Computational and Structural Biotechnology Journal 14 (2016) 238–244

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/csbj

Application of the Protein Maker as a platform purification system for

therapeutic antibody research and development
Geneviève Hélie a, Marie Parat a, Frédéric Massé b, Cory J. Gerdts c, Thomas P. Loisel a, Allan Matte a,⁎
a
Protein Purification, Human Health Therapeutics, National Research Council Canada, 6100 Royalmount Ave. Montreal, QC H4P 2R2, Canada
b
Primary Assays, Human Health Therapeutics, National Research Council Canada, 6100 Royalmount Ave. Montreal, QC H4P 2R2, Canada
c
Protein BioSolutions Inc., Suite 280, 401 Professional Drive, Gaithersburg, MD, 20879, USA

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

Article history: Within the research and development environment, higher throughput, parallelized protein purification is
Received 29 March 2016 required for numerous activities, from small scale purification of monoclonal antibodies (mAbs) and antibody
Received in revised form 8 June 2016 fragments for in vitro and in vivo assays to process development and optimization for manufacturing. Here, we
Accepted 13 June 2016
describe specific applications and associated workflows of the Protein Maker liquid handling system utilized in
Available online 16 June 2016
both of these contexts. To meet the requirements for various in vitro assays, for the identification and validation
Keywords:
of new therapeutic targets, small quantities of large numbers of purified antibodies or antibody fragments are
Parallelized protein purification often required. Reducing host cell proteins (HCP) levels following capture with Protein A by evaluating various
Antibody wash buffers is an example of how parallelized protein purification can be leveraged to improve a process
Protein A development outcome. Stability testing under various conditions of in-process intermediates, as an example,
Protein G the mAb product from a clarified harvest, requires parallelized protein purification to generate concurrent sam-
Hybridoma ples for downstream assays. We have found that the Protein Maker can be successfully utilized for small-to-mid
Process development scale platform purification or for process development applications to generate the necessary purified protein
samples. The ability to purify and buffer exchange up to 24 samples in parallel offers a significant reduction in
time and cost per sample compared to serial purification using a traditional FPLC system. By combining the
Protein Maker purification system with a TECAN Freedom EVO liquid handler for automated buffer exchange
we have created a new, integrated platform for a variety of protein purification and process development
applications.
Crown Copyright © 2016 Published by Elsevier B.V. on behalf of Research Network of Computational and
Structural Biotechnology. This is an open access article under the CC BY license (http://creativecommons.org/
licenses/by/4.0/).

1. Introduction fewer options available for generating purified quantitates of protein

in the intermediate (5–100) milligram scale. A few examples of
Purification of antibodies and antibody fragments are key activities customized solutions to this problem exist, involving integration of
in the generation of critical reagents for various in vivo and in vitro existing purification platforms such as the ÄKTA Purifier with a CETAC
assays as part of biotherapeutic lead identification and process develop- autosampler [3], ÄKTA Pure [4] or liquid handling robotics [5] have
ment. Often, it is necessary to purify large numbers of antibodies with been reported. Other solutions include the design and fabrication of
milligram yield, relatively quickly and at minimal cost. Various strate- customize robotics platform, including the Protein Expression and Puri-
gies are available to achieve such purification outcomes, and can involve fication Platform [6]. While some commercial instruments for purifica-
to various extents both automated and manual methods [1,2]. While tion of small quantities of protein have been developed, such as the
parallelized purification methods yielding sub-milligram quantities of QIAcube for purification of His-tagged proteins [7], there are few exam-
pure proteins based on packed columns, 96-well plates containing ples of commercial instruments that can be utilized for platform purifi-
small quantities of chromatographic resins or ligands immobilized to cation at milligram scale.
the surfaces of membranes have been developed, there are relatively In the context of process development applications, various
commercially available scale-down protein purification products have
been developed, including Predictor plates (GE) and Robo-columns
Abbreviations: CHO, Chinese Hamster Ovary; DPBS, Dulbecco's phosphate buffered saline; (GE and Atoll Bio). While very useful for early-stage screening of various
HC, IgG heavy chain; HCP, host cell protein; IMAC, immobilized metal ion affinity
chromatography; LC, IgG light chain; mAb, monoclonal antibody; OAA, one-armed antibody.
chromatographic conditions, the maximum size of the columns possible
⁎ Corresponding author. in these platforms (600 μL bed volume) results in a considerable gap in
E-mail address: [email protected] (A. Matte). the scale between screening and further optimization of process

http://dx.doi.org/10.1016/j.csbj.2016.06.001
2001-0370/Crown Copyright © 2016 Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).
 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244 239

