Journal of Chromatography A 1687 (2023) 463696

Contents lists available at ScienceDirect

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

Purification of hydrophobic complex antibody formats using a
moderately hydrophobic mixed mode cation exchange resin
Wolfgang Koehnlein a , Annika Holzgreve b , Klaus Schwendner a , Romas Skudas b ,
Florian Schelter a,∗
a
b

Roche Diagnostics GmbH, Nonnenwald 2, 82377 Penzberg, Germany
Merck KGaA, Frankfurter Str. 250, 64293 Darmstadt, Germany

a r t i c l e

i n f o

Article history:
Received 19 September 2022
Revised 21 November 2022
Accepted 30 November 2022
Available online 5 December 2022
Keywords:
Complex antibody formats
Downstream purification
Hydrophobicity
Mixed mode chromatography

Product-related impurities

a b s t r a c t
Immunoglobulins of complex formats possess great potential for increased biopharmaceutical efficacy.
However, challenges arise during their purification as the removal of numerous product-related impurities typically requires several expensive chromatographic steps. Additionally, many complex antibody
formats have a high hydrophobicity which impairs the use of conventional mixed mode chromatography.
In the present study, both of these challenges were addressed through the development of an innovative mixed mode resin with 2-amino-4methylpentanoic acid ligands that combines weak cation exchange
with moderate hydrophobic interactions. Supported by high throughput partition coefficient screens for
identification of preferable pH and salt concentration ranges in bind and elute mode, this mixed mode
resin successfully demonstrated efficient impurity separation from an extremely hydrophobic bispecific
antibody with a single unit operation. High purity (>97%) was obtained as a result of significant reduction of product-related impurities as well as process-related host cell proteins (>3 log scale), while
maintaining satisfactory recovery (70%). This also supports that highly hydrophobic antibody formats can
be efficiently purified using a resin with moderate hydrophobic characteristics. Studies involving additional antibodies possessing different formats and a wide range of hydrophobicity confirmed the broad
applicability of the new resin. In view of its high selectivity and robust operating ranges, as well as the
elimination of the need for an additional column step, the novel resin enables simplified downstream
processing and economic manufacturing of complex antibody formats.
© 2022 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license
( />
1. Introduction
In search of continual enhancement of recombinant monoclonal
antibodies (mAbs), development focuses on increasingly complex
molecular constructs and mechanisms of action. In this regard,
bispecific antibodies (bsAbs) or even multispecific formats have
emerged as an important class of biopharmaceuticals [1]. Their
particular property of targeting more than one antigen raises the
expectations of improved drug efficacy in comparison to conventional monospecific mAbs [2]. A great potential also arises from
the generation of a broad spectrum of distinct complex antibody
formats that may facilitate the fit for specific applications. Various
engineering strategies using full-length antibodies, specific fragment or special modules create combinatorial diversity and offer


∗

Corresponding author.
E-mail address: (F. Schelter).

more flexibility regarding valency and specificity or even enable
novel functionalities [1,3,4]. Among the numerous approaches
described for the formation of bsAbs [5], the CrossMAb technology
is a pioneer that performs exchange and crossover of antibody
domains to enable correct light chain association with the respective heavy chain counterparts [6,7]. This technology offers
stable constructs suitable for drug development and production
rendering versatile CrossMAb formats available today [8].
The opportunities for enhanced drug efficacy offered by the
various antibody formats are, however, associated with increased
challenges for product manufacturing at high yield and purity
[8,9]. In addition to common process-related impurities, like host
cell proteins (HCP) and DNA, complex antibody formats tend to
form product-related impurities at higher number and level than
standard mAbs (see Fig. 1). The more different components are
required for correct antibody formation, the more single byproducts, mis-paired forms, or aggregates may occur that appear as
low molecular weight (LMW) and high molecular weight (HMW)

/>0021-9673/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( />

W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 1. Characterization of the antibodies used. pI: Isoelectric point; HCP: host cell protein; HMW: high molecular weight; LMW: low molecular weight.


