Optimization of a platform process operating space for a monoclonal antibody susceptible to reversible and irreversible aggregation using a solution stability screening approach

Journal of Chromatography A, 1597 (2019) 100–108

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Optimization of a platform process operating space for a monoclonal
antibody susceptible to reversible and irreversible aggregation using a
solution stability screening approach
Adrian Man a , Haibin Luo a , Sophia V. Levitskaya b , Nathaniel Macapagal a ,
Kelcy J. Newell a,∗
a
b

Purification Process Sciences, AstraZeneca, One MedImmune Way, Gaithersburg, MD, 20878, USA
Analytical Sciences, AstraZeneca, One MedImmune Way, Gaithersburg, MD, 20878, USA

a r t i c l e

i n f o

Article history:
Received 19 September 2018
Received in revised form 12 March 2019
Accepted 13 March 2019
Available online 20 March 2019
Keywords:
Monoclonal antibody purification
Automation
High throughput process development

Aggregation
Scale-down
Reversible self-association

a b s t r a c t
Platform manufacturing processes are widely adopted to simplify and standardize the development and
manufacturing of monoclonal antibodies (mAbs). However, there are mAbs that do not conform to a
platform design due to instability or other protein properties leading to a negative impact on product
quality or process performance (non-platform mAb). Non-platform mAbs typically require prolonged
development times and significant deviations from the platform process to address these issues due to
the need to sequentially optimize individual process steps.
In this study, we describe an IgG2 mAb (mAb A) that is susceptible to aggregation and reversible
self-association (RSA) under platform conditions. In lieu of a sequential optimization approach, we evaluated the solution stability of mAb A across the platform operating space (solution stability screen). This
screening design was used to identify interacting parameters that affected the non-platform mAb stability. A subsequent response surface design was found to predict an acceptable operating space that
minimized aggregate formation and RSA across the entire process. This information guided the selection
of optimal parameters best suited to avoid destabilizing conditions for each process step. Substantial
time savings was achieved by focusing development around these factors including protein concentration, buffer pH, salt concentration, and excipient type. In addition, this work enabled the optimization of
a cation exchange chromatography step that removed aggregate without yield losses due to the presence
of reversible aggregation. The final optimized process derived from this study resulted in an increase in
˜
yield of 30%
over the original process while maintaining the same level of aggregate clearance to match
product quality.
Solution stability screening is readily adapted to high throughput technologies to minimize material
requirements and accelerate analytical data availability. Implementation of high throughput approaches
will further expedite process development and enable enhanced selection of candidate drugs by including
process development objectives.
© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license ( />
1. Introduction

Monoclonal antibodies (mAbs) have emerged as a rapidly growing class of therapeutic since the mid-1990s. According to the FDA
database there are more than 70 modern antibody-based therapeutic agents approved in the US and more than 500 additional

∗ Corresponding author.
E-mail address: (K.J. Newell).

products are currently in clinical development [1]. Many biopharmaceutical companies have adopted platform mAb purification
processes to simplify process development and manufacture of
mAbs [2]. Fig. 1 illustrates a common platform process with commonly used steps including; (1) affinity purification capture, (2)
low pH virus inactivation, (3) anion exchange chromatography for
process related impurity and virus removal, (4) cation exchange
chromatography for process and product related impurity removal
(5) virus filtration, and (6) formulation that utilizes (a) ultrafiltration and (b) diafiltration to generate drug substance.

/>0021-9673/© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( />0/).


A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

101

Table 1
Summary of non-ideal observations for a mAb in the platform process.
Step

1
Protein A
Chromatography
2
Low pH Inactivation

3
Anion Exchange
Chromatography
4
Cation Exchange
Chromatography
5
Virus Filtration
6A
UF1
6B
UF2

a
b
c

Condition
pH

Conductivity
(mS/cm)

Protein conc.
(mg/mL)

3.6-4.5

0.5-1.5

10-25

3.4-3.6

<5

9-23

7.3-7.6

<5

7-20

4.8-5.2

10-25

3-15

4.8-5.2

10-25

3-15

4.8-5.2

10-25

50-70

5.8-6.2

<3

125-135

Condition Categories

mAb A Observations

Low pHa
Low saltb
High protein
concentrationc

High
aggregate in product

Neutral pH
Low salt
High protein concentration
Intermediate pH
High salt
Intermediate protein
concentration

Peak tailing

Intermediate pH
High salt
High protein concentration
Intermediate pH
Low salt
Very high protein
concentration

Aggregate formation

Peak splitting, aggregate
formation,
low yield
Low
Flux
High feed pressure/Low
Flux
High feed pressure/Low
flux

pH range was categorized using following criteria: Low (<4.6), Intermediate (4.6–6.5), Neutral (6.6–8.0), High (>8).
Buffer ionic strength was categorized using following criteria: Low (<5 mS/cm), Intermediate (5–10 mS/cm), High (>10 mS/cm).
Protein concentration was categorized using following criteria: Low (<1 mg/mL), Intermediate (1–10 mg/mL), High (11–100 mg/mL) and Very High (>100 mg/mL).