conditions. Some examples of higher throughput, automated solutions time, column washing and sample elution) established using the ÄKTA
to purification process development have been reported [8,9]. While au- purification system. Sample and buffer lines were cleaned in place
tomated, sequential purification of samples is possible using a with 0.5 M NaOH and equilibrated in DPBS or appropriate buffer
chromatography system connected to an auto-sampler, this cannot be solutions. During purification, a portion of the flow-through fraction
parallelized using a single instrument, thereby reducing the possible was collected for subsequent non-reducing SDS-PAGE analysis. Protein
number of samples processed. samples were eluted in three steps, consisting of a pre-elution volume,
A specific instrument which has been designed around accomplishing elution volume and post-elution volume. Protein concentration mea-
the task of parallelized, medium scale purification is the Protein surements (A280 nm) on these fractions were used to establish the
Maker system, originated by Emerald BioStructures [10] and subse- final pooled sample.
quently developed and marketed by Protein BioSolutions. The
Protein Maker is an automated protein purification platform 2.3. Process development for mouse IgG2a purification
designed for purification of feed volumes of various sizes, from
~ 10 mL to 1 L (~ 1 mg to 100 mg) or more utilizing up to 24 chroma- The Protein Maker system was used to purify in parallel five murine
tography columns, each with an independent flow path. The main IgG2a samples from mouse Hybridomas using MabSelect SuRe and Pro-
components of the system are (i) the syringe pumps with the associ- tein G HP 1 mL HiTrap columns. For each mouse IgG2a, 20–22 mL of su-
ated 9-port valve, mixing syringe and sample lines, which together pernatant (~ 0.5 to 2 mg of mouse IgG2a, depending on titer) was
form the initial portion of the flow path, (ii) the column gantry, purified on either column using the purification method described
columns and associated tubing from the syringe pumps, which above. For protein G purifications, elution was performed in two steps,
form the subsequent portion of the flow path and (iii) the deck, first with 100 mM citrate buffer pH 3.6 and then with 100 mM
which contains up to 19 positions for SBS format plates and a dedi- glycine-HCl buffer pH 2.6. Elution fractions were neutralized using
cated waste position. 1 M Tris. The quantity of IgG2a contained in elution fractions were de-
While purification of a variety of proteins from any number of termined based on A280 nm. For protein G purification, the quantity of
sources is in principle possible with the instrument, the focus herein IgG2a obtained from the two elution steps was summed for calculating
are examples of purification of antibodies and their fragments the yield.
generated from mammalian expression systems. We have utilized the
Protein Maker as a key component of a platform purification system 2.4. Development of a post-load wash step to improve HCP removal during
that integrates automated buffer exchange implemented on a TECAN protein A purification
Freedom EVO liquid handler. This protein purification platform can
be used for both parallelized, small-medium scale purification of Data were obtained with three different antibodies expressed in
antibodies and their fragments, as well in various process development CHO cells. The Protein Maker system was used to perform parallel
applications. purifications using MabSelect SuRe 1 mL HiTrap columns. For each anti-
body, eight wash conditions were tested. For each tested condition,
2. Materials and methods 10 mL of supernatant (16.3 to 17.5 mg of antibodies) were loaded at a
residence time of 3 min. HiTrap columns were then washed and anti-
2.1. Antibody production bodies were eluted using 2.5 CV of 100 mM citrate buffer pH 3.0. Elution
fractions were neutralized using 1 M HEPES. The quantity of antibody in
Murine IgG samples were produced in hybridoma culture in IMDM elution fractions was determined by A280 nm. The quantity of HCP in
supplemented with 10% heat-inactivated FBS and mouse IL-6 by a elution fractions was measured using a CHO HCP ELISA kit (Cygnus
procedure previously described [11]. For some antibodies, cultures Technologies).
were performed transiently in Chinese Hamster Ovary (CHO) cells as
previously described [12]. Productions were harvested by centrifuga- 2.5. Buffer exchange and aseptic filtration
tion or filtration (0.22 μm or 0.45 μm) and IgG containing supernatants
stored at 4 °C until purified. Buffer exchange into DPBS following affinity purification was
performed manually either using Zeba-spin columns (Thermo-Fisher
2.2. Purification of mAbs and Fabs Scientific) by centrifugation or using PD-10 desalting columns (GE
Healthcare) by gravity according to the manufacturer's instructions.
For development of Protein Maker purification methods, protein Alternatively, sample buffer exchange using PD Miditrap G-25 columns
samples were purified using 1 mL HiTrap columns (GE Life Sciences), was automated on a TECAN Freedom EVO150® liquid handler accord-
including Protein G HP, MabSelect SuRe™ (Protein A) and Ni Sepharose ing to gravity protocols from GE Healthcare. The Freedom EVO150®
Excel™ mounted on an ÄKTA Purifier 10/100 system. Chromatographic was equipped with a liquid displacement Liquid Handler (LiHa)
profiles were monitored at 280 nm. Columns were equilibrated in configured with 8 channels (4 disposable tips and 4 washable tips), a
Dulbecco's Phosphate Buffered Saline Solution (DPBS, HyClone Robotic Manipulator (RoMa), a Tecan Vacuum (TeVac), carriers for 11
Laboratories), the sample applied at the appropriate residence time microplates, shelf for 4 microplates, one reservoir position for elution
(1 min for Ni Sepharose Excel or 3 min for MabSelect SuRe or Protein buffer, and tip carriers for hanging tips. All tubing and components of
G HP) and a portion of the flow-through fraction collected for subse- the liquid displacement system were cleaned in-place with 0.5 M
quent non-reducing SDS-PAGE analysis. Columns were washed with NaOH for at least 15 min and rinsed with sterile water prior to opera-
DPBS and bound proteins eluted with two column volumes (CV) of tions. A script was developed with the flexibility to process from 24 to
sodium-citrate buffer pH 3.6 (MabSelect SuRe), two CV of 100 mM 96 samples at once. Before starting, the storage solution from PD
glycine-HCl pH 2.6 (Protein G HP) or one CV of DPBS with 500 mM MidiTrap G25 columns was removed manually, the columns placed in
imidazole pH 8 (IMAC purification). For proteins eluted from Protein A a 24 position custom holder and the rack positioned on the Freedom
or Protein G columns, samples were pH adjusted using 1 M solutions EVO150® worktable. Purified protein samples from the Protein Maker
of sodium HEPES or Tris–HCl buffer to a final pH of 6–7. were stored in 24 deep well microplate (Seahorse Bioscience). The
Platform, parallelized purification experiments were performed system liquid was replaced with DPBS (HyClone Laboratories) and the
using the Protein Maker running the Protein Maker v2.0 software script started. First, the RoMa arm brought the column racks onto the
(Protein BioSolutions). Purification runs were performed using the TeVac and columns were equilibrated with three bed volumes using
1 mL HiTrap columns and the chromatography conditions (residence the Freedom EVO150® system liquid (DPBS). The equilibration buffer
240 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244