impurities in the product [9]. Hence, type and amount of impurities are antibody-specific and usually depend on the format.
Although, respective platform technologies have been developed
to facilitate the formation of the desired bsAb product, such as
the knob in hole (KIH) technology in case of CrossMAb [5,10,11],
the processing of complex antibody formats is still often difficult,
elaborate and limited regarding yield and purification success.
Generally, purification of complex formats may benefit from
special adaptations of the chromatography-based downstream processing established for standard mAbs [12]. The procedure typically
comprises up to four sequential steps. i.e., affinity chromatography
(mostly Protein A-based), ion exchange chromatography (IEX)
either with anions (AEX) or cations (CEX), hydrophobic interaction chromatography (HIC), and the so-called mixed mode or
multimodal chromatography (MMC) primarily combining electrostatic and hydrophobic interactions [1,9]. Respective modifications
mainly consider improved separation of product-related impurities
from the antibody monomer, as was also shown for the use of
MMC to achieve efficient bsAb purification [13]. This type of chromatography depends on the interplay of special resins with the
fine-tuned conditions of binding and elution generated by ionic
and hydrophobic interactions together with the effects of hydrogen
bonding [14]. Thereby, pH primarily controls ionic interactions and
binding capacity in favor of high recovery, whereas conductivity
(i.e. salt concentration) rather determines hydrophobic interactions
supporting separation capability rendering these two parameters
essential for optimization [15,16]. In addition, improvement in
MMC separation performance and selectivity may emerge from
the development of new functional groups/ligands and the use of
additives [17].
A further challenge for the purification of antibodies with
complex formats may arise from particular physical properties,
such as high hydrophobicity. Several bsAb with complex formats
investigated in our lab display high to extremely high hydrophobicity that counteracts successful application of available MMC.

The alternatively used two-step process of CEX and HIC, however,
increases effort, time, and costs of the manufacturing process.

To establish an efficient MMC step for highly hydrophobic antibodies, a new resin with refined characteristics is desirable
providing adequate moderate affinity for a bind and elute mode,
while maintaining favorable separation properties for impurity
elimination. Moreover, a high-quality resin should offer a broad
window of operation resulting from distinct graduation of binding
characteristics in the range of standard pH and salt conditions to
facilitate the development of robust processes.
Based on these demands, the present study investigated the usability of the newly developed mixed mode cation exchange chromatography (MMCEX) resin, Eshmuno®CMX, for the purification of
antibodies with complex formats and high hydrophobicity. In this
study, partition coefficient (Kp) screens as a high throughput tool
[18] has been used to analyze the binding behavior of antibodies
as a function of pH and salt conditions. Initial experiments indicated that functional groups of desired moderate hydrophobicity
for the resin might be found among carboxylic acids. Based on
these experiments, 2-Amino-4-methylpentanoic acid was selected.
The most hydrophobic antibody representing a trivalent bispecific (2+1) format [7] was then used to evaluate the operational
window of the novel resin for high product yield and efficient
removal of product-related as well as process-related impurities.
A refined bind and elute mode was verified by column runs
confirming efficient separation of LMW, HMW and HCP impurities.
The application space of Eshmuno® CMX was further examined
with two other antibodies covering a wider hydrophobicity range
and different impurity profiles. In summary, this study shows that
highly hydrophobic complex antibody formats can be efficiently
purified using a MMC with moderate hydrophobic characteristics.
2. Materials and methods
2.1. Antibody material
Humanized IgG1 antibodies with standard and complex formats were produced in Chinese hamster ovary (CHO) cells and

characterized by isoelectric point and impurity profiles (methods
2


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 2. Determination of relative hydrophobicity of the antibodies using commercially available pharmaceutical antibodies for linear regression analysis.

described in the following sub-sections) as shown in Fig. 1. In
addition, the relative hydrophobicity of the antibodies was determined by a standard HIC method [19] using a TSKgel® Ether-5PW
HPLC Column, 10 μm (Tosoh Bioscience). After conditioning, mAb
amounts of 20 μg (sample volume 20 μL) were applied to a gradient run (flow rate: 0.8 mL/min) consisting of eluent A (25 mM
sodium phosphate buffer with 1.5 M ammonium sulfate, pH 7.0)
and eluent B (25 mM sodium phosphate buffer, pH 7.0) starting
after 2 min equilibration with 100% eluent A and reaching 100%
eluent B after 60 min. The retention times of eluted mAbs were
determined at peak maximum. The values of commercially available pharmaceutical antibodies together with the molecule with
lowest retention time (represented by 0) and the molecule with
highest retention time (represented by 1) served for regression
analysis to determine the relative hydrophobicity of the antibodies
under study (Figs. 1 and 2).
The eluate obtained from affinity chromatography columns
(Protein A) was conditioned to pH 5 and after depth and sterile
filtration (to remove precipitates) was used as starting material for
resin investigation.

