You are on page 1of 14

Journal of Pharmaceutical Sciences 109 (2020) 656-669

Contents lists available at ScienceDirect

Journal of Pharmaceutical Sciences


journal homepage: www.jpharmsci.org

Pharmaceutical Biotechnology

Unique Impacts of Methionine Oxidation, Tryptophan Oxidation,


and Asparagine Deamidation on Antibody Stability and
Aggregation
Magfur E. Alam 1, Thomas R. Slaney 2, Lina Wu 3, Tapan K. Das 2, Sambit Kar 2,
Gregory V. Barnett 2, Anthony Leone 2, Peter M. Tessier 1, 3, 4, 5, *
1
Isermann Department of Chemical & Biological Engineering, Center for Biotechnology & Interdisciplinary Studies, Rensselaer Polytechnic Institute,
Troy, New York 12180
2
Biologics Development, Bristol-Myers Squibb, Pennington, New Jersey 08534
3
Department of Chemical Engineering, Biointerfaces Institute, University of Michigan, Ann Arbor, Michigan 48109
4
Department of Pharmaceutical Sciences, Biointerfaces Institute, University of Michigan, Ann Arbor, Michigan 48109
5
Department of Biomedical Engineering, Biointerfaces Institute, University of Michigan, Ann Arbor, Michigan 48109

a r t i c l e i n f o a b s t r a c t

Article history: Monoclonal antibodies are attractive therapeutic agents because of their impressive biological activities
Received 16 August 2019 and favorable biophysical properties. Nevertheless, antibodies are susceptible to various types of
Revised 22 October 2019 chemical modifications, and the impact of such modifications on antibody physical stability and ag-
Accepted 28 October 2019
gregation remains understudied. Here, we report a systematic analysis of the impact of methionine
Available online 31 October 2019
oxidation, tryptophan oxidation, and asparagine deamidation on antibody conformational and colloidal
stability, hydrophobicity, solubility, and aggregation. Interestingly, we find little correlation between the
Keywords:
mAb impact of these chemical modifications on antibody conformational stability and aggregation. Methio-
solubility nine oxidation leads to significant reductions in antibody conformational stability while having little
formulation impact on antibody aggregation except at extreme conditions (low pH and elevated temperature).
developability
Conversely, tryptophan oxidation and asparagine deamidation have little impact on antibody confor-
protein aggregation
mational stability while promoting aggregation at a wide range of solution conditions, and the aggre-
gation mechanisms appear linked to unique types of reducible and nonreducible covalent crosslinks and,
in some cases, to increased levels of attractive colloidal interactions. These findings highlight that even
related types of chemical modifications can lead to dissimilar antibody aggregation mechanisms, and
evaluating these findings for additional antibodies will be important for improving the systematic
generation of antibodies with high chemical and physical stability.
© 2020 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.

Introduction their long circulation times and favorable biophysical properties.4-6


The growth in antibody therapeutics has also been aided by robust
There is intense interest in developing monoclonal antibodies as methods for generating them both in vivo (e.g., immunization) and
therapeutics, which is evidenced by the growing number of in vitro (e.g., phage and yeast surface display) against seemingly any
approved antibody drugs and the large number of clinical candi- antigen of interest.7,8 These and other advances have led to anti-
dates.1-3 The power of antibody therapeutics stems from their many bodies being considered one of the standard modalities used to
attractive properties, including their natural ability to link antigen treat diverse human disorders.
binding to recruitment of immune effector functions as well as Despite the many advantages of antibody therapeutics, they are
susceptible to similar chemical and physical instabilities observed
for other types of biologics. For example, the Fc region of most
Conflicts of interest: None. antibodies contains conserved methionine (Met) residues (e.g.,
This article contains supplementary material available from the authors by request
Met252 and Met428) located at the interface of the CH2 and CH3
or via the Internet at https://doi.org/10.1016/j.xphs.2019.10.051.
* Correspondence to: Peter M. Tessier (Telephone: 734.763.1486). domains, which are particularly sensitive to oxidation.9,10 Oxidation
E-mail address: ptessier@umich.edu (P.M. Tessier). of these Met residues typically results in reduced antibody

https://doi.org/10.1016/j.xphs.2019.10.051
0022-3549/© 2020 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 657

conformational stability.11-14 In some cases, this also causes untreated (control) mAb solutions were buffer-exchanged into PBS
decreased antibody interaction with Fc receptors such as the at pH 7.4 using Zeba Spin Desalting Columns (PI89894; Thermo
neonatal Fc receptor (FcRn),14-16 which can also lead to reduced Fisher Scientific), aliquoted (~0.2-0.5 mL) and frozen at 80 C until
antibody half-life in vivo.16 Tryptophan (Trp) is also particularly evaluation.
susceptible to oxidation, and oxidation of Trp residues in antibody To achieve selective Met oxidation, H2O2 (3%, w/w) was added to
complementarity-determining regions (CDRs) has been linked to the stock mAb solutions to achieve a protein:H2O2 molar ratio of
loss of antigen binding.17-19 Moreover, Trp oxidation has been 1:100. The solutions were mixed gently and incubated at 25 C for 1
linked to reduced antibody physical stability by promoting disul- day. Afterward, a 3-fold molar excess of a Met solution (200 mM)
fide scrambling20,21 and aggregation.17,22,23 Asparagine (Asn) dea- was added to quench the oxidation reaction. Next, the mAb solu-
midation is another common type of antibody chemical tions were buffer-exchanged into PBS at pH 7.4 using Zeba Spin
modification, which has been linked to reduced affinity due to Desalting Columns, aliquoted and frozen at 80 C until evaluation.
deamidation of CDR residues24-26 as well as variable impacts on To achieve selective Trp oxidation, 2,2’-Azobis (2-
antibody conformational stability24,27,28 and aggregation.22,27 methylpropionamidine) dihydrochloride (AAPH; 2 mg/mL) was
Nevertheless, there are a number of open questions related to added to the stock mAb solutions to achieve a protein:AAPH molar
how Met oxidation, Trp oxidation, and Asn deamidation impact ratio of 1:100. Free DL-Met (10 mg/mL) was also added to the so-
antibody aggregation that are the focus of this study. First, which lution to protect the Met residues in the mAbs from oxidation. The
type of chemical modification possesses the greatest risk for pro- solutions were then mixed gently and incubated at 37 C for 3 days.
moting antibody aggregation when compared head-to-head for a At the end of the incubation period, the mAb solutions were
common set of antibodies? Second, how dependent are the levels of enriched for monomer using a semipreparative SEC column
antibody aggregation induced by each type of chemical modification €
running on an AKTA Avant 25. The enriched monomer samples
on the conditions (e.g., pH and temperature) used for evaluating were buffer-exchanged into PBS at pH 7.4 using Zeba Spin Desalting
antibody physical stability? Third, how similar or different are the Columns, aliquoted and frozen at 80 C until evaluation.
aggregation mechanisms for antibodies subjected to chemical To achieve Asn deamidation, the stock mAb solutions were
modifications of the same class (oxidation of Met vs. Trp) and titrated to pH 9.5 with Tris base. Next, the solutions were mixed
different classes (oxidation vs. deamidation)? Here we systemati- gently and incubated in the absence of light at 37 C for 1 week.
cally evaluate the impact of deamidation and selective oxidation of Afterward, the mAb solutions were buffer-exchanged into PBS at
Met and Trp residues on antibody aggregation for a wide range of pH 7.4 using Zeba Spin Desalting Columns, aliquoted and frozen
pHs (pH 3.8-7.4) and temperatures (4 C-37 C), and report unique at 80 C until evaluation.
aggregation mechanisms for each type of chemical modification. For the accelerated stability study, the mAb samples were pre-
pared at 1 mg/mL in citrate-phosphate buffer (15 mM citrate, 10
Experimental Methods mM phosphate; pH 3.8-7.4) with 0.04% w/v sodium azide. The mAb
samples were stored in 1.5 mL Eppendorf tubes (16155500, USA
Materials Scientific) and maintained at 37 C in an incubator or at ~25 C on
the bench.
The mAbs used in this study (stock concentrations of 10-15 mg/
mL) were an IgG1 (mAb 1) and an IgG4 (mAb 2), which were pro- Imaged Capillary Isoelectric Focusing
vided by Bristol-Myers Squibb (Hopewell, NJ) as formulated drug Capillary isoelectric focusing experiments were performed as
substances. Syringe filters were obtained from EMD Millipore described previously.27 Briefly, antibodies (theoretical pIs for mAb 1
(SLGV004SL), VWR (28145-501) and Thermo Fisher Scientific of 8.6-8.9 and mAb 2 of 8.1-8.4) were diluted in water to 1.5-2.5 mg/
(097203). Polystyrene plates (clear 384-well, 12565506; clear 96- mL, 4% Pharmalyte (GE Healthcare), 1% methyl cellulose (Pro-
well, 07200656; black 96-well, 07200589) were obtained from teinSimple), 8 M urea (Sigma Aldrich), and appropriate pI markers
Thermo Fisher Scientific and adhesive films (2920-0000) were (ProteinSimple). Next, the diluted samples were injected into an
obtained from USA Scientific. PCR plates (white; 04729692001) and iCE3 (ProteinSimple) system equipped with a fluorocarbon-coated
sealing foils (04729757001) were obtained from Roche Diagnostics. capillary cartridge. Each sample was prefocused for 1 min at 1500 V
Protein Thermal Shift Dye (4461146) was obtained from Applied and then focused for 10 min at 3000 V. The focusing patterns were
Sciences. Zeba Spin Desalting Columns (PI89894) and clear poly- then captured by a CCD camera at 280 nm.
styrene inserts (13622207) were obtained from Thermo Fisher
Scientific. Individually packaged cuvettes (95201005) were ob- Size-Exclusion Chromatography
tained from VWR. Potassium phosphate (205935000) was obtained For the concentration-dependent aggregation study, the mAb
from Acros Organics. Citric acid (251275), Pharmalyte (17-0456-01), samples were analyzed via size-exclusion chromatography using a
urea (57-13-6), and ammonium sulfate (7783-20-2) were obtained TSKgel SuperSW mAb HTP column (0.46  15 cm, Tosoh Bioscience)
from Sigma-Aldrich. A methyl cellulose kit (1%; 101876) was ob- and a Waters 6000 HPLC. The column was washed with 4 column
tained from ProteinSimple. Amicon 30 kDa centrifugal filters volumes of buffer (either PBS supplemented with 200 mM arginine
(UFC503096) were obtained from EMD Millipore. The 2 Laemmli at pH 7.4 or citrate-phosphate buffer at pH 3.8). Next, the samples
buffer (1610737) and 4%-15% mini-PROTEAN TGX gels (4561086) (10 mL of mAb samples at 1 or 25 mg/mL) were injected into the
were obtained from Biorad. b-Mercaptoethanol (200000622) was column without extended storage (injected <2 h after preparation)
obtained from VWR. Gelcode Blue Safe Protein Stain (1860983) and in an order that was randomized between repeats, and the area
Pierce Silver Stain Kit (24612) were obtained from Thermo Fisher under the curve was evaluated automatically using the HPLC soft-
Scientific. ware (Waters Empower Pro 2). All peaks before the main protein
peak were considered to be high-molecular-weight (HMW) species.
Methods The %HMW species was calculated as the area under the curve of
the HMW species divided by the area under the curve of the entire
Antibody Sample Preparation chromatogram.
The stock mAbs (~10 mg/mL) were provided by Bristol-Myers For the accelerated stability samples, aliquots (50 mL) were taken
Squibb at pH ~6 in their respective formulation buffers. The at 0, 1, 7, and 28 days, and 10 mL of each mAb solution was injected
658 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

