeneous large aggregates (7). The average radius of gyra-
`tion (Rg) for the samples was further determined directly from
`the SAXS curves. The determined Rg values of IgG2 and
`IgG4 at acidic pH (Fig. 5) showed similar increasing trends
`over time as the amount of HMWS observed by SE-HPLC
`(Fig. 3). When the samples were incubated in sucrose, smaller
`Rg values of IgG2 and IgG4 were observed (Fig. 5). Particu-
`larly, the SAXS curves at the initial time point as well as 2 h
`after reducing the pH were identical to the scattering curves of
`control samples (Fig. 4). This demonstrated the favorable ef-
`fect of sucrose, along with a lower ionic strength, on antibody
`stability as we have reported previously (7). IgG1 samples at
`pH 3.3 with NaCl only exhibited minimal increase of scatter-
`ing intensity in the very low q region as compared to the non-
`stressed control sample (Fig. 4). No substantial changes in the
`Rg values of IgG1 were observed over time (Fig. 5).
`Upon neutralization, the solution behaviors of the three
`IgGs were in agreement with the SE-HPLC results (Fig. 2).
`Firstly, the SAXS curves of neutralized IgG1 were identical to
`the control sample, indicating full recovery (Fig. 4). We note
`that the Rg values of IgG1 samples at pH 3.3 were slightly
`higher than the neutralized samples. We can estimate the
`average MW of the antibodies in solution from the extrapo-
`lated forward scattering intensity at zero angle (I0), from which
`we can conclude that this slight increase in Rg originates from
`the presence of a minute amount of species with a larger mass
`(Supplementary Fig. S3), and hence not merely an increase in
`dimension associated with a more extended monomer confor-
`mation. Although the SE-HPLC analyses revealed that IgG1
`with NaCl did not tend to form larger species at pH 3.3 before
`24 h (Fig. 2a), the broadened monomer peak on the SE-
`HPLC profiles along with the increased scattering intensities
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00005
`
`

`

`Subclass Specific Acid-Induced Antibody Aggregation
`
`721
`
`Fig. 3 The percentage of monomer, dimer and higher molecular weight species (HMWS) peak areas integrated from SE-HPLC profiles of acidic samples (a) and
`neutralized samples (b). In a, filled symbols/full lines represent monomers, open symbols/dotted lines are dimers and filled symbols/stapled lines are higher
`molecular weight species. The x-axis represents time of acidic incubation and the symbols at zero hour of acidic treatment represent the results of control samples.
`In b, filled symbols/full lines represent samples that are acid incubated for 2 h, open symbols/dotted lines are samples acid incubated for 5 h and filled symbols/
`stapled lines are samples acid incubated for 24 h prior to neutralization. Here, the x-axis represents time after neutralization.
`
`in the very low q region of SAXS curves and the slightly
`increased Rg values indicate that IgG1 may transiently form
`small amounts of weakly interacting dimers, when subjected to
`low pH treatment. However, in the time frame investigated,
`this transient dimerization did not lead to the formation of
`detectable amounts of stable dimers or higher oligomers. Sec-
`ondly, IgG2 could be partially recovered, as the Rg values of
`the immediately neutralized samples slightly decreased com-
`pared to the corresponding acidic samples (Fig. 5). Thirdly,
`the Rg values of neutralized IgG4 significantly increased after
`neutralization. With SAXS, it was possible to obtain measure-
`ments immediately upon acidification and neutralization,
`
`which could not be obtained using SE-HPLC. The neutral-
`ized sample at such initial time point for IgG2 exhibited a
`SAXS curve identical to that of the control sample (Fig. 4),
`thus the monomers were reverted to native state. Interestingly,
`the neutralized IgG4 samples at initial time points also exhib-
`ited decreased Rg values and decreased scattering intensities at
`the lower scattering angles (Figs. 4 and 5). This indicates that
`IgG4 also could be recovered to the monomeric state if the
`IgG4 samples were neutralized right after low pH treatment.
`However, the irreversibility of IgG4 oligomerization at later
`time points implied that IgG4 went through further structural
`changes in the process of oligomerization at pH 3.3.
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00006
`
`