rial was combined with the mixture (Eshmuno® particles) and
polymerization is started by adding Cerium (IV) nitrate. After

the grafting reaction, non-reacted components and starter are
removed by extensive washing using acidic, basic and solvent
mixtures.
The following chromatography resins were used for comparative studies: CaptoTM MMC ImpRes, (Cytiva); NuviaTM cPrimeTM
(Bio-Rad Laboratories), Eshmuno® HCX (Merck KGaA, Darmstadt,
Germany).
2.3. Execution of partition coefficient (Kp) screens
Kp screens served as a high-throughput screening tool to evaluate the binding behavior of antibodies and impurities as a function of pH and salt concentration [18]. The filter plates (96 multiwell plates, AcroprepAdvTM , 0.45μm, Polypropylen, Pall Corporation) containing 50 μL of the respective resin (50 % slurry in water (v/v)) per well and 300 μL water were prepared by a Microlab
STARlet robot (Hamilton) equipped with a shaker.
A total of 48 defined buffer conditions (six pH values x eight
salt concentrations) were investigated. The Tecan Freedom EVO®
200 liquid handling station (software Evoware®, Tecan Deutschland GmbH) created a stepwise increase in pH and sodium sulfate
molarity in the wells by combining buffer stock solutions of each
pH with increasing volumes of the respective buffer stock solution containing 1.4 M sodium sulfate. In addition, an equilibration
buffer plate and a load plate containing the protein sample were
produced. The system was equipped with one liquid handling arm
(processing volumes of 10 μL to 10 0 0 μL), one excentric gripper, a
Te-ShakeTM , a Te-StackTM for storage of microplates, a Te-SlideTM
for plate transport, an Infinite® M200 plate reader for measuring
protein concentrations (software Magellan), and an integrated centrifuge (Rotanta 46RSC, Hettich).

2.2. Chromatography resins
Eshmuno® CMX (Merck KGaA, Darmstadt, Germany prepared
by graft polymerization of 2-Amino-4-methylpentanoic acid was
used for the purification experiments with hydrophobic antibodies. Additionally, two prototypes have been prepared by graft
polymerization of a carboxyl group or 2-aminopropanoic acid.
The synthesis of the prototypes was performed by dissolving
the respective amino acid in deionized water. Then an acrylic
compound like acrylic acid chloride or acrylic acid was added
to form a monomer which is necessary for the grafting process.

Afterwards the functional group -OH containing base mate3


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Initial Kp screens evaluating the MMC resins CaptoTM MMC ImpRes, NuviaTM cPrimeTM , and Eshmuno® HCX as well as the prototypes of Eshmuno® CMX with different functional groups/ligands
were performed in a buffer of acetate, MES, HEPES, or TAPS 20
mM. Each buffer was used in its optimum buffering range and
titrated with NaOH to adjust the following pH values: 5.0, 5.7, 6.4,
7.1, 7.8, and 8.5. Additionally, buffers were combined with the following sodium sulfate concentrations: 20, 160, 300, 440, 580, 720,
860, and 10 0 0 mM. The second series of Kp screens evaluating
Eshmuno® CMX with the three hydrophobic antibodies was executed with 25 mM Tris/acetate buffer adjusted to the pH values:
4.0, 5.0, 6.0, 7.0, 8.0, and 9.0 combined with the following sodium
sulfate concentrations: 25, 75, 150, 250, 350, 500, 650, and 800
mM.
Antibody solutions used to prepare the load plate were additionally subjected to concentration and diafiltration using 30 kDa
TFF membrane Biomax® 30kDa, Pellicon® 3 (Merck KGaA, Darmstadt, Germany) in 25 mM Tris/acetate, pH 6.5 to enable subsequent adjustment of load density and pH conditions in the plates.
In the first series of Kp screens the load density accounted for 5
g/L resin, in the second series a load density of 30 g/L resin was
applied.
Execution of Kp screens on the filter plates consisted of the following steps (each with 300 μL pipetting volume): two times of
equilibration, one loading step followed by stripping and regeneration. After each step, centrifugation collected the flow-through in
a plate and measurement of protein concentration and further analytics were performed. The load (300 μL) derived from the load
plate was incubated for 30 min on a shaker to reach an equilibrium between bound and unbound state, before the flow-through
was collected in a flow-through plate.
The protein concentrations of the load plate and flow-through
plate were determined by an UV absorbance based method using the Infinite® M200 plate reader. The binding capacity of the
resin for the protein was visualized by contour plots displaying the

portion of protein flow-through. Additional analysis was performed
by SE-HPLC (Section 2.5) to determine LMW, main peak monomer,
and HMW levels displayed as flow-through contour plots.