into the TSKgel SuperSW mAb HTP column equilibrated in PBS temperature (Tonset) was determined by using methods described
buffer supplemented with 200 mM arginine at pH 7.4. The %HMW previously.29
species for each sample was calculated in the same way as
described previously. The total area under the curve for the entire Dynamic Light Scattering
chromatogram was also evaluated at each time point to determine Light scattering experiments were performed as described
the percentage antibody recovery. previously.27 Briefly, each mAb sample was buffer-exchanged 4
times into the target buffer (15 mM citrate, 10 mM phosphate at pH
SDS-PAGE 3.8 or pH 7.4). Next, the mAbs were concentrated using 30 kDa
The mAb samples were diluted to 2 mg/mL (1 mg/mL for the Amicon centrifuge filters (UFC503096, EMD Millipore) to ~30 mg/
accelerated stability samples and 40 mg/mL for the deamidated mL and filtered using 0.22 mm syringe filters (09-720-3, Thermo
[monomer and HMW] samples purified by size-exclusion chro- Fisher Scientific). Next, their concentrations were measured via UV-
matography) in citrate-phosphate buffer at pH 7.4. Next, equal absorbance at 280 nm using extinction coefficients of 1.53 and 1.59
volumes of the mAb solutions and 2 Laemmli buffer, prepared mL/(mg$cm) for mAb 1 and mAb 2, respectively, and then diluted to
with and without reducing agent (b-mercaptoethanol), were mixed a concentration of 25 mg/mL. The concentrated mAbs (80 mL) were
together. The mixtures without reducing agent were evaluated transferred to disposable cuvettes (E0030106300, Fisher Scientific)
either as is (control) or after heating at 95 C for 5 min. The mixtures and placed in a Wyatt DynaPro dynamic light scattering instru-
with reducing agent were heated for 5 min at 95 C. Next, the ment. The size distribution was measured using a regularization
samples were centrifuged at 16,000 rcf for 2 min, and 10 mL (15 mL algorithm (Wyatt instrument software) and calculated as the
for the accelerated stability samples and 5 mL for the purified average of 20 measurements. Each measurement was conducted
deamidated [monomer and HMW] samples) was loaded into each for 10 s at a laser power that yielded approximately 1 million
well of 4%-15% mini-PROTEAN TGX (4561086, Biorad) gels. Tris- counts per second.
glycine running buffer (25 mM Tris, 192 mM glycine and 0.1% SDS The experiments to determine the apparent diffusion interac-
at pH 8.3) was used and the samples were run for 1 h at 120 V. tion parameter (kD*) were performed as described previously.27
Finally, the gels were stained using either Gelcode Blue Safe Protein Briefly, the mAb samples were buffer-exchanged 4 times into
Stain or Pierce Silver Stain (1860983 or 24612, Thermo Fisher Sci- buffered solutions (15 mM citrate, 10 mM phosphate) at pH 3.8 and
entific) and then destained and imaged. 7.4. Next, the mAbs were filtered using 0.22 mm PVDF syringe filters
(SLGV004SL, EMD Millipore), and their concentrations were
measured via UV-absorbance at 280 nm. The concentrated mAbs
Differential Scanning Fluorimetry (10 mg/mL) were transferred (80 mL) to disposable cuvettes and
The mAb samples were diluted to 0.115 mg/mL in their placed in a Wyatt DynaPro dynamic light scattering instrument.
respective buffered solutions (15 mM citrate, 10 mM phosphate at The diffusion coefficients were obtained as described previously.30
pH 3.8 to 7.4). Protein Thermal Shift Dye (4461146, Applied Bio- Briefly, the autocorrelation function was fit using the method of
sciences), initially provided at a concentration of 1000, was cumulants (Wyatt instrument software) to determine average
diluted to a concentration of 40 using Milli-Q water. The mAb diffusion coefficient (DM) values, which were calculated as the
solutions (17.5 mL in each well) were dispensed into 96-well white average of 20 measurements and each measurement was con-
PCR plates (04729692001, Roche) in triplicate. Next, dye solution ducted for 10 s. The mAbs were then diluted sequentially to lower
(2.5 mL) was added to each well, and the solutions were mixed concentrations (4-9 mg/mL) in their respective buffers and evalu-
thoroughly. The plates were then sealed with foil (04729757001, ated via light scattering. The diffusion coefficients (DM) for the mAb
Roche Diagnostics) and thermal melts were performed using a solutions (4-10 mg/mL) were fit to the following equation:
LightCycler 480 real-time PCR instrument (Roche Diagnostics). The
fluorescence (Ex: 558 nm, Em: 610 nm) was measured as the plate 
DM ¼ D0 1 þ k*D c
was heated from 25 C to 95 C. Fifty acquisitions were collected per
1 C, and the heating rate was ~0.7 C/min. The apparent melting where D0 is the self-diffusion coefficient, kD* is the apparent
temperatures (Tm*) were determined by analyzing the first deriv- diffusion interaction parameter, and c is the antibody
ative of the fluorescence with respect to temperature, which concentration.
involved fitting a second-order polynomial to the major peak and
determining the temperature at the maximum. In a few cases, the Hydrophobic Interaction Chromatography
peaks could not be fit using a second-order polynomial and instead The hydrophobic interaction chromatography experiments
were deconvoluted using a bi-Gaussian fit (Origin 2019). were performed using a ProPac HIC-10 column (063653, Thermo
Fisher Scientific) and a Waters 6000 HPLC. The mAb solutions were
Differential Scanning Calorimetry diluted to 2 mg/mL in citrate-phosphate buffer at different pH
The stock mAb samples were buffer-exchanged into their values (pH 3.8 or 7.4) and mixed with an equal volume of 1.2 or 2 M
respective buffered solutions (15 mM citrate, 10 mM phosphate at ammonium sulfate solution (prepared at the same buffer condi-
pH 3.8 and 7.4) and diluted to a concentration of 1 mg/mL. Next, the tions) to obtain final concentrations of 1 mg/mL mAb and 0.6 or 1 M
antibodies (~450 mL) were loaded into the sample cell and the ammonium sulfate. Next, the mAb samples (100 mL) were injected
corresponding buffers (~450 mL) were loaded into the reference cell into the column equilibrated in citrate-phosphate buffer at pH 3.8
of a Nano DSC calorimeter (TA Instruments). Next, the cells were or 7.4 with 0.6 or 1 M ammonium sulfate. After the injection, an
heated (0.7 C/min) from 25 C to 95 C for the samples prepared at isocratic step at 0.6 or 1 M ammonium sulfate was run for 4 min
pH 7.4 and from 15 C to 75 C for the samples prepared at pH 3.8. followed by an ammonium sulfate gradient. This consisted of a
The apparent melting temperatures (Tm*) were determined by 20 min negative linear gradient (0.6 or 1 M to 0 M) at a flow rate of 1
fitting a second-order polynomial to the first peak of the thermo- mL/min, which was used to elute the mAbs for evaluation of their
grams (after subtracting the blank buffer baseline) and determining retention behavior. The area under the curve was calculated be-
the temperature at the maximum. In some cases, the peaks could tween the column void volume and the end of the run, and the
not be fit using a second-order polynomial and instead were percentage recovery of the chemically modified antibodies was
deconvoluted using a Gaussian fit (Origin 2019). The onset evaluated relative to the control.
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 659