`

`722
`
`Skamris et al.
`
`We previously reported that the intermolecular repulsive
`interactions played an important role on the stability of anti-
`bodies at acidic pH (5,7). In the present study, all three model
`mAbs also exhibited repulsive interactions at pH 3.3, which
`was evident from the slightly decreased Rg values with increas-
`ing protein concentrations at each time point (Fig. 6). This
`effect was most pronounced for the sucrose-incubated samples
`where the magnitude of repulsion correlated to the pI ranking
`of the intact IgG molecules (9.0 / 8.2 / 7.9 for IgG1 / IgG2 /
`IgG4, respectively), which we have earlier characterized using
`imaged capillary isoelectric focusing (7). The repulsive forces
`between individual IgG molecules however only slowed the
`oligomerization kinetics, as it is evident that both IgG2 and
`IgG4 continued to oligomerize over time. It is noteworthy,
`that at the initial time points of acidification in NaCl, a slight
`attraction was observed both for IgG2 and IgG4, which was
`no longer evident after longer incubation. This could imply
`that small structural changes occurred after acidic incubation
`(at least for IgG2 and IgG4) and the initial greater propensity
`of attraction may play a role for the observed greater tendency
`for IgG2 and IgG4 to commence oligomerization. This sug-
`gestion is also supported by the observation that the attractive
`forces were diminished in the sucrose buffer at pH 3.3, which
`may lead to reduced IgG2 and IgG4 oligomerization.
`More distinct differences were observed for the different
`model mAbs at neutralizing conditions. At pH 6.5, attractive
`forces were evident for both IgG2 and IgG4 with NaCl (Rg
`values increased with protein concentration in Fig. 6), whereas
`neither repulsive nor attractive forces were observed for IgG1
`solutions (Rg values were independent on protein concentra-
`tion). Again, the attractive forces observed for IgG2 and IgG4
`were more pronounced in the absence of sucrose. When in-
`cubated in sucrose buffer, the neutralized IgG1 samples even
`exhibited weak repulsive forces. Together, these subtle differ-
`ences, observed via the systematic dilution of SAXS samples,
`may significantly influence both the onset of oligomerization
`and the recovery of monomers from the acid-induced oligo-
`meric states. In addition, the presence of sucrose did not alter
`the solution behaviors of IgG2 and IgG4 with respect to re-
`versibility, i.e. IgG2 oligomerization was partially reversible
`
`Fig. 4 Comparison of the SAXS data collected at different time points. The
`SAXS curves of acidic and neutralized samples with different excipients are
`translated for comparison. To emphasize the induced changes, only the low
`angle scattering is included (See Supplementary Fig. S2 for the full-range scat-
`tering curves).
`
`Fig. 5 Rg of three IgG subclasses calculated from the SAXS curves in Fig. 4.
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00007
`
`

`

`Subclass Specific Acid-Induced Antibody Aggregation
`
`723
`
`Fig. 6 The Rg values of mAbs at different protein concentrations. Prior to SAXS measurements, the acidic samples were diluted to obtain a total of three samples,
`9, 3 and 1 mg/mL, respectively. The neutralized samples were diluted to 8.0, 2.7 and 0.9 mg/mL, respectively. The results at different protein concentrations are
`shown in different grey scales, where the white columns indicate low protein concentration and the black columns indicate high protein concentration.
`
`while IgG4 oligomerization was irreversible and continued
`after neutralization (Figs. 4 and 5), exactly as observed in the
`absence of sucrose.
`Importantly, as the pI of a protein has a major influence on
`the level of intermolecular interactions, and acknowledging
`that the variable regions can contribute much to the pI, the
`trends reported here cannot be translated to all IgG mole-
`cules. However, the presented results reveal that there are
`indeed major differences governed by the properties specific
`to each subclass.
`
`DLS Analysis of the Time Course of Monomer Recovery
`After Neutralization
`
`DLS was employed as an orthogonal method to further investi-
`gate the subclass-specific stability and aggregation reversibility.
`Taking advantage of the automated instrumental setup, data were
`collected with short time intervals, which provided a more elabo-
`rate depiction of the solution behaviors of antibodies over time.
`This enabled a study of the effect of neutralization over time.
`In a previous study from identically prepared native IgG sam-
`ples at pH 6.5, we showed that IgG1, 2 and 4 had almost iden-
`tical Rh values (7). Here, at acidic pH, the initial Rh values for
`each IgG molecule (Fig. 7) ranged from 6.0 to 7.4 nm. The first
`developments in Rh values simply occurred before the first mea-
`surement could be recorded. In general, the Rh values at acidic
`pH were in agreement with the Rg values based on SAXS data
`
`at the corresponding time points and if scaled, the values overlaid
`perfectly (Supplementary Fig. S4). The DLS data further con-
`firmed partial recovery of monomers in neutralized IgG2 after 2
`and 5 h in acidic conditions where the Rh values decrease imme-
`diately upon pH shift (Fig. 7). However, the Rh values of the
`neutralized IgG2 samples were still much higher than those of
`IgG1, indicating the presence of oligomers. In addition, we did
`not observe full recovery of IgG2 even after 18 h incubation at
`neutral pH (Fig. 7) hence a part of the IgG2 oligomers formed
`were irreversible. Interestingly, no recovery was observed upon
`neutralization of IgG2 samples incubated for 16 h in acidic con-
`ditions. The Rh values of IgG4 increased immediately upon neu-
`tralization, exactly as the SAXS derived Rg values. The observed
`discrepancy between the relative magnitude of Rh and Rg for
`freshly neutralized samples is most likely attributed to the fact
`that neutralized samples for DLS were measured 10 min after
`neutralization due to the necessity of centrifugation. An interest-
`ing pattern appears when comparing the Rh time-course for
`neutralized IgG4 after different acidic incubation times. In all
`three cases, oligomerization continued for approximately 2 h
`(Fig. 7). However, after this immediate aggregation state, the
`average sizes of the aggregates decreased as observed for samples
`incubated both for 2 and 5 h. The immediate aggregation phase
`observed both by SAXS, SE-HPLC and DLS was thus followed
`by a slower depolymerization, which could imply that IgG4 en-
`countered structural rearrangements over time after neutraliza-
`tion. IgG2 exhibited the inverse behavior depending on the
`
`Chugai Exhibit 2012
`Pfizer, Inc. v. Chugai Pharmaceutical Co. Ltd.
`IPR2017-01357
`Page 00008
`
`

`

`724
`
`Skamris et al.
`
`Fig. 7 Rh of three IgG subclasses measured by DLS after acidification and neutralization. The filled circle symbols represent the samples at pH 3.3 with NaCl