Journal of Biotechnology 242 (2017) 11–18

Contents lists available at ScienceDirect

Journal of Biotechnology
journal homepage: www.elsevier.com/locate/jbiotec

Transfer of a three step mAb chromatography process from batch to

continuous: Optimizing productivity to minimize consumable
requirements
Xhorxhi Gjoka, Rene Gantier, Mark Schofield ∗
Pall Life Sciences, 20 Walkup Dr., Westborough, MA, USA

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

Article history: The goal of this study was to adapt a batch mAb purification chromatography platform for continu-
Received 29 August 2016 ous operation. The experiments and rationale used to convert from batch to continuous operation are
Received in revised form 2 December 2016 described. Experimental data was used to design chromatography methods for continuous operation that
Accepted 2 December 2016
would exceed the threshold for critical quality attributes and minimize the consumables required as com-
Available online 6 December 2016
pared to batch mode of operation. Four unit operations comprising of Protein A capture, viral inactivation,
flow-through anion exchange (AEX), and mixed-mode cation exchange chromatography (MMCEX) were
Keywords:
integrated across two Cadence BioSMB PD multi-column chromatography systems in order to process
Continuous chromatography
Monoclonal antibody (mAb)
a 25 L volume of harvested cell culture fluid (HCCF) in less than 12 h. Transfer from batch to continu-
Process intensification ous resulted in an increase in productivity of the Protein A step from 13 to 50 g/L/h and of the MMCEX
Productivity step from 10 to 60 g/L/h with no impact on the purification process performance in term of contaminant
Cadence BioSMB PD removal (4.5 log reduction of host cell proteins, 50% reduction in soluble product aggregates) and overall
chromatography process yield of recovery (75%). The increase in productivity, combined with continuous
operation, reduced the resin volume required for Protein A and MMCEX chromatography by more than
95% compared to batch. The volume of AEX membrane required for flow through operation was reduced
by 74%. Moreover, the continuous process required 44% less buffer than an equivalent batch process. This
significant reduction in consumables enables cost-effective, disposable, single-use manufacturing.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction
Over 50 mAbs have been approved by the FDA and another 50
are currently in phase three clinical trials (Reichert, 2016). The
tremendous success of this class of drugs does come at a price, with
mAb sales recently estimated to be close to $75 billion (Ecker et al.,
2015). However, manufacturing costs are estimated to only be a
small fraction of the sales price (Kelley, 2009). This has tradition-
ally led pharmaceutical industries to focus on time to market and
adopt a conservative approach to manufacturing, (Castilho, 2014;
Gottschalk, 2013; Warikoo et al., 2012). With continuous process-
ing only being employed to reduce the overall purification time
for protein therapies whose stability throughout the purification
process is limiting (Vogel et al., 2012).
However, the biopharmaceutical landscape is changing rapidly.
Growing competition from biosimilars and multiple therapies tar-
∗ Corresponding author.
E-mail address: Mark Schofield@pall.com (M. Schofield).
geting the same disease have renewed focus upon manufacturing
cost (Konstantinov and Cooney, 2015). Many current mAb or Fc
fusion protein therapies on the market are facing patent expiration
over the next few years (Strohl et al., 2012), enabling manufactur-
ers of biosimilars greater access to market share. Along with cost,
flexibility in manufacturing is also critical as the dynamic landscape
makes it difficult to predict the quantities of drug required through-
out product lifetime and it takes many years to build a new mAb
manufacturing facility (Kamarck, 2006).
To address these combined challenges of cost and flexibility,
many biopharmaceutical companies are pursuing process inten-
sification via continuous manufacturing (Horowitz, 2010). This
endeavor has been actively encouraged by the FDA, which iden-
tifies continuous manufacturing as an opportunity to increase the
flexibility, agility and robustness of production (Lee et al., 2015).
Process intensification through continuous manufacturing
enables the elimination of hold steps which are not value added and
further simplifies the process (Warikoo et al., 2012). It also reduces
process foot print using smaller and single use equipment which is