Fig. 1. Standard platform manufacturing process for monoclonal antibodies.

A well-designed platform process is expected to robustly
remove the process and product related impurities for the majority of therapeutic mAbs used for clinical development. While many
monoclonal antibodies are easily manufactured using a platform
process with minor adjustments, difficulties arise with mAbs that

do not fit into the established platform design due to instability in solution (non-platform mAbs). This problem can result in
prolonged development times, supply chain concerns due to implementation of novel technologies and delayed clinical timelines.
Protein instability in solution can be due to multiple causes related

to mAb structural heterogeneity and result in the formation of
product related impurities [3–7]. Reversible self-association (RSA)
is a unique solution property in which native, reversible oligomeric
species are formed as a result of non-covalent intermolecular
interactions [8,9]. Unlike irreversible aggregates (aggregates), RSA
species exist at equilibrium with monomer and the modification
of solution conditions such as pH, ion concentration, or temperature can drive the equilibrium to favor monomer over RSA. If RSA
has a slow dissociation rate, it can be detected by size exclusion
chromatography. If dissociation constants are short, the primary
methods of RSA detection are analytical methods that detect the
hydrodynamic radius of a mixture of species or sedimentation rates
of multiple species such as dynamic light scattering (former) and
analytical ultracentrifugation (latter). The primary challenge of RSA
species is that they frequently behave like aggregates making it
very challenging to remove aggregates when solution conditions
favor RSA over monomer. RSA has been demonstrated to have a
large impact on cation exchange chromatography [10] and can
also impact solution viscosity [11]. Specifically, Luo et al demonstrated that elution salt concentrations that favored RSA resulted
in low column yields and the formation of irreversible aggregates
that appeared to be mediated by the presence of the chromatography resin. The formation of RSA has the potential to impact
virtually every other step of the platform process particularly if the
RSA species forms at conditions where mAb monomers are usually stable (Table 1). Until recently the experimental techniques
employed to address RSA effects upon downstream process have
been reactive rather than predictive with development timelines
being delayed while process changes were evaluated and for robust
solutions.

In this work, we used a non-platform mAb that exhibited atypical solution behaviors when manufactured with the platform
process (Table 1). A Design of Experiments (DoE) methodology was
used to screen for solution condition factors that impact these atypical behaviors and identify more stable solution conditions. The
results were applied to each process step with the goal of identifying operating conditions that would enable good performance
for the non-platform mAb with only minor deviations from the
platform process.


102

A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

2. Materials and MethodS
2.1. Chemicals and reagents
2.1.1. Buffer salts and excipients
Buffer salts were obtained from J.T. Baker (Center Valley, PA)
and were of reagent grade or higher. Concentrated stock solutions
were made by the dissolution of the solid components in reverse
osmosis distilled water (RODI). Aliquots of the stock solutions were
mixed and diluted with RODI if necessary to obtain the target buffer
concentrations
2.1.2. Reagents
The mAb A was expressed in Chinese hamster ovary (CHO)
cells made by AstraZeneca, Gaithersburg, MD. MAb A has an isoelectric point range of pH 8.5–9.0 and an approximate molecular
weight of 147 kDa. The expressed mAb A was purified by Protein A chromatography, specifically MabSelect SuRe obtained from
GE Healthcare (GE Healthcare, Piscataway, NJ). Where applicable,
the mAb A aliquots were buffer exchanged and concentrated into
the appropriate buffer solutions using tangential flow filtration
(TFF) regenerated cellulose membranes (Millipore, Billerica, MA).
Concentrated stock excipients were spiked into mAb A sample as