Ammonium Sulfate Precipitation (51%) regions, whereas mAb 2 displayed modest levels of Met
The mAb solutions were diluted to 2 mg/mL in citrate- oxidation in its Fab region (12%) and high levels in its Fc region
phosphate buffer at different pH values (pH 3.8-7.4). Stock solu- (51%). Conversely, mAbs treated with AAPH in the presence of
tions of citrate-phosphate buffer were also prepared at the same pH excess methionine displayed relatively high levels of Trp oxidation
values both with and without 3.5 M ammonium sulfate. Next, the (29%-87% in their Fab regions) and little Met oxidation (<2%). In
mAb solutions (10 mL) were added to each well of clear 96-well addition, Trp oxidation was not observed in the Fc region for either
plates (07200656, Fisher Scientific), and then ammonium sulfate mAb. Moreover, the samples incubated at high pH (pH 9.5) dis-
solutions (90 mL) were added at different concentrations (0.5-2 M). played relatively high levels of Asn deamidation in their Fc regions
Next, the solutions were mixed gently to prevent formation of air (18%-20%) and minimal levels of deamidation in the Fab regions
bubbles. The solutions were incubated at room temperature (<4%). As expected, most of the deamidation in the Fc regions
(10 min) and the absorbance (turbidity) values at 350 nm were occurred in the PENNY peptide (Asn384 and Asn389; 54.4% and
measured. The turbidity values were normalized using the 36.4% for mAb 1 and 2, respectively).31,32 While the 3 different
maximum and minimum turbidity values for a given set of samples conditions used to promote chemical degradation were relatively
(i.e., samples for the same mAb and pH). The normalized absor- selective, it is notable that treatment of mAb 1 at high pH to pro-
bance values of the mAbs (y-axis) were plotted against the mote deamidation also increased Met oxidation in the Fab region of
ammonium sulfate concentration (x-axis), and 4-parameter Boltz- mAb 1 (19% modified relative to 10% for the control), whereas this
mann sigmoidal curves (y ¼ minþ[max-min]/[1 þ exp[[CS,50 -x]/ was not observed for mAb 2.
slope]]) were fit to the resulting graphs. The midpoint of transition One of the principal goals of our study was to evaluate the ef-
(CS,50) was evaluated as a measure of the relative solubility of the fects of oxidation and deamidation on the stability of antibodies
mAbs. due to various stresses encountered by the mAbs during the
manufacturing process, which can include changes in pH, concen-
Intrinsic Tryptophan Fluorescence tration, and temperature.33,34 Therefore, we first sought to quantify
The mAb solutions were diluted to 0.25 mg/mL in citrate- the amount of soluble high molecular weight (HMW) species pre-
phosphate buffer (pH 3.8 or 7.4). Next, the mAb solutions (100 sent without extended storage as a function of pH and concentra-
mL) were dispensed into each well of black 96-well plates tion (Figs. 1 and S1). We evaluated the aggregation behavior of the
(07200589, Fisher Scientific) in triplicate. The intrinsic tryptophan control and chemically modified mAbs via size-exclusion chroma-
fluorescence emission spectra were measured from 315 to 450 nm tography at pH 3.8 (Figs. 1a and 1b) because low pH is frequently
(in 1 nm increments) after excitation at 295 nm using a Tecan used for antibody purification and is often associated with antibody
M1000 Pro plate reader. The fluorescence of the buffer solution was aggregation.35-37 At a moderately high mAb concentration (25 mg/
also measured and subtracted from the antibody fluorescence mL), the size-exclusion chromatograms revealed the presence of
spectra. HMW species for the control and chemically modified samples of
both mAbs (Fig. S1). The Trp-oxidized samples displayed the
Results highest levels of HMW species followed by the deamidated sam-
ples, and both the Met-oxidized and control samples displayed low
Oxidation and Deamidation Mediate Unique pH- and Temperature- levels of aggregation. However, at a lower mAb concentration (1
Dependent Antibody Aggregation Behaviors mg/mL), only the Trp-oxidized samples displayed significantly
higher levels of aggregation than the control (Figs. 1a and 1b).
Toward our goal of evaluating the effects of oxidation and dea- Moreover, deamidation promotes the formation of soluble aggre-
midation on antibody stability, we performed forced degradation of gates at low pH in the most pronounced concentration-dependent
2 mAbs [mAb 1 (IgG1) and mAb 2 (IgG4)] using conditions that are manner, as observed previously.27 We also found that the aggre-
relatively selective for modifying Met, Trp, or Asn residues (here- gation behavior of the same mAbs at neutral pH followed a similar
after referred to as Met-oxidized, Trp-oxidized and deamidated trend, although the levels of aggregation were higher at near-
mAbs, respectively), whereas the untreated mAbs were used as neutral pH than at low pH (Figs. 1c and 1d).
controls. We performed mass spectrometry (peptide mapping) to Next, we evaluated the aggregation behavior of the mAbs using
evaluate the extent of modification of the oxidation and deamida- a second method (dynamic light scattering) without extended
tion sites (Table 1). For the mAbs treated with H2O2, mass spec- storage (<2 h) to determine the generality of our findings (Figs. 2,
trometry analysis revealed large increases in the levels of Met S2 and S3). Light scattering revealed that the deamidated samples
oxidation relative to small increases in Trp oxidation. In particular, (25 mg/mL) displayed distinct, larger populations of particles for
mAb 1 showed high levels of Met oxidation in its Fab (91%) and Fc both mAbs at low pH (pH 3.8; Figs. 2a, 2b, S2 and S3). Moreover, the

Table 1
Summary of the Oxidation and Deamidation Levels of the Antibodies Evaluated in This Study

Methionine Residues (% Modified) Tryptophan Residues (% Modified) Asparagine Residues (% Modified)

Con. H2O2 AAPH þ Met High pH Con. H2O2 AAPH þ Met High pH Con. H2O2 AAPH þ Met High pH

mAb 1
Fab 10 91 8.7 19 <1 1.8 87 <1 <1 <1 <1 3.6
Fc 3.1 51 4.6 7.8 e e e e 2.1 2.0 2.5 18
mAb 2
Fab <1 12 1.6 <1 <1 <1 29 <1 <1 <1 <1 1.1
Fc <1 51 3.4 2.3 e e e e 2.8 2.5 2.0 20

Mass spectrometry analysis of the control (Con.), methionine (Met)-oxidized (H2O2-treated), tryptophan (Trp)-oxidized (AAPH þ Met-treated), and deamidated (high pH
treated) mAbs was performed to evaluate the level of chemical modification for each sample. The data are the average (%) levels of chemical modification observed for Met, Trp,
and Asn residues located in the Fab or Fc region. mAb 1 modifications were detected for 2 Met residues (including 1 in the CDRs), 1 Trp residue (in the CDRs), and 4 Asn residues
(including 2 in the CDRs) in the Fab region, as well as for 3 Met and 3 Asn residues in the Fc region. mAb 2 modifications were detected for 1 Met residue, 1 Trp residue (in the
CDRs), and 1 Asn residue (in the CDRs) in the Fab region, and for 3 Met and 4 Asn residues in the Fc region.
660 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

Figure 1. Size-exclusion chromatography analysis of the impact of oxidation and deamidation on antibody aggregation as a function of pH. (a-d) The antibodies were prepared at 1
and 25 mg/mL in citrate-phosphate (15 mM citrate, 10 mM phosphate) buffer at (a, b) pH 3.8 and (c, d) pH 7.4, and injected into the column without extended storage (injected <2 h
after preparation). The data shown are averages of 3 independent experiments and the error bars are standard errors. A 2-tailed Student's t-test was used to judge the statistical
significance of the difference between the chemically modified and control mAbs [p-values <0.05 (*), <0.01 (**), or <0.001 (***)].