http://dx.doi.org/10.1016/j.jbiotec.2016.12.005
0168-1656/© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
12 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18
fundamental to flexible manufacturing and the accommodation of
dynamic market demand (Fromison, 2009; Hodge, 2004).
A number of companies are now highlighting their continuous
purification strategies (Brower et al., 2014; Godawat et al., 2012;
Hernandez, 2015; Kaltenbrunner et al., 2016; Mahajan et al, 2012;
Pollock et al., 2013; Palmer, 2015; Vogel et al., 2012), but what is
largely absent from the literature is how to transfer a batch pro-
cess to continuous operation (Girard et al., 2015). Here we take
a batch mAb purification and describe the experiments and ratio-
nale required to transfer it to a continuous chromatography process
that can be operated with two Cadence BioSMB PD systems. We
show it is possible to achieve higher loading capacities with shorter
residence times by loading two or more columns in series. This
increase in capacity results from overloading the first column and
capturing unbound product in the flow through effluent on sub-
sequent columns. As buffer consumption is inversely proportional
to capacity, the increased capacity offered by multi column chro-
matography results in significantly reduced buffer consumption.
Along with increased capacity, residence times are reduced and so
is the cycle time. Increased capacity and decreased cycle both lead
to higher process step productivity expressed in gram of product
produced per liter of chromatography media and per hour (g/L/h).
This can considerably reduce the resin volume required for a purifi-
cation process. Furthermore, by leveraging the increased capacity
that is offered by continuous chromatography, the loading condi-
tions going from batch to continuous can be modified, increasing
the robustness of purification.
2. Materials and methods
2.1. Analytical methods
Host cell protein (HCP) was quantified using 3G CHO ELISA
(Cygnus Technologies, Southport, NC) assay. Soluble aggregates
were measured by HPLC-SEC using a TSK-GEL, Super SW3000,
4.6 mm × 30 cm, 4 ␮m column (Tosoh Haas, King of Prussia, PA).
Concentration of mAb in HCCF was measured using Protein A
biosensors (ForteBio, Menlo Park, CA) and an Octet Red 96 BLI sys-
tem. Concentration of mAb in post Protein A samples was measured
using a NanoDrop 8000 spectrophotometer (Thermo Scientific,
Waltham, MA).
2.2. Impact of operating parameters on purity of protein A elution
A five factor, two level DoE was designed using Minitab to deter-
mine the impact of overloading (loading the columns to saturation)
on% recovery of mAb, % mAb aggregate, elution volume and host cell
protein (HCP) concentration. The factors tested included residence
time (0.6 min vs. 4.5 min), mass loaded (10% DBC- dynamic binding
capacity- vs. 90% DBC), load concentration (2 mg/mL vs. 5 mg/mL),
and wash volume (10 column volumes (CV) versus 30 CV).
2.3. Capacity determination for multi-column bind/elute
processes
HCCF containing mAb at titer 2, 6, and 10 mg/mL was loaded
at residence times of 0.5, 1,1.5, 2, 3, and 4.5 min onto KANEKA
KanCapA 1 mL Protein A PRC columns (Pall Life Sciences, Port Wash-
ington, NY) using an AKTA Avant 25 (GE Healthcare, Uppsala, SWE).
The breakthrough curves generated from these experiments were
used to calculate the 10% DBC. Operating binding capacity (OBC) for
continuous mode was calculated by the method described in Gjoka
et al. (2015).
Protein A purified mAb (8 mg/mL concentration) titrated to pH
8.2 and 8 mS/cm was loaded at residence times of 0.75, 1.5, 2.25, 3,
4.5, and 6 min onto CMM HyperCel MMCEX 1 mL PRC columns (Pall
Life Sciences, Port Washington, NY). Tests were repeated using four
different load conditions, pH 8 & 6 mS/cm, pH 8 & 10 mS/cm, pH 8.4
& 6 mS/cm, and pH 8.4 & 10 mS/cm at residence times of 1.5 and
3 min. The breakthrough curves were used to measure the 10% DBC
at each residence time. Operating binding capacity was calculated
using the method described in Gjoka et al. (2015).
2.4. Screening conditions for operation of Mustang Q membrane
and CMM HyperCel sorbent
The impact of pH and conductivity on binding and separation of
the target mAb versus host cell protein and aggregate contaminants
was determined using a DoE based approach with response sur-
face analysis via Minitab. The AEX membrane, Mustang Q (Pall Life
Sciences, Port Washington, NY) and MMCEX sorbent, CMM Hyper-
Cel (Pall Life Sciences, Port Washington, NY), were evaluated for
binding over a broad range of conductivity and pH. CMM Hyper-
Cel sorbent was transferred into an AcroprepTM Advance 96 Filter