required. High protein concentration mAb A preparations were
achieved using a pressurized concentrator cell device (Millipore)
and ultrafiltration disc membranes (Millipore).
2.2. Analytical methods
2.2.1. Protein concentration
Protein concentration was measured using a Nanodrop spectrophotometer from Thermo Fisher Scientific (Waltham, MA). The
absorbance of the solution was measured at 280 nm and the protein concentration was calculated from the extinction coefficient
using the Beer-Lambert equation.
2.2.2. Dynamic light scattering (DLS)
The hydrodynamic radius for the IgG2 monoclonal antibody was
analyzed at multiple protein concentrations using a high throughput 384-well plate DynaPro DLS instrument from Wyatt technology
(Santa Barbara, CA) equipped with a 633 nm laser. The scattered
light was monitored at 173◦ to the incident light beam and autocorrelation functions were generated using a digital autocorrelator.
The hydrodynamic radius was calculated using the Stokes-Einstein
equation.
2.2.3. Size exclusion HPLC
Analytical high performance size exclusion chromatography
(HPSEC) was performed using a TSKgel G3000SWxl column from
Tosoh Bioscience (Tosoh Bioscience, Montgomeryville, PA) with an
Agilent HPLC 1200system from Agilent Technologies (Santa Clara,
CA, USA). Test samples were diluted to a desired concentration on
ice after storage at 2–8 ◦ C to minimize RSA dissociation in the
protein sample prior to column chromatography. The HPLC autosampler was maintained at 5 ◦ C, and the column was run at ambient
temperature. Protein sample was isocratically eluted using 0.1 M
sodium sulfate, 0.1 M sodium phosphate buffer, pH 6.8, and elution
was monitored by UV absorbance at 280 nm. RSA was measured as
a percent of the leading shoulder of the main peak relative to the
total peak area (Fig. 2C).
2.3. Statistical design of experiments
The experimental design,was built with JMP statistical software

version 10.0.0 (SAS Institute Inc., Cary, NC, USA). A concentrated

stock of mAb A was adjusted to target pH and excipient concentrations through addition of stock buffer/excipient solutions. The
time of addition was defined as T = 0 s. Unless specified otherwise,
the incubation time was ≥1 h prior to sample analysis to ensure
completion of the aggregation or RSA reaction (Data not shown).
JMP software was used for statistical analysis of the experimental
data and model determination.
Screening experiments were used to identify factors that have
an impact on stability indicating measurements including aggregate and RSA. Factors tested were either ordinal (salt type, buffer
type) or continuous (pH, protein concentration, salt concentration,
excipient concentration). Continuous factors were tested at three
levels (− 0 +) with ranges and center point values chosen based on
anticipated process conditions or prior stability information from
literature or in-house data. The screening design incorporated runs
at combinations of high (+) and low (−) values for the continuous
factors at each ordinal factor plus center point runs at each ordinal
factor.
Response surface experiments were performed to measure the
degree of impact the factors have on the stability attributes and
identify interactions between continuous factors. An on-face central composite design incorporated runs at combinations of high
(+), center point (0) and low (-) values for each factor.
2.4. Bench scale chromatography experiments
2.4.1. Chromatography equipment and materials
Laboratory scale chromatography experiments were performed
using a GE Healthcare ÄKTA Explorer 100 using Unicorn software version 5.2 (GE Healthcare, Piscataway, NJ, USA). The cation
exchange resin used in this study was Capto SP ImpRes obtained
from GE Healthcare (GE Healthcare, Piscataway, NJ, USA). Cation
exchange chromatography (CEX) resins were packed into 1.15 cm
inner diameter (ID) Vantage columns (Millipore, Billerica, MA, USA)

to a bed height of approximately 12 cm and operated at a linear
velocity of 150 cm/hour.
2.4.2. Cation exchange chromatography
The CEX columns were equilibrated with 3 column volumes
(CVs) of 50 mM sodium acetate pH 4.5. The pH and conductivity of
the column effluent were monitored through built-in ÄKTA probes
to confirm column equilibration. The column was loaded with mAb
material to a residence time of 5 min. After loading, the column was
washed with 3 column volumes (CV) of 50 mM sodium acetate, pH
4.5. The protein bound on the column was eluted over a 20 CV linear salt gradient from 0 to 500 mM in 50 mM sodium acetate pH
4.5. The elution peak was fractionated in half-CV fractions based
on A280 collection criteria of 100 mAU. The absorbance of the protein was monitored at A280 by built-in ÄKTA probes. In-line pH and
conductivity were also monitored for all test runs. All runs were
carried out at room temperature (20–25 ◦ C). Cleaning in-place (CIP)
was conducted using 3 CV 50 mM sodium acetate, 1 M sodium chloride, pH 5.0 followed by 1 N sodium hydroxide. The columns were
stored in 20% ethanol after each run. Final CEX test conditions were
demonstrated using a 20 mM CaCl2 step elution.
3. Results and discussion
3.1. Observations of atypical solution behaviors during
purification of mAb A using a platform purification process
Monoclonal antibody A (mAb A) was purified using a standard
platform process (Fig. 1). While other monoclonal antibodies can
be purified under these conditions with acceptable process performance and product quality, the process standards for performance
and quality could not be achieved with mAb A (Table 1).