(Tosoh Bioscience) using an UV based detector (at 280 nm) on
the UltiMateTM 30 0 0 RSLC instrument (Thermo Fisher Scientific).
Analysis was performed with a mobile phase of 0.20 M potassium
phosphate and 0.25 M potassium chloride, pH 6.2 and a flow rate
of 0.3 mL/min at room temperature (19-26°C). The injected sample
volume was 10 μL (5 g/L) resulting in 50 μg per load. Samples of
lower concentration were used undiluted.
2.6. Analytical capillary electrophoresis (CE-SDS)
To examine obtained samples under denatured conditions, the
capillary electrophoresis system Caliper Labchip GXII (PerkinElmer)
was used to execute conventional SDS-PAGE in a chip format. The
included single components were separated according to molecular
weight by disconnecting all non-covalently bound molecules from
each other. Analysis was performed using the HT Protein Express
Reagent Kit (PerkinElmer) with a sample volume of 5 μL (1 g/L)
under non-reducing conditions (pH 8.7). Samples of lower concentration were used undiluted. Samples were labeled by a fluorescent
dye to enable laser detection.
2.7. HCP electrochemiluminescent immunoassay (ECLIA)
The residual Chinese Hamster Ovarian HCP (Host Cell Protein)
content in the samples was determined by the ECLIA-HCP assay on
a cobas ®e 801 immunoassay analyzer (Roche Diagnostics). The assay was based on a sandwich principle by using biotinylated polyclonal CHO HCP-specific antibodies to bind the target. This complex was fixed to the solid phase by streptavidin-coated microparticles. Thereafter, a ternary sandwich complex was formed by addition of a second polyclonal CHO HCP-specific antibody labeled
with a ruthenium complex that allowed chemiluminescence detection. The assay displayed a detection limit for CHO HCP of 2 ng/mL,
a quantification limit of 7.5 ng/mL and a linear measuring range up
to 10 0 0 ng/mL.
2.8. DNA measurement
DNA originating from host cells was measured for the column

runs using an automated quantitative PCR method performed in
the 96-well format by the FLOW Flex system (Roche Diagnostics)
consisting of three modules: The FLOW PCR SETUP Instrument
used as FLOW primary sample handling system, the MagNa Pure
96 system for automated DNA extraction using the MagNA Pure LC
Total Nucleic Acid Isolation Kit – High Performance, and the LightCycler® 480 system for DNA amplification and quantification using the Residual DNA CHO Kit specific for highly conserved CHO
DNA regions. The method had a detection limit of 0.4 pg/mL and
a quantification limit of 4 pg/mL with a linear measuring range up
to 40 0 0 pg/mL.

2.4. Eshmuno® CMX column runs
Column runs for MAB A using a bind and elute mode were
performed on columns with a bed height of 20 cm and a diameter of 1 cm. The residence time was 4 min corresponding to a
linear flowrate of 300 cm/hour. The dynamic binding capacity at
pH 5.0 and 10% breakthrough resulted in 64 g/L. Equilibration was
done with 5 column volumes (CV) of equilibration buffer (50 mM
sodium acetate, 25 mM sodium sulphate, pH 5.0) followed by load
application (density 10 g/L). The following washing step (wash 1:
100 mM MES, pH 5.5, 3.5 CV) served to override the buffer capacity of the resin. With the next washing step, the pH was shifted
to basic conditions (wash 2: 15 mM Tris/acetate, pH 9.0, 5 CV) and
the last washing step served to prepare elution conditions (wash
3: 50 mM sodium citrate, 100 mM sodium sulfate, pH 6.1, 6 CV).
The product was eluted with elution buffer (100 mM sodium succinate, 300 mM sodium sulfate, 10 CV). Two runs with different
elution pH levels were performed (run 1 at pH 6.0, run 2 at pH
6.2). Finally, the column was cleaned in place (CIP) (100 mM arginine, pH 10.9, 5 CV followed by 1 M sodium hydroxide, 16 CV).