Plate (PN 8129, Pall Life Sciences, Port Washington, NY). A post
Protein A mAb sample at 1 mg/mL in the appropriate buffer was
added to each well (Pezzini et al., 2011; Toueille et al., 2011). Fil-
ter plates were incubated at room temperature while shaking for
2 h. Flow through was collected and analyzed for recovery of mAb
and level of contaminants. For Mustang Q membrane screening of
flow through operating conditions, AcroPrepTM Advance 96-Well
Filter Plates with Mustang Q membrane (PN 8171, Pall Life Sciences,
Port Washington, NY) were washed with binding buffer and loaded
directly with 1 mg/mL post Protein A mAb samples. Flow through
fraction was analyzed for HCP concentration. The best perform-
ing conditions from high throughput screening were verified in
dynamic mode on 1 mL PRC pre-packed columns (Pall Life Sciences,
Port Washington, NY).
2.5. Cadence BioSMB PD chromatographic system
Two Cadence BioSMB PD systems were employed to oper-
ate the multi-column chromatography experiments performed in
this study. Each system is capable of operating up to 16 columns
simultaneously through software that enables the operation of sim-
ulated moving bed chromatography. The core technology of the
machine is the disposable valve cassette which contains 240 valves
and features eight inlets, and six outlets. Valves are activated by
air pressure applied by solenoid actuators onto a fluoroelastomer
sheet. When air pressure is applied the valves are closed, when
the pressure is released pressure in the liquid flow-path forces the
valves open. The system is equipped with seven pumps capable of
operating from 1 to 200 mL/min. The standard valve cassette flow
path is 1 mm in diameter and can accommodate flow rates up to
70 mL/min. A larger 3 mm valve cassette is available for flow rates
above 70 mL/min.
The system has four UV sensors with wavelengths in
200–850 nm range, four conductivity sensors with a range of
1 ␮S/cm to 200 mS/cm, and two pH probes with a range from 2
to 12.
3. Results and discussion
Streamlining of the purification process began by assessing the
order of unit operations. The batch process had previously been
operated with the sequence of operations, Protein A, MMCEX in
bind and elute mode to AEX in flow through mode, Fig. 1. The
MMCEX sorbent is eluted via an increase in conductivity, however,
the AEX membrane demonstrates improved HCP reduction at low
conductivity. To accommodate this, the elution from the MMCEX
sorbent has to be diluted three-fold to lower the conductivity from
22 mS/cm to 7 mS/cm. The dilution requires additional buffer and
 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18
Viral
Protein A
va on
Load to 30 mg/mL at Hold for 30 minutes
4 minutes residence between pH 3.5 – 3.7
Fig. 1. Order of unit operations for a batch process.
Table 1
Effect of order of unit operations on purity.
Order of unit operations HCP log reduction % Aggregates
KanCapA – CMM HyperCel – Mustang Q 4.1 1.1
KanCapA – Mustang Q – CMM HyperCel 3.9 1.5
a system for buffer addition and mixing. By modifying the order of
unit operations to Protein A followed by AEX and finally MMCEX we
can eliminate the dilution step. When we compare the two different
sequences of operations we can see an overall similar performance,
Table 1. The small advantage in purification performance by oper-
ating MMCEX sorbent before AEX membrane is outweighed by the
advantages of eliminating the dilution step. This makes integration
of the two polishing steps easier and enables the elimination of a
system for buffer addition and mixing from the process. Operation
without a dilution step also reduces the time required to load AEX
membrane as only 1/3rd of the volume has to be applied making the
batch process more efficient and less time consuming. This process
sequence, Protein A sorbent followed by intermediate purification
using an AEX membrane and polishing with MMCEX sorbent, is our
starting point as we begin to consider transfer of the process from
batch to continuous operation.
3.1. Optimization of protein A capture step
In order to convert the process from batch to continuous oper-
ation a DoE was performed on Protein A chromatography media
(KANEKA KanCapA). The effects of overloading, wash volume, load
concentration and residence time were assessed with regards to
the purification performance (data not shown). This experiment
shows that a short wash (10 CV divided by 3 wash steps) along
with low residence time and overloading has minimal effect on the
HCP concentration post Protein A.
To investigate the impact of the multi-column chromatography