A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

103

Fig. 2. Analytical methods employed to measure the purification process design space. A) DLS data demonstrating the changes in hydrodynamic radius of a 10 mg/mL mAb
A sample at 25 ◦ C (monomer) and at 5 ◦ C (multimer). B) HPSEC data used to measure irreversible aggregate levels. C) HPSEC data demonstrating RSA when using SEC method
with refrigerated autosampler and higher protein concentration load.

As shown in Table 1, aggregate formation is a significant source
of decreased product purity during several platform steps including
Protein A Chromatography, Low pH inactivation, Cation exchange
chromatography and Ultrafiltration and Diafiltration. For example, observations of mAb A during low pH activation demonstrated

aggregation rates up to 0.9% aggregate/hour (Supplemental Fig. 1).
This rate of aggregation is 9 times higher than our action limits for
aggregation rate (0.1% aggregate/hour).
High protein concentration conditions are also typical during
bind and elute chromatography steps and ultrafiltration steps of the


104

A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

Fig. 3. Design space for mAb A purification process [1] Protein A [2] Low pH [3] Anion exchange [4] Cation exchange [5] Virus Filtration [6a] UF1 [6b] UF2 indicating where
aggregation and RSA occur. A) Platform conditions, B) Platform condition with sucrose, C) Platform condition with calcium chloride.

platform process (steps 1, 4 and 6 in Table 1). While conventional
mAbs exhibited acceptable stability within these protein concentrations, Luo et al. demonstrated that peak splitting observed
during mAb A cation exchange chromatography purification was
due to formation of additional aggregates during elution from the
column. The generation of aggregates results in lower product
purity and consequential yield decrease during aggregate removal
steps such as the cation exchange chromatography step. Aggregation has additional impacts on process performance such as

reduced flux during viral filtration and ultrafiltration (steps 5 and
6 in Table 1).
Early observations suggested that aggregate generation and
advanced aggregation rates might be the cause of the performance issues (i.e. peak tailing and splitting) observed during
cation exchange chromatography (step 4 of Table 1). However,
the observations of peak tailing and splitting still occurred even
when the feed aggregate levels were greatly reduced. Further
investigation revealed that mAb A exhibited on-column reversible
self-association (RSA) as well as irreversible aggregate formation
[10]. RSA and aggregate levels can be measured using DLS and
HPSEC respectively (Fig. 2). Refrigeration of the HPLC auto-sampler
enabled the use of HPSEC to visualize both RSA and aggregate
levels (Fig. 2). RSA formation was demonstrated to be favored
at both refrigerated temperatures and high protein concentrations.
It was hypothesized that mAb A non-conformance to the platform process is due to susceptibility to RSA and aggregation.
Development of a mAb A process therefore required identification of solution conditions where acceptable stability can be
achieved.

3.2. Determination of the downstream operating space
A design of experiment (DoE) approach was used to screen
several factors including pH, buffer type, salt type and salt concentration to determine the impact, if any, on mAb A stability as
measured by RSA and aggregate formation. After identifying factors that impact stability, a follow-up response surface experiment
using these factors generated a predictive model to determine the
operating space that minimizes RSA and aggregate formation.
This operating space was defined at a mAb A protein concentration of 10 mg/mL. While protein concentration is a factor that
was shown to cause aggregation and RSA with increasing concentration (Data not shown), a 10 mg/mL concentration is within the
expected range for most of the platform steps (steps 1–5 in Table 1)
and therefore statistical models were created at this control protein
concentration.
Fig. 3a plots where RSA and aggregation is predicted to occur at

combinations of solution pH and NaCl concentration which were
determined to be significant from the screening study. To illustrate a potential purification design space, limits of less than 40%
RSA and aggregate were assigned to identify a pH and NaCl concentration range where the non-ideal solution behaviors would be
minimized. RSA and aggregate formation occur under distinct conditions with RSA occurring at neutral to high pH in a salt-dependent
manner and with aggregates forming at low pH conditions also in
a salt-dependent manner. The plot shows a narrow pH range, from
approximately pH 5–6, and narrows with increasing salt concentration where RSA and aggregation are avoided. Consequently, only
one (step 6a in Table 1) out of the six platform process steps is
suitable for mAb A purification.