3. Results and discussion
3.1. Evaluation of standard MMC for the purification of a highly
hydrophobic molecule
Initial experiments evaluated the impact of antibody hydrophobicity on the usability of available MMC resins for the purification

process. Kp screens performed with the eluate of Protein A chromatography over the range of standard pH (5-8.5) and salt concentrations (0-10 0 0 mM) were used to investigate the binding properties of CaptoTM MMC ImpRes, a common available MMC resin,
for three antibodies with increasing relative hydrophobicity determined by standard HIC (Fig. 2). Obtained contour plots showed a
satisfactory window for binding and flow-through reflecting suitable elution conditions for tocilizumab, the antibody with low hydrophobicity (0.06) (Fig. 3A). MAB C, the antibody with moderate

2.5. Size exclusion high performance liquid chromatography
(SE-HPLC)
The separation of HMW and LMW from the antibody monomers
in the samples was analyzed by SE-HPLC under native conditions
using a TSKgel® UP-SW30 0 0 column, 2 μm, 4.6 mm x 300 mm
4


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 3. The impact of antibody hydrophobicity on the usability of different standard mixed mode resins. A) Kp screen contour plots of three antibodies with increasing
relative hydrophobicity (0.06, 0.33, 0.74) using Capto MMC Impress. B) Kp screen contour plots for the highly hydrophobic MAB A performed on three different mixed mode
resins.

hydrophobicity (0.33) displayed a smaller window with a narrowed
transition zone in the low salt range, but still appeared acceptable.
In contrast, the highly hydrophobic MAB A (0.74) showed efficient
binding, however, completely lacked an elution window under the
pH and salt conditions used.
Two further tested MMC resins, NuviaTM cPrimeTM and Eshmuno® HCX, provided a similar picture regarding a strong binding of the highly hydrophobic antibody without offering suitable
conditions for elution (Fig. 3B). Only for Eshmuno® HCX, a small
elution window at high pH and moderate salt concentration was
indicated, however, at unsatisfactory recovery level. Overall, the
tested MMC resins were not applicable for the processing of the

hydrophobic bsAb under study, as they impaired efficient elution
and recovery. This outcome may be explained by the fact that
all three standard MMC resins use phenyl groups as hydrophobic
functional groups/ligands that mediate strong hydrophobic interaction.

acids containing a short hydrophobic linear or branched hydrocarbon part were found suitable as possible ligands for Eshmuno®
CMX. This new MMC resin material is based on weak cation exchange groups and displays a pronounced three-dimensional structure in the form of tentacles that offers extended steric access for
binding and is, therefore, supposed to favor the separation of target
molecules with moderate or minor differences (Fig 4A).
The goal was to complement these binding properties with an
appropriate component providing the required level of hydrophobic interaction. To this end, three prototypes of the new resin
containing functional groups/ligands with different hydrophobicity
were tested in Kp screens with the highly hydrophobic MAB
A. A carboxyl group, 2-aminopropanoic acid, and 2-Amino-4methylpentanoic acid were selected as functional groups/ligands
based on increasing hydrocarbon parts to build resins with increasing hydrophobicity (Fig. 4B). The contour plots obtained for
the carboxy group and 2-aminopropanoic acid were quite similar
and showed wide areas with low or no binding capacity, indicating
a low modulatory impact especially of pH (Fig. 3B). These binding
characteristics were not suitable to generate adequate operational windows. In contrast, 2-Amino-4-methylpentanoic acid, the
most hydrophobic functional group/ligand examined displaying a
branched hydrocarbon part, showed a pronounced effect of pH
resulting in a more graduated contour plot with separated areas of
efficient binding and elution over a wider transition zone. A high
binding rate appeared at low salt content over the range > pH 4.5,
while increasing salt concentration decreased binding affinity and
yielded efficient elution in the range of moderate salt and neutral
to weak acidic pH conditions. These results indicated a suitable operational window for a bind and elute mode for MAB A along with
favorable preconditions for the elimination of the included impurities. Thus, 2-Amino-4-methylpentanoic acid could be identified
as an appropriate functional group/ligand to provide moderate hy-