mode of operation on Protein A binding capacity, multiple single
column breakthrough experiments were employed as described in
Gjoka et al. With six single column breakthrough experiments, the
operating binding capacity in multi-column mode of operation can
be approximated over the complete design space. This data pre-
sented as a contour plot, using the distance method of interpolation,
(Fig. 2) shows a 50% or greater increase in capacity is available in
continuous multi-column mode over the range of residence times
13
Mixed Mode Anion Exchange
Ca on
Membrane
Exchange
Load to 30 mg/mL
Dilute and flow through at
Bind at pH 7.6 and 7 mS/cm
pH 7.6 and 7 mS/cm
Elute at pH 7.8 and 22 mS/cm
Mixed Mode
Anion Exchange
Ca on
Membrane
Exchange
Load to 30 mg/mL
Flow through at pH 7.6
Bind at pH 7.6 and 7 mS/cm
and 7 mS/cm
Elute at pH 7.8 and 22 mS/cm
that were tested when compared to a single column operated in
batch mode.
The design space for column loading capacity was then used to
select the optimal binding capacity for a continuous multi-column
chromatography process that can accommodate an upstream flow
rate of 12 mL/min at a mAb titer of 4 mg/mL which is imposed on the
capture step by prior unit operations. When the process was devel-
oped only 5 mL pre-packed columns with a 5 cm bed height were
available. Applying 12 mL/min onto these columns would give a lin-
ear flow rate of 720 cm/h which is in excess of the recommended
maximum linear velocity for the Protein A sorbent (500 cm/h). To
be able to accommodate this load flow rate, we split the flow across
two pairs of columns connected in series, meaning that at any given
point in time there are always four columns receiving feed. With
this configuration the columns are loaded at 360 cm/h (0.83 min
RT per column or 1.66 min total RT as two columns are loaded in
series). We can see from Fig. 2 that this gives a 35 mg/mL operat-
ing binding capacity. This compares favorably to the batch process
which was loaded using a 4 min residence time and a capacity of
27 mg/mL (calculated by taking 60% of the DBC at 10% of protein
breaking through).
To make a complete chromatography cycle, four additional
columns are required to operate the non-load steps (washes, elu-
tion, CIP and re-equilibration). Thus, the Protein A step is operated
with eight 5 mL columns (configuration of columns shown in Fig. 7)
with a similar overall performance to a predicted four column pro-
cess with 10 mL columns. This flexibility is a feature that is unique
to the Cadence BioSMB PD valve block.
3.2. Low pH viral inactivation
In order to perform post Protein A low pH viral inactivation
semi-continuously, a sub-batch cycle to cycle strategy is employed.
Eight Protein A elution fractions (the fractions from a complete
cycle) are pooled into a first surge tank over the course of one
Cadence BioSMB PD Protein A cycle. At the end of the Protein A
cycle, this pool is transferred to a second stirred surge tank where
acidification to pH 3.6 +/− 0.1, 30 min hold, and a neutralization
step with 0.5 M Tris is performed to bring it to pH 8.2 & 6 mS/cm.
The neutralized viral inactivated material is then transferred into
third surge tank before the end of the second Cadence BioSMB PD
Protein A cycle. These steps were repeated for every Protein A cycle
using external programmable pumps to transfer product between
surge tanks cyclically for the duration of the experiment. Dosing
14 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18
Fig. 2. Binding capacity of The Protein A sorbent KANEKA KanCapA in (a) Batch mode and (b) Continuous mode with two columns loaded in series.
Table 2
Options considered for flow through operation of AEX membrane adsorber.
Strategy # 1 2 3 4
Membrane Volume (mL) 10 5 0.86 0.86
# Cartridges 1 2 2 2
Load capacity (g/mL) 2.5 2.5 1.4 2.8
Regeneration Scheme None None <1 h CIP 1 h CIP
was performed using two external pumps that are always running
at a fixed flow rate through the Cadence BioSMB PD valve block.
The valves within the cassette are used to direct acid and base to
the second surge tank or to recirculate the acid and base to their
respective containers.
3.3. Evaluation of load conditions – anion exchange membrane
adsorber operated in flow through mode
The key metrics used to assess the performance of the polish-
ing steps are HCP and aggregate reduction. Here an AEX membrane
adsorber (Mustang Q) is employed in flow through mode to lower
the HCP concentration. The ability of this membrane to remove HCP
over a broad design space was investigated using a combination of
96 well membrane plates and a 0.86 mL membrane capsule. First,