Trp-oxidized samples displayed a distinct population of particles sample of mAb 1 also displayed reduced recovery (~75% recovery)
for mAb 1 and modestly increased polydispersity (~21%) for mAb 2 from the size-exclusion column relative to the control (~90% re-
compared to the control (~10% polydispersity; Figs. 2a, 2b, S2 and covery; Fig. S4). This aggregation behavior was not observed when
S3). By contrast, the Met-oxidized samples of both mAbs were the incubation temperature was lowered to 25 C (Figs. 3c and 3d),
similar to the controls. Increased polydispersity in dynamic light as Met-oxidized samples displayed low levels of aggregation at
scattering may indicate the presence of dimers or trimers while 25 C that were similar to the controls. At pH 4.5, we also observed
distinct populations of larger species often represent larger ag- that Met-oxidized samples were aggregation prone at 37 C but not
gregates.38 At near-neutral pH (pH 7.4), the deamidated samples at 25 C, although the levels of aggregation were much lower (<20%
also displayed a distinct, larger population of particles for mAb 2 after 28 days at 25 C; Fig. S6) than at pH 3.8. At higher pH values
(but not for mAb 1) relative to the control samples. By contrast, the (pH 6 and 7.4), all of the mAb samples displayed modest changes in
same pH condition resulted in high levels of polydispersity for the aggregation after incubation for 28 days at 37 C (Figs. 4a, 4b and S7)
Trp-oxidized samples of mAb 1 (~32%) relative to mAb 2 (~13.8%) and 25 C (Figs. 4c, 4d and S7). The highest levels of aggregation at
and the controls (~8% and ~9% for mAb 1 and 2, respectively; pH 6 and 7.4 corresponded to the Trp-oxidized samples followed by
Figs. 2c, 2d, S2 and S3). The Met-oxidized samples at pH 7.4 showed the deamidated samples, whereas the Met-oxidized and control
little change relative to the control samples for both mAbs. Taken samples displayed the lowest levels of aggregation. At these pH
collectively, the light scattering and size-exclusion chromatography values, the chemically modified antibodies were recovered from
results suggest that deamidation and Trp oxidationdwithout pro- the size-exclusion column at high levels (>95% recovery) that were
longed incubation (<2 h)denhance aggregation of both mAbs at a similar to those observed for the controls (Fig. S5). These findings
range of solution conditions. demonstrate that the Met-oxidized samples are particularly
We next sought to evaluate the effects of temperature and pH on aggregation-prone at elevated temperature (37 C) and low pH (pH
the aggregation behavior of the mAbs during extended storage. 3.8), but are resistant to aggregation at lower temperatures (25 C)
Thus, we incubated the control and modified mAbs at an elevated or higher pH values (pH 6 and 7.4).
temperature (37 C) and room temperature (25 C) as a function of
pH (pH 3.8-7.4) for 7-28 days (Figs. 3, 4, S4-S7). We first evaluated Unique Mechanisms Mediate Enhanced Antibody Aggregation Due
the aggregation behavior of the mAbs at low pH (pH 3.8) during to Oxidation and Deamidation
incubation for 7 days via size-exclusion chromatography (Fig. 3). As
expected, the mAbs displayed high levels of aggregation at 37 C There are multiple possible mechanisms by which oxidation and
and low pH during extended storage. Interestingly, Met-oxidized deamidation can potentially enhance aggregation. One such
samples displayed the largest increase in HMW species relative to mechanism is that these chemical modifications can lead to the
the control at 37 C after 7 days (Figs. 3a and 3b). The Met-oxidized formation of covalently cross-linked species between different
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 661

Figure 2. Dynamic light scattering analysis of the impact of oxidation and deamidation on antibody aggregation as a function of pH. (a-d) The antibodies were prepared at 25 mg/
mL in citrate-phosphate buffer at (a, b) pH 3.8 and (c, d) pH 7.4, and their size distributions were measured using dynamic light scattering. The data shown are representative
examples of 3 independent experiments.

antibody domains.20,39-41 Therefore, we used SDS-PAGE to evaluate samples as pH was reduced. Interestingly, the Met-oxidized sam-
the formation of reducible and nonreducible cross-linked species of ples displayed significant decreases in Tm* (2.5 C-6.5 C) relative to
the control, oxidized, and deamidated mAbs (Fig. 5). We observed the controls for a wide range of pH values (pH 3.8-7.4). For example,
that the Trp-oxidized samples of both mAbs displayed HMW spe- the apparent melting temperatures of the Met-oxidized samples
cies (>260 kDa) for nonreducing conditions and that the level of were significantly lower than the untreated controls at pH 3.8
aggregation was higher for mAb 1 than for mAb 2. Intrinsic Trp (41.7 C relative to 47.8 C for mAb 1 and 36.2 C relative to 42.3 C for
fluorescence (Fig. S8) and mass spectrometry (Table 1) confirmed mAb 2). However, the differences in apparent melting temperature
that the Trp-oxidized samples for mAb 1 were modified to a greater between Trp-oxidized, deamidated, and control (untreated) sam-
extent than for mAb 2, which potentially explains the higher ag- ples were much smaller (<1 C). We also evaluated the apparent
gregation levels for mAb 1. Moreover, the Trp-oxidized samples in melting temperatures (Tm*) and the onset temperature of unfolding
the presence of reducing agent did not display cross-linked species (Tonset) using DSC and observed similar trends (Figs. S12 and S13).
for either mAb (Fig. 5). Interestingly, the deamidated samples, These results suggest that oxidation and deamidation of the mAbs
especially for mAb 1, displayed nonreducible cross-linked species in this study result in modest decreases in antibody folding stability
(~80-90 kDa). Moreover, similar behavior was observed after and, therefore, are unlikely to promote aggregation via mechanisms
incubating the antibodies at pH 7.4 and an elevated temperature that involve significant antibody unfolding. The notable exception
(37 C) for 1 week (Fig. S9). In addition, we find that deamidated is for Met-oxidized samples at low pH (pH 3.8) and elevated tem-
mAb 1 monomers and HMW speciesdwhich were isolated using perature (37 C), a condition at which low antibody folding stability
size-exclusion chromatographyddisplay similar levels of non- results in high levels of aggregation (Fig. 3).
reducible species (Fig. S10). These findings reveal the conditions It is also possible that chemical modification of solvent-exposed
that promote Trp oxidation lead to the formation of reducible, residues leads to increased antibody self-association and reduced
disulfide-linked species, whereas the conditions that promote colloidal stability of natively (or near natively) folded antibodies.
deamidation lead to the formation of nonreducible cross-linked Therefore, we evaluated the apparent diffusion interaction
species, and that mAb 1 is more prone to the formation of non- parameter (kD*) of the mAbs as a function of pH (Figs. 7, S14 and
reducible crosslinks. S15). These measurements involved evaluating the diffusion coef-
Another potential mechanism by which oxidation or deamida- ficient of the mAbs as a function of antibody concentration using
tion can promote aggregation is through reduction of antibody dynamic light scattering.30 Interestingly, oxidation and deamida-
folding stability and enhancement of antibody unfolding. To test tion had unique, antibody-specific impacts on the kD* values at low
this possibility, we evaluated the apparent melting temperature pH (pH 3.8). Deamidation led to large and significant decreases in
(Tm*) of the mAbs as a function of pH (Fig. 6 and S11). As expected, kD* values for both mAbs, which corresponds to less repulsive (or
the antibody conformational stabilities decreased for all of the more attractive) colloidal interactions. By contrast, Met-oxidized
662 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

Figure 3. Size-exclusion chromatography analysis of the impact of oxidation and deamidation on antibody aggregation as a function of temperature and incubation time at pH 3.8.
(a-d) The antibodies were prepared at 1 mg/mL in citrate-phosphate buffer (pH 3.8) and incubated at (a, b) 37 or (c, d) 25 C for 7 d before analysis by size-exclusion chroma-
tography. The data shown are averages of 3 independent experiments and the error bars are standard errors.

Figure 4. Size-exclusion chromatography analysis of the impact of oxidation and deamidation on antibody aggregation as a function of temperature and incubation time at pH 7.4.
(a-d) The antibodies were prepared at 1 mg/mL in citrate-phosphate buffer (pH 7.4) and incubated at (a, b) 37 or (c, d) 25 C for 28 d before analysis by size-exclusion chroma-
tography. The data shown are averages of 3 independent experiments and the error bars are standard errors.
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 663