3.2. Development of a novel MMC resin with moderate hydrophobic
interaction
The impurity profile of MAB A, containing in addition to high
amounts of HMW and HCP also a significant LMW fraction (Fig. 1),
required processing in a bind and elute mode. Therefore, the development of a new resin suitable for highly hydrophobic molecules
aimed to provide a broad operating window that clearly identifies
proper conditions for binding and elution. This is facilitated by a
wide transition zone with gradual differences from binding to nonbinding conditions for the antibody monomer across standard pH
and salt concentrations. In this regard, the findings with available
MMC resins indicated that the hydrophobic interaction component
had to be weakened towards moderate binding affinity. In search of
structures providing desired moderate hydrophobicity, carboxylic
5


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 4. Development of a novel mixed mode resin for the purification of hydrophobic molecules. A) Structure of the three prototypes with tentacle chemistry. B) Kp screen
contour plots for three prototypes of Eshmuno® CMX with different functional groups/ligands of increasing hydrophobicity using the highly hydrophobic MAB A.

drophobic interactions. In combination with weak cation exchange
groups, the novel Eshmuno® CMX resin showed promising binding
characteristics for the design of robust purification processes for
highly hydrophobic antibodies with complex formats.

pH range with slightly raised ionic strength was found suitable to
prepare elution while simultaneously removing further LMW. An
appropriate elution window could then be reached with increasing

salt concentrations exceeding 250 mM. At a closer view, the data
indicated a balance between purity and recovery in this zone
that was more dependent on pH than on salt conditions. A more
acidic pH appeared to support higher purity due to improved
HCP reduction at the cost of lower product recovery, while a shift
towards neutral pH pointed to increased recovery associated with
slightly increased HCP content. Thus, the contour plots served
as an overall survey to assess how individual performance parameters can be modulated to obtain desired product yield and
quality. Although Kp screens are operated as a static process, they
provide useful information for the design of a dynamic column
process.

3.3. Evaluation of Eshmuno® CMX for the purification of a highly
hydrophobic bsAb with a complex format
In the next step, the space for suitable modes of purification
for the MAB A should be further specified, also in regard to the
separation capacity for the major impurities HMW, LMW, and HCP
that were present at relatively high levels (Fig. 1). To this end,
a Kp screen was repeated focusing on salt concentrations of 0800 mM over the range of pH 4-9 (Fig. 5) that largely reproduced
the outcome for protein binding regarding the initial experiments
with 2-Amino-4-methylpentanoic acid as functional group/ligand.
In addition, the screen enclosed contour plots reflecting the binding behavior of the main peak and of the individual impurity types.
Obtained results revealed a suitable window for HMW reduction
around pH 5 and low salt concentration, for LMW reduction at low
salt and basic conditions around pH 8, and for HCP reduction at
basic pH values largely independent from the salt concentration.
Importantly, these patterns did not show significant impurity flowthrough in the window identified for efficient main peak elution,
thus indicating a high selectivity of the new resin.
The combination of the contour plots provided the basis to
identify the ranges of adequate conditions for each step of the

purification process. Thus, optimal conditions for antibody binding were found at acidic conditions around pH 5 and low ionic
strength, at which part of HMW could be simultaneously removed.
As a next step, a shift to basic pH 8-9 at the same conductivity
appeared appropriate to remove LMW and HCP fractions, while
the antibody was still strongly bound. The weak acid to neutral

3.4. Bind and elute column purification process for a highly
hydrophobic bsAb using Eshmuno® CMX
Based on the Kp screen results, a procedure for an Eshmuno®
CMX column run was tested selecting the following bind and elute
mode. The resin was equilibrated at pH 5.0 and low ionic strength.
Loading under these conditions should result in efficient capturing of the antibody. Thereafter, incubation at pH 5.5 intended to
overcome the buffer capacity of the resin. Following a shift to pH
9.0, a second washing step aimed to eliminate LMW and HCP from
the still tightly bound antibody. A third wash step at pH 6.1 and
slightly increased conductivity (100 mM sodium sulfate) served to
prepare elution of the product that was done with 300 mM salt.
Due to the assumed impact of pH, runs at two different elution pH
values, pH 6.0 (run 1) and pH 6.2 (run 2), were performed. The
protocol resulted in the elution profiles shown in Fig. 6A. The respective purification process is documented by analytical SE-HPLC
6


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 5. Kp screen contour plots of the highly hydrophobic MAB A on Eshmuno® CMX displaying the flow-through of total protein, host cell protein (HCP), high molecular
weight (HMW) impurities, main peak, and low molecular weight (LMW) impurities.