the HCP removal performance was evaluated at different conduc-
tivity and pH load conditions using a fixed load capacity of 0.5 g/mL
of membrane. Fig. 3 shows fold reduction in HCP after passing Pro-
tein A purified mAb through an AEX membrane under a range of
conditions from pH 7.2–8.4 and conductivity from 5 to 10 mS/cm.
At conductivities ranging from 5 mS/cm to 7 mS/cm and pH 8 to
pH 8.4, it is possible to achieve a greater than 60 fold reduction in
HCP. To determine the load capacity of the AEX membrane, up to
5 g of mAb was loaded onto a 0.86 mL capsule at four conditions,
pH 8 & 5 mS/cm, pH 8 & 7 mS/cm, pH 8.4 & 5 mS/cm and pH 8.4 &
7 mS/cm. These experiments (data not shown) indicate that within
this design space, loading Protein A purified mAb beyond 2.5 g/mL
of membrane results in an exponential increase in pressure across
the device. This pressure increase occurs before breakthrough of
HCP.
With this load capacity limitation, multiple strategies listed in
Table 2 that could accommodate a continuous load onto an AEX
membrane were considered. An oversized device (a 10 mL capsule)
could be employed that is large enough to polish the complete batch
without being regenerated and recycled (Table 2, strategy 1). Con-
sidering the total load volume to be processed is 25 L of HCCF with
a mAb titer of 1 mg/mL, and the AEX membrane loading capacity
defined in this study is 2.5 g/mL of membrane, this first strategy
could be used to accommodate the load. Alternatively, two AEX
5 mL capsule configured in parallel could be loaded and discarded
after use (Table 2, strategy 2). Strategies 3–5 rely on cycling and
regenerating a 0.86 mL capsule which can be operated at flow rates
up to 10 membrane volumes per minute. These options are feasi-
5 ble since the average flow rate directly post viral inactivation is less
0.86 than 5 mL/min. The number of devices required depends on two fac-
3 tors: the load capacity (2.5 g/mL) and the amount of time required
1.4 to regenerate the device (1 h). Here it would be physically possible
1 h CIP to operate with two 0.86 mL capsules and have one device being
loaded while the other device is being washed, regenerated and
re-equilibrated. Strategies 3 and 4 in Table 2 are both options that
use 2 smaller capsules for purification. Using two capsules makes
integration with the mixed-mode resin difficult because the cycle
time for the membrane adsorbers has to be synchronized with the
cycle times for the mixed-mode resin. This means the AEX mem-
brane adsorber cycle time has to be a multiple of the MMCEX cycle
time. In strategy 3, one AEX membrane adsorber cycle is exactly
twice as long as the cycle for the MMCEX process. This strategy is
feasible but requires the CIP step to be shortened to less than 1 h
which increases the possibility of performance loss over time. The
other option is to create a method with a cycle duration equal to
three times that of the mixed-mode cycle which enables us to per-
form CIP for more than an hour but requires each capsule to be
loaded to 2.8 g/mL and increases the risk of failure. Therefore, the
fifth strategy outlined in Table 2, using 3 capsules each loaded to
1.4 g/mL, was selected because it provided the best compromise in
terms of minimizing membrane volume, ensuring the capsules are
regenerated thoroughly, and avoiding operation at high pressure.
This strategy (5) embraces the concept of continuous processing,
reducing the size of consumables by performing more cycles. This
also gives a clear path for how best to operate processes over a
longer time that would be required to integrate with, for example,
a continuous upstream perfusion cell culture process.
3.4. Evaluation of load and elution conditions – CEX mixed-mode
resin operated in bind and elute mode
A MMCEX sorbent (CMM HyperCel) is employed in the process
to reduce mAb aggregates and to further reduce HCP concentration.
Originally, for the batch process a compromise was made between
capacity and HCP reduction, Fig. 4. Loading at higher pH results in
greater HCP reduction at the binding step, but this is at the expense
of capacity. The batch process was designed around column vol-
ume. The strategy was to purify the complete batch (25 L HCCF at
1 g/L mAb titer) via a single cycle with a column of 1 L in volume
or less. To accommodate this, loading conditions selected for the
batch process are pH 7.6 and 7 mS/cm conductivity. The DBC at
10% breakthrough under these conditions is 45 mg/mL. This allows
for an operating binding capacity of 27 mg/mL (60% of the DBC at
10% breakthrough).
 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18 15