A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

Fig. 4. Impact of excipients on a) aggregation at pH 3.5 and b) RSA at pH 5.5 on a
50 mg/mL mAb A sample.

The results in Fig. 3a show that either the design space needs
to be widened by the addition of stabilizing excipients or that the
purification process would need to be modified to shift the majority
of process steps into an operating space that minimizes RSA and
aggregation.
3.3. Excipients can modulate the purification design space
Stability of proteins may be enhanced in the presence of excipients through conformational or colloidal mechanisms [12]. Sugars
such as trehalose and sucrose have been shown to prevent attractive electrostatic interactions between monomeric molecules [13].
A patent was filed demonstrating that mAb A RSA was mitigated
by the addition of sucrose [14], [15]. Aggregation mediated through
hydrophobic interactions can be mitigated in the presence of arginine and propylene glycol [16,17]. Urea can stabilize proteins
through hydrogen bonding [18]. Many of these excipients are currently used in platform formulations based on demonstrated drug
substance safety, efficacy and stability.
A screening study to determine the effect of four excipients,

sucrose, propylene glycol (PG), urea and arginine, upon RSA and
aggregation was performed and the results shown in Fig. 4. Based
on data from the response surface design model identifying pH,
NaCl concentration and protein concentration as factors impacting
aggregation and RSA, two test conditions were defined (50 mg/mL
protein concentration pH 3.5 or pH 5.5 in the presence of 0.1 M
NaCl). Excipient concentrations tested were demonstrated to be
effective with other monoclonal antibodies under development
(Data not shown). Mab A is susceptible to aggregation and RSA at
pH 3.5 and pH 5.5 respectively. The aggregate and reversibly associated species were measured and the percent change in each species

105

relative to the negative control (mAb A in the absence of excipient)
calculated. Results from this study are summarized in Fig. 4. Significant mAb A aggregation occurs at pH 3.5 but the presence of
sucrose mitigates the increase by 8%. The other excipients tested
appear to exacerbate aggregation. RSA occurring at pH 5.5 is mitigated by sucrose, propylene glycol and arginine but is most effective
with sucrose. Urea is shown to be ineffective in stabilizing mAb A at
either of the risk conditions for aggregation and RSA. Sucrose was
the only excipient tested that mitigated both aggregation and RSA
formation.
The data in Fig. 4 suggests that sucrose has the potential for
widening the mAb A purification design space. A response surface
design experiment was performed with the inclusion of sucrose as a
factor in addition to pH and NaCl concentration. Applying the same
criteria for aggregation and RSA as the model described in Fig. 3a,
b demonstrates that 0.3 M sucrose broadens the design space. The
result is that four of the six platform steps (steps 1, 4, 5 and 6 in
Table 1) occur in the broader safe operating space.
While addition of sucrose to all process steps can mitigate the

non-ideal solution behaviors observed with mAb A, there are process challenges and risks associated with the use of sugars as
a manufacturing raw material. These challenges include: higher
viscosity liquids [19], a carbon source for adventitious biologic
contamination, and a higher risk for unacceptable endotoxin levels in the drug substance. Sucrose addition may be appropriate
for unit operations like low pH inactivation (step 2 in Table 1)
where exposure time to sucrose would be limited and risk of microbial contamination low at this pH condition. Sucrose addition to
the Protein A chromatography (step 1 in Table 1) low pH elution buffer may also prevent aggregate formation under platform
conditions. However, filtration steps (steps 5 and 6 in Table 1)
with pressure considerations or polishing chromatography steps
(steps 3 and 4 in Table 1) with potentially longer product exposure time to sucrose are at greater risk to the aforementioned
challenges.
3.4. Changing salts used in the cation exchange chromatography
step can modulate the design space location
The response surface design shows that RSA and aggregation
increases with increasing sodium chloride concentration. The salt
type was not a significant factor as different salts (Ex. calcium chloride) had a similar impact on RSA and aggregation (Supplemental
Fig. 2) at equivalent molar concentrations. While this data suggests that selection of salt type cannot modulate the design space
with respect to RSA and aggregation at equivalent molarities, salt
type can impact elution conditions during cation exchange chromatography [20]. Cation exchange elution salt concentration is
dependent upon the strength of the cationic charge. Because the
binding between the mAb and the resin ligand is electrostatic, the
required concentration of salt to elute the mAb varies based on the
respective charge of the cation.
A study was performed to evaluate the efficacy of different salt
types in resolving mAb A monomeric product from high molecular weight (HMW) impurity species and to determine the required
salt concentration to achieve resolution. Mab A was loaded onto an
equilibrated cation exchange column, washed and eluted with salt
as previously described (Section 2.4.2). This chromatographic procedure was used to screen different salt types, including potassium
chloride, sodium sulfate, sodium acetate, sodium citrate, arginine
hydrochloride, magnesium chloride and calcium chloride, in addition to the platform sodium chloride. The cation exchange outputs