was allowed to enter the packed bed completely and the flow-through observed comparable results in terms of product yield and sample
discarded in the TeVac waste. Samples were pipetted onto the columns purity when using the same purification method on an ÄKTA purifica-
using disposable filter tips (Tecan) and time was allowed for them to tion platform and the Protein Maker (results not shown).
enter the packed bed by gravity. Column racks were tapped 4× times A key factor in establishing the capture step is to determine the
on the Te-Vac by the RoMa arm to remove any droplets and transported residence time to achieve optimal capture of the product on the column
on top of a 24 deep well block for elution. The LiHa pipetted 1.5 mL of of interest. In the case of Protein G affinity chromatography, using a
elution buffer (DPBS) to each MidiTrap, respectively, using disposable residence time of 3 min, no IgG was found in the flow-through of
tips and the eluate containing the protein of interest was collected by hybridoma-generated mAb samples as determined by non-reducing
gravity. Subsequent aseptic filtration was performed by centrifugation SDS-PAGE with Sypro Ruby staining. For larger volumes of feeds con-
using either sterile Multiscreen 0.22 μm 96-well plates (Millipore) or taining small quantities of product, it is necessary to split samples
deep well 0.22 μm 96-well plates (Corning) stacked with a sterile re- over two or more columns in order to reduce the total time required
ceiver plate. for the binding step. Using such an approach, ~200–250 mL of product
(~ 30 mg of protein in the case of Protein A) can be passed over 1 mL
2.6. Analytical methods columns in an overnight run.
We have found that using the approach of collecting elution
Proteins were quantitated based on A280 nm values obtained using fractions in three steps (pre-elution (0.5 CV), main elution (0.5–
a NanoDrop 2000 spectrophotometer (Thermo Scientific) and concen- 1.5 CV) and post-elution (1.5+ CV) volumes) allows for optimization
tration values were corrected based on calculated extinction coefficients of the quantity of purified product in a minimum volume, allowing for
derived from the protein sequence. Non-reducing SDS-PAGE analysis maximum product concentration. At the elution step, the practical out-
was performed using 4–12% Bis-Tris NuPage gels (Novex, Thermo- come is to obtain ≥ 80% of the purified protein in a minimum volume
Fisher scientific) and stained with Sypro Ruby protein gel stain suitable for manual or automated buffer exchange as the second step
(Thermo-Fisher Scientific) as recommended by the manufacturer. in the workflow shown in Fig. 2.
SDS-PAGE gels were imaged using a ChemiDoc MP imaging system
(Bio-Rad Laboratories).