Fig. 3. AEX membrane, Mustang Q, HCP reduction performance.

In continuous mode a higher capacity is expected because
columns are loaded in series. This allows the columns to be
overloaded without loss of product as product is captured on a sub-
sequent column. Therefore, the loading conditions can be modified
when transferring from a batch to continuous process. To find the
best compromise between HCP reduction and operating binding
capacity a column based DoE was performed using extended pH
conditions, Fig. 4. This DoE enables the identification a new opti-
mum design space. Using loading conditions of pH 8.1 and 6 mS/cm
conductivity a 45 mg/mL capacity at 1% breakthrough is calculated.
A 40 mg/mL operating loading capacity was used for the actual
MMCEX process to take a modest margin. These buffer conditions
used in continuous multi-column mode decrease the HCP level in
elution to 500 ppm compared to 625 ppm obtained in batch mode.
In addition, making the same modification to the loading condi-
tions onto the membrane adsorber (upstream of MMCEX) results in
increased HCP reduction from the membrane adsorber (60 fold HCP
reduction, compared to 30 fold with the batch conditions, Fig. 3).
Column elution DoEs were performed to determine the optimal elu-
tion condition from MMCEX for yield, HCP reduction and removal
of mAb aggregates, data not shown. From these experiments elu-
tion conditions of pH 7.8 and 22 mS/cm conductivity are selected to
deliver greater than 80% product recovery, >10 fold HCP reduction
and a final aggregate clearance of 50%.
With this information, along with the process constraints from
the previous process, the MMCEX process can be defined within
the context of continuous operation. To generate a balanced cycle,
six × 5 mL columns are required in total, where two columns are
loaded in series with a residence time of 1.5 min and four columns
are required to operate the non-load steps.
3.5. Implementation of the continuous integrated
chromatography platform
Two Cadence BioSMB PD systems were connected to operate the
entire chromatography train. The first Cadence BioSMB PD system
performed the capture step with eight × 5 mL Protein A columns
along with the dosing for the low pH viral inactivation step. The
second Cadence BioSMB PD operates the combined polishing steps.
AEX membrane is operated with three devices with a 120 min cycle