Figure 5. SDS-PAGE analysis of the impact of oxidation and deamidation on the for-
mation of reducible and nonreducible crosslinks. (a, b) The antibody samples for (a)
mAb 1 and (b) mAb 2 at pH 7.4 in citrate-phosphate buffer were analyzed using SDS-
PAGE without extended storage (<2 h after preparation). The gels shown are repre-
Figure 6. Evaluation of the impact of oxidation and deamidation on antibody
sentative examples of 3 independent experiments.
conformational stability as a function of pH. (a, b) The apparent melting temperatures
(Tm*) for (a) mAb 1 and (b) mAb 2 in citrate-phosphate buffer as a function of pH. An
extrinsic dye (protein thermal shift dye) was used to obtain the apparent melting
samples showed either little change (mAb 1) or a significant in- temperatures (first unfolding transition) using differential scanning fluorimetry. The
crease (mAb 2) in kD* values. In addition, Trp oxidation led to data shown are averages of 3 independent experiments and the error bars are standard
significantly increased (mAb 1) or decreased (mAb 2) kD* values. At errors. A 2-tailed Student's t-test was used to judge the statistical significance of the
difference between the chemically modified and control mAbs [p-values <0.05 (*),
a higher pH value (pH 7.4), little dependence of kD* values on
<0.01 (**), or <0.001 (***)].
chemical modifications was observed except for deamidation of
mAb 2, which resulted in decreased (more negative) kD values.
These findings reveal that antibody colloidal interactions are
Interestingly, Met oxidation led to minor (mAb 1) and considerable
particularly sensitive to chemical modifications at low pH.
(mAb 2) increases in the formation of acidic species. Given that
deamidation leads to the highest levels of acidic species while Trp
Different Chemical Modifications Uniquely Impact Antibody oxidation generally leads to the highest levels of aggregation, it
Hydrophobic Interactions and Solubility appears that formation of acidic species is not a primary determi-
nant of the aggregation behavior displayed by these antibodies.
Chemical modifications may alter several antibody properties, Next, we evaluated the hydrophobicity of the antibodies using
including their charge, charge distribution, hydrophobicity, and analytical hydrophobic interaction chromatography (HIC) at low pH
solubility.31,42,43 Therefore, we sought to evaluate if oxidation and (pH 3.8) and near-neutral pH (pH 7.4; Fig. 8). Interestingly, the
deamidation adversely affected these antibody properties, and controls displayed significant differences in elution profiles (elution
thereby, contributed to aggregation. We first evaluated the pres- using a negative gradient from 0.6 M to 0 M ammonium sulfate).
ence of acidic or basic species in the control and stressed antibodies mAb 1 displayed higher hydrophobicity and a more complex
using imaged capillary isoelectric focusing (Table 2 and Fig. S16). elution profile (2 main peaks) than mAb 2 (1 main peak). Moreover,
We observed that the various chemical modifications led to the hydrophobicity of mAb 1 was much more sensitive to chemical
increased levels of acidic species compared to the control antibody modifications than that of mAb 2. At low pH (pH 3.8), Met and Trp
samples, whereas the same modifications had little impact on the oxidation of mAb 1 led to significant decreases in hydrophobicity,
levels of basic species. The deamidated samples of both mAbs whereas deamidation led to little change in hydrophobicity
displayed the highest levels of acidic species, which is expected due (Fig. 8a). At higher pH (pH 7.4), the mAbs generally were more
to the formation of negatively charged moieties after deamida- hydrophobic than at low pH (Figs. 8c and 8d), but the relative
tion.31 In addition, oxidation typically promotes the formation of trends in hydrophobicities were generally similar at both pH values.
polar moieties such as methionine sulfoxide, hydroxytryptophan, The chemically modified antibodies displayed relatively high
and N-formylkynurenine,42 and the oxidized samples of both mAbs (>80%) and similar recoveries from the HIC column relative to their
generally displayed increased formation of acidic species. controls (Fig. S17). We also observed similar HIC results using a
664 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

We also evaluated the relative solubilities of the chemically


modified mAbs in ammonium sulfate (Figs. 9 and S19). We adapted
a previously reported, high-throughput method for evaluating the
relative solubility of antibodies.44 This approach involves titrating
antibody solutions over a wide range of ammonium sulfate con-
centrations to identify those that promote antibody precipitation.
As expected, we found that the solubility of all the mAb samples, as
judged by turbidity, decreased as a function of ammonium sulfate
concentration (Figs. 9a, 9b, 9c, 9d and S19). We quantified the
relative solubility, in terms of the ammonium sulfate concentration
that led to half maximal values of the turbidity (CS,50), at multiple
pH values (Figs. 9e, 9f and S19). The Met-oxidized samples gener-
ally displayed little change or increased solubility, consistent with
their reduced hydrophobicity relative to the control samples as
judged by HIC (Fig. 8). The Trp-oxidized samples displayed more
complex behavior, as their solubilities increased in some cases (e.g.,
mAb 1 at pH 7.4) and decreased in other cases (e.g., mAb 2 at pH 7.4;
Figs. 9e and 9f). Moreover, the deamidated samples displayed little
change in solubility. In addition, the antibody solubilities at pH 4.5
were similar to those at pH 3.8, whereas the solubilities at pH 6
were similar to those at pH 7.4 (Figs. 9 and S19). While there is little
correlation between the solubilities of the Trp-oxidized and dea-
midated mAbs in ammonium sulfate and their propensities to
aggregate (in the absence of ammonium sulfate), the findings for
the Met-oxidized samples suggest that their increased solubilities
in ammonium sulfate may be linked to their reduced hydropho-
bicities and low aggregation propensities.

Discussion

Our experimental findings are summarized in Table 3 and


deserve further consideration. One interesting aspect of our find-
ings is that the mechanisms by which chemical modifications lead
Figure 7. Evaluation of the impact of oxidation and deamidation on antibody colloidal
to antibody aggregation are weakly correlated with changes in
interactions as a function of pH. (a, b) The apparent diffusion interaction parameters
(kD*) were evaluated using dynamic light scattering for (a) mAb 1 and (b) mAb 2 as a antibody conformational stability. While the simplest explanation
function of pH. The experiments were performed in citrate-phosphate buffer at the for the impact of chemical modifications on antibody aggregation is
reported pH values. The data shown are averages of 3 independent experiments and that they lead to reduced conformational stability and thereby
the error bars are standard errors. A 2-tailed Student's t-test was used to judge the
increased aggregation, our results reveal significantly different
statistical significance of the difference between the chemically modified and control
mAbs [p-values <0.05 (*), <0.01 (**), or <0.001 (***)].
aggregation mechanisms. The molecular basis for our findings ap-
pears explainable (at least in part) based on the sites in the mAbs
that were modified. For the Met-oxidized samples, the modifica-
broader salt gradient (from 1 to 0 M ammonium sulfate; Fig. S18).
tions were primarily observed at sites in the Fc region that are
These findings suggest that the reduced hydrophobicity of the Met-
known to reduce antibody conformational stability,9,10 as we
oxidized samples may be responsible for their low levels of self-
observed in this work (Fig. 6). The moderate reduction in stability is
association and aggregation, while the aggregation behaviors
not sufficient to promote aggregation except at conditions of low
observed for the Trp-oxidized and deamidated samples do not
pH (pH 3.8; apparent Tm reduced from 47.8 C to 41.7 C for mAb 1
appear to be mediated by increased antibody hydrophobicity.
and from 42.3 C to 36.2 C for mAb 2) and elevated temperature
(37 C). Under these particular conditions, the conformational sta-
Table 2 bility of the Met-oxidized samples (apparent Tm of 36.2 C-41.7 C) is
Summary of the Levels of Acidic and Basic Species of the Antibodies Evaluated in This
too low to prevent antibody aggregation. Conversely, the methods
Study
used in this work for selective Trp oxidation primarily lead to
Acidic Species (%) Main Species (%) Basic Species (%) modification of a single Trp residue in the CDRs (which is different
mAb 1 for each mAb), and this limited modification appears to promote
Control 41.5 52.3 6.2 aggregation via mechanisms that are distinct from those linked to
H2O2 43.4 50.4 6.3 reduced antibody conformational stability. Likewise, Asn deami-
AAPH þ Met 61.5 29.7 8.9
High pH 91.2 6.9 1.9
dation occurs primarily in the Fc region at the well-known Asn sites
mAb 2 in the PENNY peptide, and the lack of conformational destabiliza-
Control 28.7 63.9 7.4 tion (as observed previously27) also suggests that deamidation
H2O2 48.0 46.0 6.1 promotes aggregation via mechanisms distinct from those linked to
AAPH þ Met 45.5 47.1 7.5
reduced conformational stability.
High pH 61.7 34.5 3.8
Therefore, the aggregation mechanisms for the mAbs subject to
Imaged capillary isoelectric focusing (iCIEF) was performed on the control, Met- Trp oxidation and Asn deamidationdwhich cannot be easily
oxidized (H2O2-treated), Trp-oxidized (AAPH þ Met-treated) and deamidated
(high pH-treated) mAbs to determine the level of charged species for each sample.
explained by impacts of the chemical modifications on conforma-
The main peak represents unmodified antibody, and the acidic and basic species tional stabilitydalso deserve further consideration. We find that
represent charged species. The data are for 1 to 3 replicates. Trp oxidation leads to formation of reducible crosslinks, which are
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 665

Figure 8. Evaluation of the impact of oxidation and deamidation on antibody hydrophobicity as a function of pH. (a-d) The antibodies were prepared at 1 mg/mL in citrate-
phosphate buffer with 0.6 M ammonium sulfate at (a, b) pH 3.8 and (c, d) pH 7.4, and injected into the column without extended incubation (<2 h after preparation). The col-
umn was equilibrated at the same conditions as the antibody samples, and then a gradient of ammonium sulfate (0.6 to 0 M) was used to elute the antibodies. The chromatograms
shown are representative examples of 3 independent experiments.