3. Results

3.1. Development and implementation of protein maker platform

purification methods

Our strategy for the development and implementation of affinity

purification methods is summarized in Fig. 1. Key factors that influence
the purification process, including residence time for product capture
and elution volume are determined using the ÄKTA purification runs.
The essential value of these runs is to provide the absorbance trace
during the purification in order to quickly converge to appropriate
starting conditions for purification. The essential features of the
different affinity purification methods are similar for Protein G, Protein
A and IMAC based purification method development. Overall, we

Fig. 2. Protein G purification workflow for purification of mAbs from hybridoma

Fig. 1. Protein Maker purification development strategy. supernatants.
 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244 241

3.2. Platform purification applications

3.2.1. Small scale purification of mAbs from hybridoma supernatants

In order to generate purified mAbs for cell-based assays, we have
developed a Protein G based purification workflow using Protein G HP
as the capture resin (Fig. 2). Each purification cycle permits 23 mAb
samples along with one murine IgG control sample, to be purified
using Protein G, buffer exchanged into DPBS and aseptically filtered
prior to aliquoting in bar coded tubes for storage. The purpose of the
control sample is to act as a sentinel for the purification run via determi-
nation of the recovery of purified protein. In principle, this sample will
help to troubleshoot problems in executing the purification method,
as for example, a change in column binding capacity resulting from ex-
cessive numbers of clean in place (CIP) cycles.
In most cases, approximately 0.1–1 mg of purified mAb were
obtained per hybridoma supernatant, mainly with a concentration
range of approximately 0.75 to 0.95 mg/mL. Non-reducing SDS-PAGE
followed by staining with Sypro Ruby, a high-sensitivity protein stain,
revealed no IgG in the flow-through fraction (Fig. 3B). The purity of
the resulting purified mAbs was also verified by non-reducing SDS-
PAGE (Fig. 3A).
In early experiments, samples were buffer exchanged manually
using Zeba-spin centrifugal columns. While this approach minimizes
sample dilution, it is manually intensive, and sub-optimal within a
high throughput protein purification workflow (Fig. 2). Using PD Fig. 4. Non-reducing SDS-PAGE and Sypro Ruby staining of CHO-produced and IMAC
purified Fab samples using the Protein Maker. Lanes 1, 9, 17, molecular weight markers,
MidiTrap G-25 columns in gravity mode, we implemented an automat-
in kDa; lanes 2, 10, CHO supernatants of expressed Fabs; Lanes 3,4,11,12, flow-through
ed approach to buffer exchange samples following affinity purification, fractions; lanes 5,6,13,14, wash fractions; lanes 7,8,15,16, elution fractions. The ~50 kDa
reducing the manual labor, time and cost associated with this step. A species in the F/T fraction represents light chain (LC) dimers.
consequence of this change was increased dilution of the purified sam-
ple by ~1.5-fold compared to the Zeba-spin buffer exchange.
G or Protein A. The purity of representative Fab samples as determined
by non-reducing SDS-PAGE is shown in Fig. 4.
3.2.2. Antibody fragments — purification of Fabs
We applied the Protein Maker platform to the purification of a series 3.3. Process development applications
of his-tagged Fab constructs generated for in vitro studies. Other than
the buffer composition for column washing and sample elution, the 3.3.1. Evaluation of protein A vs protein G resin for purification of murine
overall strategy is similar to that applied to Protein G or Protein A IgG2a
purification methods. Using this approach, up to 20 mg of purified Fab Protein G resin is typically considered as the first choice to purify
from ~200 mL CHO culture has been obtained. Recoveries post-buffer antibodies produced from mouse Hybridomas, especially when the
exchange are similar to that obtained for mAbs purified using Protein IgG subclass is unknown. Indeed, Protein G has strong affinity for all
mouse IgG subclasses, whereas Protein A has strong affinity for only
certain subclasses, specifically IgG2a and IgG2b. For purification of
mouse IgG2a, the choice of one resin vs the other is not straightforward.
According to some manufacturers, mouse IgG2a has strong binding
affinity for both Protein A and Protein G, however, protein A columns
have a higher binding capacity and could be a better choice especially
for larger scale purification. To determine which of Protein A or Protein
G is the better choice to purify mouse IgG2a, we used the Protein Maker