Fig. 4. Extension of MMCEX (CMM HyperCel) sorbent design space made possible by continuous operation.
16 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18

Process #1 Process #2
A post KanCapA post VI/depth/sterile post CMM
B post KanCapA post VI/depth/sterile post CMM

Mab concentration (mg/mL)

14 14

Mab Concentration (mg/mL)

12 12

10 10

8 8

6 6

4 4

2 2

0 0
60 180 300 420 540 660 0 120 240 360 480
Time (min) Time (min)

Host Cell Protein Aggregates

C Expt. #1 #2
Process Expt. #2 #1
Process
D Process
Expt. #1 #2 Process
Expt. #2 #1
100 5
90
80 4
70

% Aggregates
HCP (ppm)

60 3
50
40 2
30
20 1
10
0 0
60 160 260 360 460 560 60 160 260 360 460 560
Time (min) Time (min)

Fig. 5. (a) mAb concentration versus time during continuous operation for process #1. (b) mAb concentration versus time during continuous operation for process #2. (c)
Residual HCP levels after polishing steps (d) Residual aggregate levels after polishing steps.

time. The MMCEX sorbent cycle is 40 min and requires six columns.
These two polishing operations are operated on a single Cadence
BioSMB PD system with a single pump used to flow-through AEX
membrane and load onto two MMCEX columns in series. To accom-
modate this on a single Cadence BioSMB PD, three MMCEX sorbent
cycles are performed per AEX membrane cycle. This again shows
the flexibility and intensification offered by this approach. Unit
operations are combined even though they have different number
of columns and cycle times.
Two continuous integrated processes (process 1 and process 2
in Fig. 5) have been performed to test the performance of contin-
uous processing with this configuration. For both processes, prior
to capture step, 20 Liters of clarified CHO supernatant were con-
centrated using an in line concentrator down to 5 Liters and a
final mAb concentration of 4 mg/mL. Each process was operated
for a duration of around 10 h. Each Protein A and MMCEX column
was cycled 15 times, while AEX membrane capsules were cycled
5 times over the course of each experiment. The mAb production Fig. 6. Impact of continuous processing on consumables.
rate, or throughput, was approximately 2 g/h. Sample aliquots were
taken every hour for analysis of critical quality attributes. Aggre-
gate and HCP concentration after the final polishing step are shown of the surge tanks between steps was lowered in process 2 resulting
in Fig. 5. Overall, the continuous purification train resulted in a 4.5 in further reduction of the overall process time.
log reduction in HCP (from 500,000 ppm to 10 ppm), a 50% reduc- Fig. 6 summarizes the benefits of processing a 25 L HCCF sam-
tion in aggregates (from 2 to 3% aggregates in the load down to ple volume in continuous processing mode compared to batch
1–1.5% after an isocratic elution from the mixed-mode resin) and mode. Both of these processes assume the same sequence of unit
75% total yield. There was no significant time dependent perfor- operations being performed under optimized conditions for each
mance observed in the quality of the final polished mAb. The final respective mode of operation. For the batch mode of operation,
HCP concentration was approximately five-fold lower compared it is assumed that the columns are ideally sized to perform the
to the batch process confirming that modifying the design space purification in a single cycle. Under these assumptions, the over-
when transferring from batch to continuous processing can be used all process time from start to finish is similar for both continuous
to enhance purification performance. The aggregate clearance was and batch modes of operation. However, the amount of sorbent
50% for both batch and continuous operation. Notably, the volume required to operate the continuous process is considerably reduced.
For the Protein A sorbent, continuous operation results in a 94%
 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18
Fig. 7. Chronograms showing the 3 chromatography processes, (a) Protein A, (b) AEX and (c) MMCEX. Column positions are shown by the radial “spokes” (Column 1, C1,
through Column 8, C8). The outer annular ring shows the material the column is receiving and the inner annular ring shows where the flow out of the column is directed.
The columns proceed around the cycle in clockwise direction.
reduction of resin volume. For the polishing steps, total AEX mem-
brane volume would be reduced by 74%, while the MMCEX resin
volume would be reduced by 96%. This is a significant reduction
in the amount of resin required and could result in consider-
able cost savings. Buffer consumption would also be reduced by
44% by transferring from batch to integrated continuous process-
ing. The buffer consumption reduction comes from operating with
increased capacity, combining unit operations and intensification
of the process through both changing the order of unit operations
and reducing the number of column volumes required for some of
the process steps. The dramatic reduction in buffer consumption
might be a reason to convert facilities that are limited by buffer
production to switch to continuous processing. These results show
a compelling case for continuous production. However, an over-
all cost benefit analysis between batch and continuous for various
scales of production will be addressed in a subsequent publication.
4. Conclusion
Four mAb purification unit operations, Protein A, low pH viral
inactivation, AEX, and MMCEX were optimized for batch pro-
cessing and converted into continuous unit operations using two
Cadence BioSMB PD continuous purification systems. The study
demonstrates that significant operational efficiency and quality
advantages can be realized by moving from single column batch
processing to multi-column processing on a Cadence BioSMB PD
system. The study also demonstrates that further operational effi-
ciency is made possible by connecting and operating multiple unit
operations together and those continuous processes can be oper-
ated at steady state in order to produce a homogeneous, high purity