presumably disulfide bonds and may form due to disulfide although it had little effect on the rate of Met oxidation.41 Various
scrambling. One potential mechanism for this phenomenon is that observations have been reported for the effects of sucrose on the
chemical oxidation of Trp can generate Trp radicals with delo- oxidation of other proteins, including slight protection for
calized electrons, which can in turn reduce disulfide bonds45,46 and interleukin-7 Fc fusion protein52 and subtilisin,53 no effect for tu-
lead to disulfide scrambling. Another potential mechanism by mor necrosis factor receptor 1,54 and slight worsening for gran-
which Trp oxidation may lead to disulfide scrambling is through ulocyte colony-stimulating factor55 and recombinant activated
partial unfolding at or near Trp oxidation sites, which in turn may factor VII.51 Moreover, sucrose has been shown to decrease the rate
lead to the exposure of buried cysteines and thereby increase the of Asn deamidation (or formation of acidic species) for mAbs56 and
likelihood of disulfide scrambling.47 Indeed, our mass spectrometry acrylodan-conjugated glucose-binding protein.50 Whether the
results are potentially consistent with this mechanism because addition of excipients, such as sucrose or salts, would differentially
mAb 1 displays higher levels of Trp di-oxidationdwhich can modify the aggregation propensity of oxidized and deamidated
disrupt the Trp indole ring and cause local structural changes mAbs requires additional studies in the future.
through formation of N-formylkynureninedand this may lead to The formation of nonreducible crosslinks during antibody
greater amounts of reducible aggregates for mAb 1 relative to mAb storage at high pH (pH 9.5) to promote Asn deamidation also de-
2. It is also notable that previous studies of photo-oxidation of serves further consideration. The fact that these crosslinks (which
antibodies found that disulfide bond shuffling can occur via primarily form for mAb 1) are nonreducible suggests mechanisms
oxidation of Trp residues located in close proximity to intra- other than disulfide scrambling, although cases of the formation of
molecular disulfide bonds through electron transfer or singlet ox- reduction-resistant disulfide bonds through disulfide scrambling
ygen pathways, and such disulfide scrambling also promotes have been reported.20 It is notable that previous studies have
antibody aggregation.20,21,48 However, whether similar or unique identified similar types of nonreducible crosslinks for mAbs incu-
mechanisms mediate disulfide scrambling for chemical oxidation bated at similar conditions as those used in our study for promoting
and photo-oxidation of antibodies is unclear and requires more deamidation, namely elevated temperature (e.g., 37 C-45 C) and
detailed studies in the future. high pH (pH 8-10).40,57-60 Moreover, the observed degree of
It is also important to consider that the aggregation behavior of crosslink formation was significantly different for different anti-
mAbs is strongly influenced by the type of stress and the specific bodies,57 as we observed in our study. One potential mechanism for
formulation conditions. In particular, sucrose and other sugars have the generation of nonreducible species involving antibody heavy
been shown to improve the conformational stability of proteins and and light chains is thioether bond formation in the hinge region.
antibodies through the excluded volume effect and thereby reduce Previous studies have shown that disulfide bonds at high pH can
the rate of aggregation.49-51 In the case of photodegraded mAbs, decompose via a b-elimination reaction,61-63 which leads to in-
sucrose was also observed to reduce the rate of aggregation, termediates that can form nonreducible thioether bonds between
666 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

Figure 9. Evaluation of the impact of oxidation and deamidation on antibody solubility in the presence of ammonium sulfate as a function of pH. (a-d) The antibodies were
evaluated at 0.2 mg/mL in citrate-phosphate buffer at (a, b) pH 3.8 and (c, d) pH 7.4 as a function of ammonium sulfate concentration. The turbidity values were measured and
normalized using the maximum and minimum turbidity values for a given set of samples (i.e., samples for the same mAb and pH). (e-f) The ammonium sulfate midpoints of
transition (CS,50) for each mAb and pH condition. The data shown are averages of 4 independent experiments and the error bars are standard errors. A 2-tailed Student's t-test was
used to judge the statistical significance of the difference between the chemically modified and control mAbs [p-values <0.05 (*), <0.01 (**), or 0.001 (***)].

modified cysteines in the CH1 and CL domains.39,59 This mechanism hydrophobicities of the mAbs in different ways and lead to distinct
has been reported for IgG1s,39 and the IgG1 antibody in our study HIC retention behaviors.66 For Met oxidation, our mass spectrom-
(mAb 1) shows much higher levels of such nonreducible crosslinks etry analysis revealed that the oxidation levels of the 2 mAbs were
between heavy and light chain than the IgG4 antibody in our study similar in the Fc regions but markedly different in the Fab regions
(mAb 2). We speculate that the different disulfide bonding patterns (Table 1), which suggests that the Fab region is primarily respon-
between CH1 and CL in IgG1 and IgG4 antibodies result in different sible for the observed differences in hydrophobicity for the Met-
susceptibilities to thioether bond formation, which will need to be oxidized samples. For Trp oxidation, the oxidized samples for the
tested using additional antibodies in the future. 2 mAbs display markedly different HIC retention behavior despite
Another interesting aspect of our findings is the HIC retention the fact that both mAbs primarily undergo Trp oxidation at one site
behavior of the oxidized mAbs. Previous studies have also found in the CDRs, which is different for each mAb. One likely explanation
that Met and Trp oxidation lead to the formation of species with for this behavior is that mAb 1 displays much higher levels of Trp
reduced hydrophobicity,43 as expected based on the nature of the di-oxidation (N-formylkynurenine or dihydroxytryptophan) rela-
modifications.64,65 While we found that Met and Trp oxidation led tive to mAb 2, which is expected to be more hydrophilic than Trp
to significantly reduced HIC retention times for mAb 1, these mono-oxidation (hydroxytryptophan) that was primarily observed
modifications had little impact on the HIC behavior for mAb 2. This for mAb 2. These speculative ideas require additional experimen-
observation suggests that local chemical or structural changes, in tation to be properly evaluated.
addition to changes in global antibody properties, caused by these The HIC retention behavior of the deamidated mAbs also de-
chemical modifications may induce changes in the surface serves further consideration. Previous studies have found that
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 667

Table 3
Summary of the Experimental Analysis of the Effects of Chemical Modifications on Various Antibody Biophysical Properties

Met Oxidation Trp Oxidation Asn Deamidation

pH 3.8 pH 7.4 pH 3.8 pH 7.4 pH 3.8 pH 7.4

mAb 1
Aggregation (SEC, %HMW) NC NC þþ þþþ þa þþþ
Aggregation (DLS, HMW species) NC NC þþþ þþ þþþ NC
Accelerated Stability (37 C, %HMW) þþþ þ ‒ þ þþþ þ
Reducible HMW Species (SDS-PAGE)  NC  þþþ  NC
Nonreducible Species (SDS-PAGE)  NC  NC  þþþ
Conformational Stability (DSF, Tm*) ‒‒ ‒‒ NC NC NC NC
Colloidal Stability (kD*) NC NC þþ ‒ ‒‒ NC
Hydrophobicity (HIC, retention time) ‒‒‒ ‒‒‒ ‒‒ ‒‒‒ NC NC
Solubility (ammonium sulfate, CS,50) NC þþþ NC þþ NC NC
mAb 2
Aggregation (SEC, %HMW) NC NC þþ þþþ þ þa þþþ
Aggregation (DLS, HMW species) NC NC NC NC þþþ þþþ
Accelerated Stability (37 C, %HMW) þþþ NC NC þþ þþþ þ
Reducible HMW Species (SDS-PAGE)  NC  þþ  NC
Nonreducible Species (SDS-PAGE)  NC  NC  þ
Conformational Stability (DSF, Tm*) ‒‒ ‒‒‒ NC ‒‒ NC NC
Colloidal Stability (kD*) þ NC ‒ NC ‒‒‒ ‒
Hydrophobicity (HIC, retention time) ‒ ‒ þ NC NC NC
Solubility (ammonium sulfate, CS,50) þ þþþ ‒‒ ‒‒ ‒ þ

The observed differences between the Met-oxidized, Trp-oxidized, or Asn-deamidated mAbs relative to the control mAbs at pH 3.8 and pH 7.4 are summarized for mAbs 1 and
2. A plus sign (þ) denotes an increase in the measured value of a particular biophysical property relative to the control mAb (e.g., wild-type mAb 1), whereas a minus sign (‒)
denotes a decrease. More than one plus or minus sign indicate larger changes.
, not evaluated; NC, no change.
a
At high concentration only.