Fig. 3. Non-reducing SDS-PAGE and Sypro Ruby staining of ten different hybridoma-
produced antibodies purified using Protein G HP with the Protein Maker (A) purified
mAbs, (lanes 2–11)(B) flow-through fractions (lanes 12–21), revealing no unbound
mAb (arrow). The mAb species (2HC + 2LC) is indicated with an arrow, and minor
species include 2HC (~ 100 kDa), half-antibody (~ 75 kDa) and free LC (~ 25 kDa) are Fig. 5. Quantities of mouse IgG2a (μg) purified from mouse hybridoma supernatants using
visible for some samples. Molecular weight markers (lanes 1, 22) are shown in kDa. either Protein A or Protein G.
242 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244

to purify five mouse IgG2a samples in parallel on these columns. The

total quantity of mouse IgG2a obtained with either resin following
purification is summarized in Fig. 5. For all tested mouse IgG2a, the
quantity of purified protein obtained from Protein A was at least 50%
higher than those obtained from Protein G purification.

3.3.2. Development of a post-load wash step to improve HCP removal

during protein A purification
Removing impurities post-Protein A chromatography still
represents a significant challenge to purification process development
in order to achieve the required drug substance specifications suitable
for patient administration (HCP b 100 ppm). Using a post-load wash
step is a key means to achieve HCP clearance, since it has been demon-
strated that HCP associates with antibodies and co-elute during the
elution step. Basic pH and wash additives such as arginine seem to
improve HCP removal during the Protein A chromatography wash
step by disrupting interactions between the antibody and the HCPs
[13]. Here we evaluated both Tris and phosphate-based arginine wash
buffers at different pH values (7 to 9) in comparison to conventional
wash buffer (DPBS or citrate pH 5.0) using 1 mL HiTrap MabSelect
SuRe columns. These various wash conditions (Table 1) were tested in
parallel with 3 different antibodies (CHO supernatants) using the
Protein Maker. Yield and HCP levels obtained for each wash condition
are presented in Fig. 6. The use of basic wash buffers containing arginine
showed improved HCP removal (1.7 to 2.4-fold) compared to the Fig. 6. Effect of wash buffers on (A) purification yields and (B) HCP levels. Wash
conventional DPBS wash for the three antibodies tested, with detrimen- conditions — A: PBS; B: 100 mM Citrate, 50 mM NaCl pH 5.0; C: 50 mM Tris, 100 mM
Arginine, 50 mM NaCl pH 7.0; D: 50 mM Tris, 100 mM Arginine, 50 mM NaCl pH 8.0; E:
tal effects on purification yields only for one antibody (mAb3). 50 mM Tris, 100 mM Arginine, 50 mM NaCl pH 9.0; F: 10 mM Phosphate, 100 mM
Arginine, 50 mM NaCl pH 7.0; G: 10 mM Phosphate, 100 mM Arginine, 50 mM NaCl
3.3.3. Stability hold of clarified harvest for a mAb pH 8.0; H: 10 mM Phosphate, 100 mM Arginine, 50 mM NaCl pH 9.0. *: Yields below
In a series of experiments, a clarified mAb harvest was held at either 80% — #: HCP reduction vs PBS greater than 1.7-fold.
2–8 °C or 19–23 °C and sampled at various time points corresponding to
t = 0, 1, 2, 3, 5 and 7 days before purification using MabSelect SuRe we selected the Protein Maker as a platform using HiTrap columns,
followed by buffer exchange to DPBS in order to evaluate changes in including Protein G (Protein G HP), Protein A (MabSelect SuRe) and
the product using an analytical assay panel. The purification perfor- IMAC (Ni Sepharose Excel), and have applied these to feeds from either
mance at time points, t = 0 and t = 7 days, is shown in Table 2. hybridoma or CHO supernatants ranging from 10 mL to 200 mL vol-
Purification recoveries based on the measured product titer were at umes. A key outcome has been for purification of small quantities
least 85%. Analysis of charge variants and glycosylation profiles at each (b1 mg) of purified mAbs using the Protein G method described here,
time point at either temperature revealed a significant decrease in acidic with greater than 1000 mAbs purified and utilized as reagents for
charge variants upon storage at 19–23 °C, corresponding to a drop in the in vitro screening assays.
sialic acid content of the mAb (results not shown). This assessment For the development of platform-based purification methods on the
made it possible to define the manufacturing hold time duration Protein Maker, several parameters that influence the overall purifica-
necessary for process control during GMP manufacturing. tion outcome required evaluation. One of the most critical parameters
is the residence time used in the binding step, as most of the overall
4. Discussion time required to execute the method involves binding of the product
to the packed bed. Through several purification campaigns utilizing
During early-stage therapeutic antibody R&D projects, it is often Protein G HP with murine IgG's, we have established that a three-
necessary to purify large numbers of samples that will be used as minute residence time ensures capture of the product, as revealed by
reagents for in vivo or in vitro screening assays. The nature of the non-reducing SDS-PAGE analysis. Another critical parameter is to
purification strategy employed is determined by the quantity and final establish the elution volume range for the product. The elution volume
concentration of purified protein required; the number of samples will depend on the column volume and the quantity of product bound
that need to be purified per unit time (throughput), the supernatant to the packed bed, with larger elution volumes being required as one
volume and initial product titer, as well as the availability, if any, of approaches the dynamic binding capacity limit of the column. The opti-
automated liquid handling instrumentation. While clearly there is mal elution volume is a balance between the desired product concentra-
more than one possible solution to achieve the purification objectives, tion, yield and recovery of the purified product. One rule of thumb