drug substance from start to finish.
In this study, processing continuously enabled a significant
reduction in consumable needs (volume of chromatography media,
buffer volume). With such a drastic reduction, it follows that con-
tinuous processing will have a proportionate impact on facility
footprint leading to a reduction in both operating and capital
expenditure. Reduction in capital expenditure could prove valuable
when advancing drugs through clinical trials because only a fraction
of drugs make it through the FDA. Reducing the capital expendi-
ture for each potential treatment during clinical trials reduces the
cost associated with the high risks of failure during clinical trials.
Furthermore, this type of process intensification, where multiple
columns are cycled far more frequently than in batch processing,
paves the way for dynamic process control using multi-variate data
17
analysis in combination with several in-line process analytical tech-
nologies. This means that in the near future, real-time release of
drug product will be possible. This will reduce the need for large
inventories of a given drug substance and ultimately lower the cost
of that drug to the patient.
References
Brower, M., Hou, Y., Pollard, D., 2014. Monoclonal Antibody Continuous Processing
Enabled by Single Use. Continuous Processing in Pharmaceutical
Manufacturing.
Castilho, L.R., 2014. Continuous Animal Cell Perfusion Processes: the First Step
Toward Integrated Continuous Biomanufacturing. Continuous Processing in
Pharmaceutical Manufacturing.
Ecker, D.M., Jones, S.D., Levine, H.L., 2015. The Therapeutic Monoclonal Antibody
Market. In MAbs, Vol. 7. Taylor & Francis, pp. 9–14 (January, No. 1).
Fromison, J., 2009. Disposables in clinical manufacturing. Am. Pharm. Rev. 12,
20–27.
Girard, V., Hilbold, N.J., Ng, C.K., Pegon, L., Chahim, W., Rousset, F., Monchois, V.,
2015. Large-scale monoclonal antibody purification by continuous
chromatography, from process design to scale-up. J. Biotechnol. 213, 65–73.
Gjoka, X., Rogler, K., Martino, R.A., Gantier, R., Schofield, M., 2015. A
straightforward methodology for designing continuous monoclonal antibody
capture multi-column chromatography processes. J. Chrom. A 1416, 38–46.
Godawat, R., Brower, K., Jain, S., Konstantinov, K., Riske, F., Warikoo, V., 2012.
Periodic counter-current chromatography–design and operational
considerations for integrated and continuous purification of proteins.
Biotechnol. J. 109 (12), 1496–1508.
Gottschalk, U., 2013. Biomanufacturing: time for change? Pharm. Bioprocess. 1 (1),
7–9.
Hernandez, R., 2015. Continuous manufacturing: a changing processing paradigm.
Biopharm. Int. 28 (April (4)), Retrieved from: http://www.
biopharminternational.com/continuous-manufacturing-changing-processing-
paradigm.
Hodge, G., 2004. Disposable components enable a new approach to
biopharmaceutical manufacturing. Biopharm. Int. 17, 38–49.
Horowitz, A., 2010. Is the reign of batch processing coming to an end? Bioprocess
Online, Retrieved from: http://www.bioprocessonline.com/doc/is-the-reign-
of-batch-processing-coming-to-0001.
Kaltenbrunner, O., Diaz, L., Hu, X., Shearer, M., 2016. Continuous bind-and-elute
protein A capture chromatography: optimization under process scale column
constraints and comparison to batch operation. Biotechnol. Progr. 32 (4),
938–948.
Kamarck, M.E., 2006. Building biomanufacturing capacity −the chapter.
Kelley, B., 2009. Industrialization of mAb Production Technology: the
Bioprocessing Industry at a Crossroads. In MAbs, Vol. 1. Taylor & Francis, pp.
443–452 (September, No. 5).
Konstantinov, K.B., Cooney, C.L., 2015. White paper on continuous bioprocessing.
may 20–21, 2014 continuous manufacturing symposium. J. Pharm. Sci. 104 (3),
813–820.
Lee, S.L., O’Connor, T.F., Yang, X., Cruz, C.N., Chatterjee, S., Madurawe, R.D.,
Woodcock, J., 2015. Modernizing pharmaceutical manufacturing: from batch
to continuous production. J. Pharm. Innov. 10 (3), 191–199.
Mahajan, E., George, A., Wolk, B., 2012. Improving affinity chromatography resin
efficiency using semi-continuous chromatography. J. Chrom. A 1227, 154–162.
18 X. Gjoka et al. / Journal of Biotechnology 242 (2017) 11–18
Palmer, E., 2015. Vertex, J&J, GSK, Novartis all working on continuous
manufacturing facilities. FiercePharma, Retrieved from: http://www.
fiercepharma.com/supply-chain/vertex-j-j-gsk-novartis-all-working-on-
continuous-manufacturing-facilities.
Pezzini, J., Joucla, G., Gantier, R., Toueille, M., Lomenech, A.M., Le Sénéchal, C.,
Garbay, B., Santarelli, X., Cabanne, C., 2011. Antibody capture by mixed-mode
chromatography: a comprehensive study from determination of optimal
purification conditions to identification of contaminating host cell proteins. J.
Chromatogr. A 1218, 8197–8208.
Pollock, J., Bolton, G., Coffman, J., Ho, S.V., Bracewell, D.G., Farid, S.S., 2013.
Optimising the design and operation of semi-continuous affinity
chromatography for clinical and commercial manufacture. J. Chrom. A 1284,
17–27.
Reichert, J.M., 2016. Antibodies to Watch in 2016. In MAbs, Vol. 8. Taylor & Francis,
pp. 197–204 (February, No. 2).
Strohl, William R., Strohl, Lila M., 2012. Therapeutic Antibody Engineering: Current
and Future Advances Driving the Strongest Growth Area in the Pharmaceutical
Industry. Elsevier, pp. 443.
Toueille, M., Uzel, A., Depoisier, J.F., Gantier, R., 2011. Designing new monoclonal
antibody purification processes using mixed-mode chromatographysorbents. J.
Chromatogr. B 879, 836–843.
Vogel, J.H., Nguyen, H., Giovannini, R., Ignowski, J., Garger, S., Salgotra, A., Tom, J.,
2012. A new large-scale manufacturing platform for complex
biopharmaceuticals. Biotechnol. Bioeng. 9, 3049–3058.
Warikoo, V., Godawat, R., Brower, K., Jain, S., Cummings, D., Simons, E., Johnson, T.,
Walther, J., Yu, M., Wright, B., McLarty, J., Karey, K.P., Hwang, C., Zhou, W.,
Riske, F., Konstantinov, K., 2012. Integrated and fully continuous production of
recombinant therapeutic proteins. Biotechnol. Bioeng. 109, 3018–3029.