of step yield, product volume and product quality were measured
for each run.
The elution chromatograms for each salt type are shown in Fig. 5.
The mAb A monomer (peak 1) is eluted first during the salt gra-


106

A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

Fig. 5. Impact of Salt Species on Separation of Monomer and High Molecular Weight (HMW) Impurities during Cation Exchange Chromatography. Monomer is the early
eluting peak while the HMW impurities, annotated with the percent HMW, are late eluting.

Table 2
Summary of process modifications for mAb A purification and applicable HTPD tools for future non-platform mAb case studies.
Step

mAb A Observations

New mAb A
Observations

Modified Conditions

Potential HTPD methods

1
Protein A
Chromatography
2

Low pH Inactivation
3
Anion Exchange
Chromatography

High aggregate in
product

Reduced aggregate in
product

96-well stability study,
DLS plate reader,
HTSEC

Aggregate formation

No aggregate formation

Low pH
Low salt,
High conc.
Sucrose

Peak tailing

No Peak tailing

TeChrom Robo-column
purification

4
Cation Exchange
Chromatography
5.
Virus Filtration
6A.
Ultrafiltration 1

Peak splitting,
aggregate formation,
low yield
Low
flux
High feed
pressure/Low flux

Peak splitting, No
aggregate formation,
Increased yield
Increased
Throughput
Reduced Feed pressure/
Increased flux

Neutral pH
Low salt
High conc.
Sucrose
Intermediate pH

Low salt

6B
Ultrafiltration 2

dient followed by the HMW species (peak 2). This is due to the
HMW species having a greater affinity to the cation exchange ligand. The yield percentage of each peak can be calculated from the
relative peak areas. The yield of the HMW species is annotated on
each chromatogram in Fig. 5. The pool of each peak was collected
and analyzed for product quality (Data not shown). Comparing
each of the chromatographic runs with respect to monomer peak
purity and step yield, calcium chloride elution achieved both high
monomer purity (98%) and high yield (82%). The run with calcium
chloride also shows a low level (13.7%) of the HMW species and is an
indication that this elution salt condition minimizes the formation
of HMW species. This observation is consistent with the response
surface design model that shows RSA and aggregation occurring at
high salt concentration. Peak elution with calcium chloride occurs
at approximately 25 mM while other salts (Ex. sodium acetate)

Intermediate pH
Low salt
High conc.
Intermediate pH
Low Salt
Arginine
Very high conc.

Filterplate throughput
experiments

DLS flowthrough
cuvette, High
throughput tangential
flow filtration

that require a high salt concentration (approximately 150 mM) for
elution show a higher yield (36.4%) of the HMW species. While
other salt types can achieve some combination of (1) high step
yield, (2) high product purity or (3) minimal HMW species formation, calcium chloride achieves all three desired objectives. Further
confirmation of calcium chloride efficacy for mAb A purification
was shown by a subsequent cation exchange chromatography run
using a calcium chloride 20 mM isocratic step elution where high
yield and high product quality were achieved (Supplemental Fig. 3).
Additionally, the product volume with the calcium chloride elution
was 25% less compared to the platform sodium chloride condition
allowing for manufacturing facility fit benefits. Fig. 3c illustrates
how the use of calcium chloride shifts the design space location for
the cation exchange step (step 4 in Table 1) into a safe operating
area to avoid aggregation.


A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

Fig. 6. Optimized operating space for mAb A purification process [1] Protein A [2]
Low pH [3] Anion exchange [4] Cation exchange [5] Virus Filtration [6a] UF1 [6b]
UF2 indicating where aggregation and RSA is predicted to occur. Proposed optimized
mAb A process using sucrose during Protein A chromatography and calcium chloride
during cation exchange chromatography.