deamidation can lead to increased or decreased HIC retention times Acknowledgments


depending on the formation of succinimide67-71 or aspartic acid
(and isoaspartic acid),43,70-73 respectively. The degradation condi- The authors thank members of the Tessier lab for their assis-
tion (high pH) used in this study to generate the deamidated mAbs tance editing the manuscript. The authors would also like to thank
typically leads to the formation of aspartic acid and isoaspartic acid Pritesh Patel and Shannon Flagg for providing the iCIEF data. In
due to the rapid hydrolysis of the succinimide intermediate formed addition, they thank Helen Zha for allowing them to use her DSC
during deamidation of Asn at basic pH.67,70,74 The fact that we calorimeter. This work was supported by Bristol-Myers Squibb and
observed little impact of deamidation on the HIC retention the Albert M. Mattocks Chair (to P.M.T.).
behavior may be linked to the low levels of deamidation observed
in the Fab region for both mAbs (Table 1). Furthermore, the fact that
our HIC retention behavior did not correlate well with the observed
References
aggregation propensity of the chemically modified mAbs is
consistent with previous studies.75,76 The disconnect between HIC 1. Ecker DM, Jones SD, Levine HL. The therapeutic monoclonal antibody market.
measurements and aggregation behavior suggests that hydropho- mAbs. 2015;7(1):9-14.
bic interactions are not a primary determinant of aggregation for 2. Grilo AL, Mantalaris A. The increasingly human and profitable monoclonal
antibody market. Trends Biotechnol. 2019;37(1):9-16.
the mAbs in our study and that other mechanisms (e.g., covalent 3. Nelson AL, Dhimolea E, Reichert JM. Development trends for human mono-
crosslinking and colloidal instability) play a more significant role in clonal antibody therapeutics. Nat Rev Drug Discov. 2010;9:767.
mediating aggregation. 4. Tiller KE, Tessier PM. Advances in antibody design. Annu Rev Biomed Eng.
2015;17(1):191-216.
5. Maynard J, Georgiou G. Antibody engineering. Annu Rev Biomed Eng. 2000;2(1):
339-376.
Conclusions 6. Carter PJ. Potent antibody therapeutics by design. Nat Rev Immunol. 2006;6(5):
343-357.
The impact of chemical modifications on antibody physical 7. Winter G, Griffiths AD, Hawkins RE, Hoogenboom HR. Making antibodies by
phage display technology. Annu Rev Immunol. 1994;12(1):433-455.
stability can be variable and difficult to predict. In this work, we 8. Chao G, Lau WL, Hackel BJ, Sazinsky SL, Lippow SM, Wittrup KD. Isolating and
found that antibody conformational stability is weakly correlated engineering human antibodies using yeast surface display. Nat Protoc.
with aggregation except in extreme cases and that aggregation can 2006;1(2):755-768.
9. Liu D, Ren D, Huang H, et al. Structure and stability changes of human IgG1 Fc
occur via multiple mechanisms such as chemical crosslinking and as a consequence of methionine oxidation. Biochemistry. 2008;47(18):5088-
reduced colloidal stability. Our findings underscore the importance 5100.
of evaluating the physical stability of antibodies at different con- 10. Chumsae C, Gaza-Bulseco G, Sun J, Liu H. Comparison of methionine oxidation
in thermal stability and chemically stressed samples of a fully human mono-
ditions relevant to antibody manufacture, purification and formu- clonal antibody. J Chromatogr B. 2007;850:285-294.
lation to understand the effects of chemical modifications on 11. Wang S, Ionescu R, Peekhaus N, Leung J, Ha S, Vlasak J. Separation of post-
aggregation as well as evaluating a combination of properties (e.g., translational modifications in monoclonal antibodies by exploiting subtle
conformational changes under mildly acidic conditions. J Chromatogr A.
crosslinking, charge, polarity, hydrophobicity and solubility) to
2010;1217(42):6496-6502.
elucidate the mechanisms of antibody aggregation. Future work 12. Liu H, Gaza-Bulseco G, Xiang T, Chumsae C. Structural effect of deglycosylation
should evaluate the generality of our findings by expanding these and methionine oxidation on a recombinant monoclonal antibody. Mol
studies to include additional antibodies. We expect these studies Immunol. 2008;45(3):701-708.
13. Burkitt W, Domann P, O’Connor G. Conformational changes in oxidatively
will be important for improving the generation and production of stressed monoclonal antibodies studied by hydrogen exchange mass spec-
therapeutic antibodies with high chemical and physical stability. trometry. Protein Sci. 2010;19(4):826-835.
668 M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669