Fig. 6. Bind and elute purification process of the highly hydrophobic MAB A on Eshmuno® CMX. A and C) Elution profiles of run 1 (elution pH 6.0) and run 2 (elution
pH 6.2). B and D) overlays of the analytical size exclusion chromatography performed for run 1 and run 2. Colour coding: black = load, blue = wash 1, pink = wash 2,
green = eluate, brown = post eluate.

7


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 7. CE-SDS data obtained for the bind and elute purification process of the highly hydrophobic MAB A in two runs using non-reduced conditions. Colour coding:
black = load, blue = wash 1, pink = wash 2, brown = eluate, green = post-eluate
Table 1
Removal of product-related and process-related impurities by Eshmuno® CMX column runs.
Step
definition

IgG
(g/L)

Native purity by SE-HPLC (Sum in %)
HMW

Main peak

LMW

1


15.8
5.9
6.4
0.6

10.5
1.2
2.3
14.2

83.6
98.2
97.2
84.8

5.9
0.6
0.5
1.1

Load
Run 1: eluate elution pH 6.0
Run 2: eluate elution pH 6.2
Run 2: fraction post end pooling criteria

HCP
(ng/mg)

DNA
(pg/mg)


Step yield (%)

Purity by CE-SDS (Sum in %)

12353
2.3
3.3
31

5759
1367
1375
4168

64.9
71.2
7.5

HMW

Main peak

LMW

0.07

94.3
97.4
97.2
93.2

5.6
2.6
2.8
6.8

1
Data obtained from conditioned Protein A eluate pH 5.0 after depth and sterile filtration (to remove precipitates)HCP: host cell proteins; HMW: high molecular
weight; LMW: low molecular weight; LoD: limit of detection.

data for flow-through samples collected from each process step
that indicate successful impurity separation regarding one HMW
pool and three LMW peaks from eluted main peak (Fig. 6B). Additional analysis performed under denaturing conditions using CESDS demonstrated an expected increase in the LMW portion detecting nine different types of small fragments, thus confirming a
complex profile of LMW impurities (Fig. 7A and B). In contrast, the
HMW species had almost completely disappeared in comparison to
native conditions (Fig. 6A).
Quantitative evaluation (Table 1) revealed that the static Kp
screen process evaluated by contour plots yielded expedient results
that were transferable to a dynamic column process. Obtained data
reinforced high selectivity of Eshmuno® CMX by efficient removal
of LMW (10-fold reduction of native LMW to levels ≤ 0.6%) and
HMW (5-fold reduction of native HMW to levels ≤ 2.3%) impurity fractions, resulting in the desired increase of main peak purity
reaching levels of > 97% with a recovery rate in the range of 70%.
Similarly, the reduction of process-related HCP was very successful
(> log 3 scale) reaching values of 2-3 ng/mg protein which were
close to the detection limit of the assay. The results for elution pH

6.0 and pH 6.2 showed minor differences, which may suggest an
impact of the small pH shift, as deduced from the Kp screen data.
Thus, pH 6.0 resulted in lower yield but higher purity, whereas pH

6.2 provided higher yield but lower purity regarding slightly increased HMW and HCP levels.
Since the Kp screens give an overview of the binding conditions
and provide a window of operation, protocols other than the presented ones might be feasible for successful column runs. Especially, in the elution window alternative conditions to that finally
chosen might be suitable for similar purification results. At pH 6,
efficient elution of pure main peak appears also possible with salt
concentration ranges higher than 300 mM. However, increased salt
concentrations are usually regarded as unfavorable, as they may
negatively impact following process steps as well as the sustainability of the process and in case of halides (e.g. chloride) also the
resistance of the stainless-steel process equipment .
Overall, the purification outcome obtained with Eshmuno®
CMX together with the results for HMW, LMW and HCP meet
the requirements for impurity removal and are in the range that
can be expected for the application of a two-step protocol with
CEX and HIC. However, the combination of two columns is usually disadvantageous considering overall effort and product yield.
Thus, the new MMC resin providing high-quality purification for
the highly hydrophobic bsAb in a single column step offers the
potential to simplify downstream processing. Only host cell DNA
remained at a moderate level, which in our view was not sur8


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig. 8. Kp screen contour plots of MAB B on Eshmuno® CMX displaying the flow-through of total protein, host cell protein (HCP), high molecular weight (HMW) impurities,
main peak, and low molecular weight (LMW) impurities.

prising since DNA dissociation from the product usually requires
strong acidic pH and was not expected under the conditions used
in the present study (>pH 5). This shortcoming may be counterbalanced by using a successive standard flow-through AEX polishing at high capacity and product recovery (around 95%) [20], that
may combine DNA removal with efficient virus reduction in one
step.