4. Conclusion

In this paper, we describe a case study where atypical solution
behaviors, determined to be reversible self-association (RSA) and
aggregation, were observed during platform processing of mAb A
resulting in unacceptable product quality and yield for a platform
commercial manufacturing process. An intensive study was performed to better understand these non-ideal solution behaviors
and to develop a downstream process for this non-platform mAb.
Through the use of screening and response surface design of experiments, factors impacting the non-ideal solution behaviors were
identified and statistical models to predict RSA and aggregation levels within ranges of these factors were generated. It was shown that
pH, salt concentration, protein concentration and sucrose are factors that determine where non-ideal solution behaviors occur. With
this knowledge, an optimized process for mAb A was proposed.
Because of the mAb A stabilizing properties of sucrose, inclusion
of sucrose to the elution buffer of the Protein A chromatography
step (step 1 in Table 2) and to the acidification titrant of the low pH
inactivation step (step 2 in Table 2) is expected to reduce aggregate
formation currently observed at platform conditions. Manufacturing concerns about the risk of adventitious biologic contamination
associated with high sugar concentrations are mitigated by the
subsequent process steps where sucrose would be diluted during neutralization after low pH treatment and flow-through anion
exchange chromatography (step 3 in Table 2) and removed during
cation exchange chromatography (step 4 in Table 2). Knowledge
of the salt concentration effect on RSA and aggregation drove further experiments to screen salt types that can elute mAb A from
the cation exchange column at a low molar concentration. The use
of 20 mM calcium chloride eluted mAb A product of high product quality and yield without aggregate induced peak splitting
observed with the platform sodium chloride elution salt. Consequential lower salt concentration in the mAb A product have
additional process benefits where low salt concentrations may
improve flux and reduce feed pressure during viral filtration and
ultrafiltration (steps 5 and 6 in Table 2) [21]. Fig. 6 illustrates how
leveraging the combination of excipient and salt type results in
a manufacturing process that is only modestly different from the

107

platform but now allows for four (steps 1, 4, 5 and 6 in Table 2)
out of the six platform process steps to be within a stable operating
space for mAb A purification. While low pH inactivation (step 2 in
Table 2) and anion exchange chromatography (step 3 in Table 2)
still occur within the aggregation and RSA risk space respectively,
the high molecular weight species formed at these steps are within
the impurity removal capabilities of the subsequent downstream
steps.
This strategy for optimizing a process for a non-platform mAb
was a reactive approach to purification problems observed during conventional process development. While successful, delayed
timelines and significant additional investment in lab scale development time were an unintended consequence. Most of the
development time was attributed to performing the bench scale
experiments (section 2.4) and analytical methods (section 2.3).
For future mAb and non-mAb drug products, the use of the
design of experiment approach described here could be used in a
proactive manner where candidate molecules are initially screened
to assess solution properties. This approach will need to (1) minimize the amount of protein needed for these studies and (2)
generate data quickly to enable candidate drug selection within
the existing development timelines. Both objectives can be readily
accomplished by leveraging existing high throughput scale-down
models and analytics, collectively referred to as high-throughput
process development (HTPD). Our lab has implemented several
HTPD methods that greatly reduce experiment time [22]. Some of
these techniques and their applicability to process steps in this case
study are shown in Table 2. This data can then determine suitability
for further process development and applicability to our platform
process thus reducing costly development time. The model for our
development lab of the future will link these HTPD tools together
to enable the proactive screening of all biopharmaceuticals that fit

within the platform paradigm to assess the molecule’s process fit
and enable better planning for development resources and timeline.
Acknowledgements
The authors would like to acknowledge Matthew Dickson, Yuling Li, Timothy Pabst, Irina Ramos, David Robbins and Min Zhu (all
currently or formerly at AstraZeneca).
Appendix A. Supplementary data
Supplementary material related to this article can be found, in
the online version, at doi: />03.021.
References
[1] G. Walsh, Second-generation biopharmaceuticals, Eur. J. Pharm. Biopharm. 58
(2004) 185–196.
[2] A.A. Shukla, J. Thommes, Recent advances in large-scale production of
monoclonal antibodies and related proteins, Trends Biotechnol. 28 (2010)
253–261.
[3] W. Wang, Protein aggregation and its inhibition in biopharmaceutics, Int. J.
Pharm. 289 (2005) 1–30.
[4] W. Wang, S. Singh, D.L. Zeng, K. King, S. Nema, Antibody structure, instability,
and formulation, J. Pharm. Sci. 96 (2007) 1–26.
[5] S.J. Shire, Z. Shahrokh, J. Liu, Challenges in the development of high protein
concentration formulations, J. Pharm. Sci. 93 (2004) 1390–1402.
[6] E.Y. Chi, S. Krishnan, T.W. Randolph, J.F. Carpenter, Physical stability of
proteins in aqueous solution: mechanism and driving forces in nonnative
protein aggregation, Pharm. Res. 20 (2003) 1325–1336.
[7] W. Wang, C.J. Roberts, Protein aggregation – mechanisms, detection, and
control, Int. J. Pharm. 550 (2018) 251–268.
[8] R. Esfandiary, A. Parupudi, J. Casas-Finet, D. Gadre, H. Sathish, Mechanism of
reversible self-association of a monoclonal antibody: role of electrostatic and
hydrophobic interactions, J. Pharm. Sci. 104 (2015) 577–586.