14. Houde D, Peng Y, Berkowitz SA, Engen JR. Post-translational modifications 44. Yamniuk AP, Ditto N, Patel M, et al. Application of a Kosmotrope-based solu-
differentially affect IgG1 conformation and receptor binding. Mol Cell Prote- bility assay to multiple protein therapeutic classes indicates broad use as a
omics. 2010;9(8):1716-1728. high-throughput screen for protein therapeutic aggregation propensity.
15. Pan H, Chen K, Chu L, Kinderman F, Apostol I, Huang G. Methionine oxidation in J Pharm Sci. 2013;102(8):2424-2439.
human IgG2 Fc decreases binding affinities to protein A and FcRn. Protein Sci. 45. Arenas A, Lo pez-Alarco
n C, Kogan M, Lissi E, Davies MJ, Silva E. Chemical
2009;18(2):424-433. modification of lysozyme, glucose 6-phosphate dehydrogenase, and bovine eye
16. Bertolotti-Ciarlet A, Wang W, Lownes R, et al. Impact of methionine oxidation lens proteins induced by peroxyl radicals: role of oxidizable amino acid resi-
on the binding of human IgG1 to FcRn and Fcg receptors. Mol Immunol. dues. Chem Res Toxicol. 2013;26(1):67-77.
2009;46(8-9):1878-1882. 46. Hawkins CL, Davies MJ. Generation and propagation of radical reactions on
17. Hensel M, Steurer R, Fichtl J, et al. Identification of potential sites for trypto- proteins. Biochim Biophys Acta. 2001;1504(2):196-219.
phan oxidation in recombinant antibodies using tert-butylhydroperoxide and 47. Brych SR, Gokarn YR, Hultgen H, Stevenson RJ, Rajan R, Matsumura M. Char-
quantitative LC-MS. PLoS One. 2011;6(3):e17708. acterization of antibody aggregation: role of buried, unpaired cysteines in
18. Wei Z, Feng J, Lin H-Y, et al. Identification of a single tryptophan residue as particle formation. J Pharm Sci. 2010;99(2):764-781.
critical for binding activity in a humanized monoclonal antibody against res- 48. Vanhooren A, Devreese B, Vanhee K, Van Beeumen J, Hanssens I. Photoexci-
piratory syncytial virus. Anal Chem. 2007;79(7):2797-2805. tation of tryptophan groups induces reduction of two disulfide bonds in goat a-
19. Hageman T, Wei H, Kuehne P, et al. Impact of tryptophan oxidation in lactalbumin. Biochemistry. 2002;41(36):11035-11043.
complementarity-determining regions of two monoclonal antibodies on 49. Meyer JD, Nayar R, Manning MC. Impact of bulking agents on the stability of a
structure-function characterized by hydrogen-deuterium exchange mass lyophilized monoclonal antibody. Eur J Pharm Sci. 2009;38(1):29-38.
spectrometry and surface plasmon resonance. Pharm Res. 2018;36(1):24. 50. Sahni N, Chaudhuri R, Hickey JM, et al. Preformulation characterization, sta-
20. Wecksler AT, Yin J, Lee Tao P, Kabakoff B, Sreedhara A, Deperalta G. Photo- bilization, and formulation design for the acrylodan-labeled glucose-binding
disruption of the structurally conserved cys-cys-trp triads leads to reduction- protein SM4-AC. J Pharm Sci. 2017;106(5):1197-1210.
resistant scrambled intrachain disulfides in an IgG1 monoclonal antibody. 51. Soenderkaer S, Carpenter JF, van de Weert M, Hansen LL, Flink J, Frokjaer S.
Mol Pharm. 2018;15(4):1598-1606. Effects of sucrose on rFVIIa aggregation and methionine oxidation. Eur J Pharm
s E, Staelens S, Vanhooren A, Deckmyn H, Hanssens I, Majer Z. Role of the
21. Illye Sci. 2004;21(5):597-606.
trp-disulfide triads in the UV light induced degradation of a monoclonal 52. Lim JY, Kim NA, Lim DG, Eun C, Choi D, Jeong SH. Biophysical stability of hyFc
antibody scFv. Int J Biochem Res Rev. 2014;4:367-385. fusion protein with regards to buffers and various excipients. Int J Biol Mac-
22. Luo Q, Joubert MK, Stevenson R, Ketchem RR, Narhi LO, Wypych J. Chemical romol. 2016;86:622-629.
modifications in therapeutic protein aggregates generated under different 53. DePaz RA, Barnett CC, Dale DA, Carpenter JF, Gaertner AL, Randolph TW. The
stress conditions. J Biol Chem. 2011;286(28):25134-25144. excluding effects of sucrose on a protein chemical degradation pathway:
23. Joubert MK, Luo Q, Nashed-Samuel Y, Wypych J, Narhi LO. Classification and methionine oxidation in subtilisin. Arch Biochem Biophys. 2000;384(1):123-132.
characterization of therapeutic antibody aggregates. J Biol Chem. 2011;286(28): 54. Roy S, Mason BD, Scho € neich CS, Carpenter JF, Boone TC, Kerwin BA. Light-
25118-25133. induced aggregation of type I soluble tumor necrosis factor receptor. J Pharm
24. Vlasak J, Bussat MC, Wang S, et al. Identification and characterization of Sci. 2009;98(9):3182-3199.
asparagine deamidation in the light chain CDR1 of a humanized IgG1 antibody. 55. Yin J, Chu J-W, Ricci MS, Brems DN, Wang DIC, Trout BL. Effects of excipients on
Anal Biochem. 2009;392(2):145-154. the hydrogen peroxide-induced oxidation of methionine residues in gran-
25. Yan B, Steen S, Hambly D, et al. Succinimide formation at Asn 55 in the ulocyte colony-stimulating factor. Pharm Res. 2005;22(1):141-147.
complementarity determining region of a recombinant monoclonal antibody 56. Alekseychyk L, Su C, Becker GW, Treuheit MJ, Razinkov VI. High-throughput
IgG1 heavy chain. J Pharm Sci. 2009;98(10):3509-3521. screening and analysis of charge variants of monoclonal antibodies in multiple
26. Qiu H, Wei R, Jaworski J, et al. Engineering an anti-CD52 antibody for enhanced formulations. SLAS Discov. 2017;22(8):1044-1052.
deamidation stability. mAbs. 2019;11:1266-1275. 57. Jiskoot W, Beuvery EC, de Koning AAM, Herron JN, Crommelin DJA. Analytical
27. Alam ME, Barnett GV, Slaney TR, Starr CG, Das TK, Tessier PM. Deamidation can approaches to the study of monoclonal antibody stability. Pharm Res.
compromise antibody colloidal stability and enhance aggregation in a pH- 1990;07(12):1234-1241.
dependent manner. Mol Pharm. 2019;16(5):1939-1949. 58. Zhang Q, Schenauer MR, McCarter JD, Flynn GC. IgG1 thioether bond formation
28. Tang L, Sundaram S, Zhang J, et al. Conformational characterization of the in vivo. J Biol Chem. 2013;288(23):16371-16382.
charge variants of a human IgG1 monoclonal antibody using H/D exchange 59. Cohen SL, Price C, Vlasak J. b-elimination and peptide bond hydrolysis: two
mass spectrometry. mAbs. 2013;5(1):114-125. distinct mechanisms of human IgG1 hinge fragmentation upon storage. J Am
29. Menzen T, Friess W. High-throughput melting-temperature analysis of a Chem Soc. 2007;129(22):6976-6977.
monoclonal antibody by differential scanning Fluorimetry in the presence of 60. Zhang Q, Flynn GC. Cysteine racemization on IgG heavy and light chains. J Biol
surfactants. J Pharm Sci. 2013;102(2):415-428. Chem. 2013;288(48):34325-34335.
30. Nicoud L, Lattuada M, Yates A, Morbidelli M. Impact of aggregate formation on 61. Nashef AS, Osuga DT, Lee HS, Ahmed AI, Whitaker JR, Feeney RE. Effects of alkali on
the viscosity of protein solutions. Soft Matter. 2015;11(27):5513-5522. proteins. Disulfides and their products. J Agric Food Chem. 1977;25(2):245-251.
31. Pace AL, Wong RL, Zhang YT, Kao Y-H, Wang YJ. Asparagine deamidation 62. Florence TM. Degradation of protein disulphide bonds in dilute alkali. Biochem
dependence on buffer type, pH, and temperature. J Pharm Sci. 2013;102(6): J. 1980;189(3):507-520.
1712-1723. 63. Galande AK, Trent JO, Spatola AF. Understanding base-assisted desulfurization
32. Yan Q, Huang M, Lewis MJ, Hu P. Structure based prediction of asparagine using a variety of disulfide-bridged peptides. Pept Sci. 2003;71(5):534-551.
deamidation propensity in monoclonal antibodies. mAbs. 2018;10(6):901-912. 64. Robinson JR, Karkov HS, Woo JA, Krogh BO, Cramer SM. QSAR models for
33. Shukla AA, Hubbard B, Tressel T, Guhan S, Low D. Downstream processing of prediction of chromatographic behavior of homologous Fab variants. Biotechnol
monoclonal antibodiesdapplication of platform approaches. J Chromatogr B. Bioeng. 2017;114(6):1231-1240.
2007;848:28-39. 65. Jetha A, Thorsteinson N, Jmeian Y, Jeganathan A, Giblin P, Fransson J. Homology
34. Shukla AA, Tho € mmes J. Recent advances in large-scale production of mono- modeling and structure-based design improve hydrophobic interaction chro-
clonal antibodies and related proteins. Trends Biotechnol. 2010;28(5):253-261. matography behavior of integrin binding antibodies. mAbs. 2018;10(6):890-
35. Mazzer AR, Perraud X, Halley J, O’Hara J, Bracewell DG. Protein A chromatog- 900.
raphy increases monoclonal antibody aggregation rate during subsequent low 66. King C, Patel R, Ponniah G, et al. Characterization of recombinant monoclonal
pH virus inactivation hold. J Chromatogr A. 2015;1415:83-90. antibody variants detected by hydrophobic interaction chromatography and
36. Low D, O ’leary R, Pujar NS. Future of antibody purification. J Chromatogr B. imaged capillary isoelectric focusing electrophoresis. J Chromatogr B.
2007;848:48-63. 2018;1085:96-103.
37. Hober S, Nord K, Linhult M. Protein A chromatography for antibody purifica- 67. Ouellette D, Chumsae C, Clabbers A, Radziejewski C, Correia I. Comparison of
tion. J Chromatogr B. 2007;848:40-47. the in vitro and in vivo stability of a succinimide intermediate observed on a
38. Nobbmann U, Connah M, Fish B, et al. Dynamic light scattering as a relative tool therapeutic IgG1 molecule. mAbs. 2013;5:432-444.
for assessing the molecular integrity and stability of monoclonal antibodies. 68. Kwong MY, Harris RJ. Identification of succinimide sites in proteins by N-ter-
Biotechnol Genet Eng Rev. 2007;24(1):117-128. minal sequence analysis after alkaline hydroxylamine cleavage. Protein Sci.
39. Tous GI, Wei Z, Feng J, et al. Characterization of a novel modification to 1994;3(1):147-149.
monoclonal antibodies: thioether cross-link of heavy and light chains. Anal 69. Valliere-Douglass J, Wallace A, Balland A. Separation of populations of antibody
Chem. 2005;77(9):2675-2682. variants by fine tuning of hydrophobic-interaction chromatography operating
40. Kroon DJ, Baldwin-Ferro A, Lalan P. Identification of sites of degradation in a conditions. J Chromatogr A. 2008;1214:81-89.
therapeutic monoclonal antibody by peptide mapping. Pharm Res. 1992;9(11): 70. Cao M, Mulagapati SHR, Vemulapalli B, et al. Characterization and quantifica-
1386-1393. tion of succinimide using peptide mapping under low-pH conditions and hy-
41. Shah DD, Zhang J, Maity H, Mallela KMG. Effect of photo-degradation on the drophobic interaction chromatography. Anal Biochem. 2019;566:151-159.
structure, stability, aggregation, and function of an IgG1 monoclonal antibody. 71. Valliere-Douglass J, Jones L, Shpektor D, et al. Separation and characterization
Int J Pharm. 2018;547(1-2):438-449. of an IgG2 antibody containing a cyclic imide in CDR1 of light chain by hy-
42. Torosantucci R, Scho € neich C, Jiskoot W. Oxidation of therapeutic proteins and drophobic interaction chromatography and mass spectrometry. Biotechnol Appl
peptides: structural and biological consequences. Pharm Res. 2014;31(3):541- Biochem. 2002;18(2):8.
553. 72. Wakankar AA, Borchardt RT, Eigenbrot C, et al. Aspartate isomerization in the
43. Boyd D, Kaschak T, Yan B. HIC resolution of an IgG1 with an oxidized Trp in a complementarity-determining regions of two closely related monoclonal an-
complementarity determining region. J Chromatogr B. 2011;879(13):955-960. tibodies. Biochemistry. 2007;46(6):1534-1544.
M.E. Alam et al. / Journal of Pharmaceutical Sciences 109 (2020) 656-669 669

73. Harris RJ, Kabakoff B, Macchi FD, et al. Identification of multiple sources of charge 75. Shieh IC, Patel AR. Predicting the agitation-induced aggregation of mono-
heterogeneity in a recombinant antibody. J Chromatogr B. 2001;752(2):233-245. clonal antibodies using surface tensiometry. Mol Pharm. 2015;12(9):3184-
74. Sreedhara A, Cordoba A, Zhu Q, Kwong J, Liu J. Characterization of the isom- 3193.
erization products of aspartate residues at two different sites in a monoclonal 76. Jain T, Sun T, Durand S, et al. Biophysical properties of the clinical-stage anti-
antibody. Pharm Res. 2012;29(1):187-197. body landscape. Proc Natl Acad Sci U S A. 2017;114(5):944-949.

You might also like