Table 1
Wash conditions tested for Protein A post-load HCP removal.

Wash 1 (5 CV) Wash 2 (5 CV) Wash 3 (5 CV)

1 PBS PBS PBS

2 PBS 100 mM Citrate, 50 mM NaCl pH 5.0 PBS
3 PBS 50 mM Tris, 100 mM Arginine, 50 mM NaCl pH 7.0 PBS
4 PBS 50 mM Tris, 100 mM Arginine, 50 mM NaCl pH 8.0 PBS
5 PBS 50 mM Tris, 100 mM Arginine, 50 mM NaCl pH 9.0 PBS
6 PBS 10 mM Phosphate, 100 mM Arginine, 50 mM NaCl pH 7.0 PBS
7 PBS 10 mM Phosphate, 100 mM Arginine, 50 mM NaCl pH 8.0 PBS
8 PBS 10 mM Phosphate, 100 mM Arginine, 50 mM NaCl pH 9.0 PBS
 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244 243

Table 2
Chromatographic Performance Table — MabSelect SuRe purification, clarified harvest stability hold study for t = 0 and t = day 7. The global process purification yield is calculated as the
ratio of the final yield over the initial load on the column.

PD-10 buffer exchange and

HiTrap Mab Select SuRe aeptic filtration

Load Elution Load Final

Storage temperature Time point Volume Quantity Quantity Step yield quantity quantity Step yield Global process
(°C) (Day) (mL) (mg) (mg) (%) (mg) (mg) (%) yield (%)