108

A. Man et al. / J. Chromatogr. A 1597 (2019) 100–108

[9] D. Some, J. Pollastrini, S. Cao, Characterizing reversible protein association at
moderately high concentration via composition-gradient static light
scattering, J. Pharm. Sci. 105 (2016) 2310–2318.
[10] H. Luo, N. Macapagal, K. Newell, A. Man, A. Parupudi, Y. Li, Y. Li, Effects of
salt-induced reversible self-association on the elution behavior of a
monoclonal antibody in cation exchange chromatography, J. Chromatogr. A
1362 (2014) 186–193.
[11] P.S. Sarangapani, J. Weaver, A. Parupudi, T.M.D. Besong, G.G. Adams, S.E.
Harding, P. Manikwar, M.M. Castellanos, S.M. Bishop, J.A. Pathak, Both
reversible self-association and structural changes underpin molecular
viscoelasticity of mAb solutions, J. Pharm. Sci. 105 (2016) 3496–3506.
[12] D.S. Goldberg, S.M. Bishop, A.U. Shah, H.A. Sathish, Formulation development
of therapeutic monoclonal antibodies using high-throughput fluorescence
and static light scattering techniques: role of conformational and colloidal
stability, J. Pharm. Sci. 100 (2011) 1306–1315.
[13] J.K. Kaushik, R. Bhat, Why is trehalose an exceptional protein stabilizer? An
analysis of the thermal stability of proteins in the presence of the compatible
osmolyte trehalose, J. Biol. Chem. 278 (2003) 26458–26465.
[14] M. Dimitrova, N. Mody, in: Antibody Formulations, U.S.P. Office, USA, 2012
(Ed.).
[15] G.V. Barnett, V.I. Razinkov, B.A. Kerwin, S. Blake, W. Qi, R.A. Curtis, C.J. Roberts,
Osmolyte effects on monoclonal antibody stability and
concentration-dependent protein interactions with water and common
osmolytes, J. Phys. Chem. B 120 (2016) 3318–3330.

[16] K. Tsumoto, D. Ejima, Y. Kita, T. Arakawa, Review: why is arginine effective in

suppressing aggregation? Protein Pept. Lett. 12 (2005) 613–619.
[17] A. Apicella, P. Heunemann, S. Bolisetty, M. Marascio, A. Gemperli Graf, L.
Garamszegi, R. Mezzenga, P. Fischer, C.J. Plummer, J.A. Manson, The influence
of arginine on the response of enamel matrix derivative (EMD) proteins to
thermal stress: towards improving the stability of EMD-based products, PLoS
One 10 (2015) e0144641.
[18] Q. Zou, S.M. Habermann-Rottinghaus, K.P. Murphy, Urea effects on protein
stability: hydrogen bonding and the hydrophobic effect, Proteins 31 (1998)
107–115.
[19] P. Longinotti, H. Corti, Viscosity of concentrated sucrose and trehalose
aqueous solutions including the supercooled regime, J. Phys. Chem. Ref. Data
37 (2008).
[20] W.R. Melander, Z. el Rassi, C. Horvath, Interplay of hydrophobic and
electrostatic interactions in biopolymer chromatography. Effect of salts on the
retention of proteins, J. Chromatogr. 469 (1989) 3–27.
[21] J. Luo, Y. Wan, Effects of pH and salt on nanofiltration—a critical review, J.
Membr. Sci. 438 (2013) 18–28.
[22] S.T. Evans, K.D. Stewart, C. Afdahl, R. Patel, K.J. Newell, Optimization of a
micro-scale, high throughput process development tool and the
demonstration of comparable process performance and product quality with
biopharmaceutical manufacturing processes, J. Chromatogr. A 1506 (2017)
73–81.