N/A 0 50 16.6 14.8 89 13.6 13.6 99.9 82.0

2–8 °C 7 50 16.5 14.0 85 12.8 12.6 98.7 76.6
19–23 °C 7 50 16.6 15.3 92 14.0 13.7 97.5 82.6

would be to establish the elution volume such that ≥80% of the purified purification scheme presented here, Fabs are purified via his-tags
protein is captured in one fraction. using IMAC, although Fab purification can be achieved in some in-
In many antibody affinity purification work-flows, purification is stances using protein A and more commonly using the CH1 domain of
followed by buffer exchange into a formulation buffer, including PBS, the Heavy Chain (HC) or the conserved domain of the Light Chain (LC)
in order to minimize protein aggregation as a result of unfavorable as capture modes. Other therapeutic antibody formats, including one-
conditions of pH or ionic strength. While it is possible to perform the armed antibodies, hybrids, bi-specifics and various Fc-fusion molecules
buffer exchange step in high-throughput mode using 96-well plates can also be purified in platform mode with the appropriate affinity
[14], this only applies to samples having a small volume, typically less resin. Examples of readily available affinity chromatography resins
than 130 μL. Initially, we performed this step in a more manually and how they can be applied to various antibody purification require-
intensive manner using Zeba-spin buffer exchange columns. In order ments is summarized in Table 3.
to increase the throughput and operational efficiency of the buffer ex- Both the post-load wash step for optimization of HCP removal from
change step, a more automated approach was developed using PD Protein A as well as the clarified harvest hold stability study offer
MidiTrap G-25 columns on a TECAN Freedom EVO 150 liquid handling examples of how the Protein Maker can be effectively utilized in the
system. The integration of an automated buffer exchange step in a 24- context of protein purification process development. Minimizing host
sample format, the same format as the Protein Maker, increases the cell protein levels at the Protein A capture step through modification
overall sample throughput, decreases manual manipulations and of wash buffer composition can improve the purification process [13,
enhances the consistency of results. 15,16]. One cycle of parallelized purification using the Protein Maker,
In addition to platform-based purification of full-size IgG's, there is requiring approximately 3 h, was sufficient to purify 24 samples
often a requirement to purify IgG fragments, including Fabs and scFv's, (eight different wash conditions for three different antibodies). Execut-
as well as to evaluate the effectiveness of different affinity purification ing the same experiment using one ÄKTA purification system (i.e. 24
resins for purification of the same product. For purification of murine sequential purifications on an ÄKTA Purifier) would have required ap-
IgG2a, the choice of Protein A vs Protein G purification requires experi- proximately 37.5 h. In this experiment, the Protein Maker not only in-
mental verification prior to establishing the final process to be used. By creased purification throughput by 12.5-fold but also permitted
performing all of the purification experiments in parallel using the Pro- purification of samples in an unbiased manner by having the capacity
tein Maker, a few hours were sufficient to determine that protein A is to purify all of the samples at the same time. Indeed, under conditions
the better choice for purification of murine IgG2a samples. Indeed, the where a mAb is unstable in the clarified harvest, the ability to purify
quantity of purified protein obtained from Protein A was at least 50% multiple samples in parallel offers a distinct advantage over serial
higher than those obtained from Protein G purification, although in purification using conventional purification equipment, minimizing
the absence of additional data we cannot offer a definitive explanation misinterpretation of data due to sample degradation. As the effective-
for this result. Executing the same experiment using one ÄKTA purifica- ness of optimized wash buffers and their effect on purification yields
tion system would require approximately five-fold more time than seem to vary depending on the antibody, the same wash conditions
performing the purification experiments in parallel with the Protein cannot be used for all samples. Rather, post-load wash conditions
Maker. It is noteworthy that these 5 antibodies were subsequently should be optimized for each antibody as part of purification process de-
purified at larger (2 L) scale using protein A resin with greater than velopment. Due to its ability to process multiple samples in parallel, the
80% recovery for four out of the five antibodies processed. In the Protein Maker represents a preferred instrument for performing wash
buffer screening. Moreover, as multiple wash conditions can be evaluat-
ed in parallel, it is possible to minimize variability in sample handling
that could bias data interpretation.
Table 3
Possible platform affinity purification modes for antibodies and antibody fragments using
Therapeutic antibody products often pose various challenges
the Protein Maker. during the development of upstream and downstream processes,
including degradation, modification of Fc or Fd glycans, reduction of in-
Capture
Format Isotype Species mode Resin
tramolecular disulfides, deamidation of Asn and Gln residues, as well as
aggregation [17]. In order to de-risk the overall process, an important
mAb, bispecific IgG1, IgG2, IgG4 Human Fc Protein A
component of the transition between upstream and downstream pro-
mAb, bispecific IgG2 Murine Fc Protein A
mAb, bispecific IgG1, IgG2, IgG3, IgG4 Human Fc Protein G cessing is to understand the stability of the product post-clarification
mAb, bispecific IgG1, IgG2, IgG3 Murine Fc Protein G but prior to the initial capture purification step. The removal of sialic
OAA N/A Human Fc Protein A, G acid from mAbs by the action of extracellular CHO sialidase is an
scFv Human LC (VL) Protein L example of the kinds of post-production, pre-purification modifications
Hybrid Human Fc Protein A, G
Hybrid Human HC (CH1) CH1
that are possible [18]. The utility of the Protein Maker in this application
Hybrid Human LC (κ, CL) Kappa-Select is its ability to purify several samples in parallel, for example, if the clar-
Fab Human HC (CH1) CH1 ified harvest is to be held at various temperatures or at different pH
Fab Human LC (κ, CL) Kappa-Select values, or if multiple harvests are to be tested in parallel and purified
Fab Human LC (λ, CL) Lambda-Fab Select
under the same controlled conditions.
244 G. Hélie et al. / Computational and Structural Biotechnology Journal 14 (2016) 238–244

5. Conclusions [4] Holenstein F, Eriksson C, Erlandsson I, Norrman N, Simon J, Danielsson A, et al.

Automated harvesting and 2-step purification of unclarified mammalian cell-
culture broths containing antibodies. J Chromatogr A 2015;1418:103–9.
We have found the Protein Maker to be a versatile tool for a number [5] Alm T, Steen J, Ottosson J, Hober S. High-throughput protein purification under de-
of purification problems, ranging from small scale mAb and Fab purifica- naturing conditions by the use of cation exchange chromatography. Biotechnol J
2007;2:709–16.
tion to various applications in process development. We have devel- [6] Gonzalez R, Jennings LL, Knuth M, Orth AP, Klock HE, Ou W, et al. Screening the
oped platform purifications methods using Protein G, Protein A and mammalian extracellular proteome for regulatorsof embryonic human stem cell
IMAC and have applied these to several projects. Use of the Protein pluropotency. Proc Natl Acad Sci U S A 2010;107:3552–7.
[7] McGraw J, Tatipelli VK, Feyijinmi O, Traore MC, Eangoor P, Lane S, et al. A semi-
Maker and TECAN Freedom EVO150 together has resulted in reducing
automated method for purification of milligram quantities of proteins on the
both the required personnel time and creation of an integrated QIAcube. Protein Expr Purif 2014;96:48–53.
workflow that includes both protein purification and automated buffer [8] Kelley BD, Switzer M, Bastek P, Kraimarczyk JF, Molnar K, Yu T, et al. High-
throughput screening of chromatographic separations: IV. Ion-exchange. Biotechnol
exchange in a 24-sample format. Future implementation of the UV
Bioeng 2008;100:950–63.
monitoring capability will add an additional dimension to the capabili- [9] Teeters M, Bezila D, Alread P, Velayudhan A. Development and application of an
ties of this instrument, permitting 24-channel monitoring of the protein automated, low-volume chromatography system for resin and condition screening.
absorbance signal, thereby expediting utilization of the instrument for Biotechnol J 2008;3:1212–23.
[10] Smith ER, Begley DW, Anderson V, Raymond AC, Haffner TE, Robinson JI, et al. The
various process development applications. Protein Maker: an automated system for high-throughput parallel purification.
Acta Crystallogr F 2011;67:1015–21.
Acknowledgements [11] Yokoyama WM, Li Z. Monoclonal antibody supernatant and ascites fluid production.
Curr Protoc Immunol 2001 [40:II:2.6:2.6.1–2.6.9].
[12] Raymond C, Robotham A, Spearman M, Butler M, Kelly J, Durocher Y. Production of
We wish to thank our colleagues within the Protein Purification α2,6-sialylated IgG1 in CHO cells. MAbs 2015;7:571–83.
Team who have provided valuable feedback on the various methods [13] Chollangi S, Parker R, Singh N, Li Y, Borys M, Li Z. Development of robust antibody
purification by optimizing protein-A chromatography in combination with
that have been developed. We also wish to thank our colleagues in precipitation methodologies. Biotechnol Bioeng 2015;112:2292–304.
various projects who have provided us with materials with which we [14] Ying W, Levons JK, Carney A, Gandhi R, Vydra V, Rubin AE. Semiautomated sample
have developed the various methods described. In particular we thank preparation for protein stability and formulation via buffer exchange. J Lab Autom
May 12 2015 [pii: 2211068215585484].
Maria Jaramillo and Anne Marcil for their encouragement. This research [15] Aboulaich N, Chung WK, Thompson JH, Larkin C, Robbins D, Zhu M. A novel ap-
was supported by the National Research Council Canada. proach to monitor clearance of host cell proteins associated with monoclonal anti-
bodies. Biotechnol Prog 2014;30:1114–24.
[16] Sisodiya V, Lequieu J, Rodriguez M, McDonald P, Lazzareschi KP. Studying host cell
References
protein interactions with monoclonal antibodies using high throughput protein A
chromatography. Biotechnol J 2012;7:1233–41.
[1] Cummins E, Luxenberg DP, McAleese F, Widom A, Fennell BJ, Damanin-Sheehan A,
[17] Vasquez-Ray M, Lang DA. Aggregates in monoclonal antibody manufacturing
et al. A simple high-throughput purification method for hit identification in protein
processes. Biotechnol Bioeng 2011;108:1494–508.
screening. J Immunol Methods 2008;339:38–46.
[18] Gramer MJ, Goochee CF, Chock VY, Brousseau DT, Sliwkowski MB. Removal of sialic
[2] Su B, Hrin R, Harvey BR, Wang YJ, Ernst RE, Hampton RA, et al. Automated high-
acid from a glycoprotein in CHO cell culture supernatant by action of an extracellular
throughput purification of antibody fragments to facilitate evaluation in functional
CHO cell sialidase. Biotechnology (N Y) 1995;13:692–8.
and kinetic based assays. J Immunol Methods 2007;322:94–103.
[3] Yoo D, Provchy J, Park C, Schulz C, Walker K. Automated high-throughput protein
purification using an AKTApurifier and a CETAC autosampler. J Chromatogr A
2014;1344:23–30.