The third molecule investigated was MAB C, a glycoengineered
mAb with a standard format and moderate relative hydrophobicity of 0.33 displaying a more usual impurity profile at a lower
level (Fig. 1). In a similar manner as found for the two bsAbs,
efficient protein binding was detected in the low salt range and
elution seemed feasible at conditions of higher salt and pH 6 or
above (Fig. 9). HMW and LMW reduction occurred under nearly
the same conditions, mainly at acidic pH and at low salt up to pH
8, where the antibody was tightly bound. Therefore, loading at pH
5 and subsequent washing at a pH 8 seemed suitable to remove
significant fractions of HMW and LMW together with HCP. Favorable elution conditions appeared around 400 mM salt and pH 6.
Again, a lower HCP content could be expected at lower pH values.
The results showed that the binding properties of this mAb on Eshmuno® CMX had changed in comparison to CaptoTM MMC ImpRes
(Fig. 3). The reduced hydrophobic interaction of Eshmuno® CMX
resulted in a consistent and wide transition zone with increasing
salt concentrations and, thus, may constitute a practical alternative
to resins with a stronger hydrophobic component as well as to the
application of pure cation exchange resins.
In view of the presented examples, the application of Eshmuno® CMX appeared feasible in a similar bind and elute mode
offering high selectivity for efficient purification of the two complex bispecific antibody formats as well as of the standard mAb.
In all three cases, the binding window was found in the range of
pH 5 without salt addition, where simultaneously HMW fractions
could be washed out. A shift to highly basic conditions served for
impurity removal. The optimal pH level for this step depended on

the pI of the antibody, which determined the range in which the
antibody was still strongly bound due to cationic interaction. Thus,
for MAB B having a pI of 8.0 the wash step was suitable at pH 7.5,
whereas for MAB A with a pI of 9.4 the wash step could be ex-

3.5. General applicability of Eshmuno® CMX for the purification of
hydrophobic molecules
The applicability of Eshmuno® CMX for the purification of
hydrophobic antibodies was further evaluated by performing Kp
screens with other molecules, such as MAB B, a bsAb of a symmetric 1+1 format with a high relative hydrophobicity of 0.58.
Among the LMW components, the impurity profile included a ¾
antibody fragment (Fig. 1). Again, obtained contour plots revealed
good loading conditions for protein binding in the range of pH 5
and low salt, while favorable elution conditions could be deduced
in the range of pH 5-6 and elevated salt concentrations higher
than 350 mM (Fig. 8). HMW behavior was similar as observed for
MAB A, where flow-through appeared under acidic conditions. The
major difference appeared for LMW flow-through, which was only
low at basic conditions but increased around pH 5 and 150 mM
salt. Therefore, loading at pH 5 and subsequent washing appeared
suitable to partially remove HMW and LMW fractions. As strong
antibody binding occurred up to pH 7.5, a washing step at this pH
level was useful to reduce further LMW in addition to a part of
the HCP fraction. In terms of the elution window in the range of
pH 5 to pH 6, more acidic conditions seem to provide the possibility for higher product purity due to lower HCP content, however,
most probably at the cost of lower recovery.
9


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Fig 9. Kp screen contour plots of MAB C on Eshmuno® CMX displaying the flow-through of total protein, host cell protein (HCP), high molecular weight (HMW) impurities,
main peak, and low molecular weight (LMW) impurities.

ecuted at pH 9.0. Especially, HCP could be removed in part with
this washing step at elevated pH and low salt concentration because all three antibodies showed sufficient binding under these
conditions. In contrast, strong HCP binding occurred in the acidic
pH range largely independent from the salt concentration, while
the final HCP content in the elution pool was mainly influenced
by the elution pH. Type, amount, and binding behavior of LMW
were more specific for the individual antibody, so that their removal occurred under different conditions. Kp screens identified
best conditions for elution in a window of pH 5-6 with salt concentrations around 300 mM and above. This window fell in a wide
transition zone of gradual binding differences indicating that elution conditions regarding optimal selectivity and product recovery
can be carefully adapted and balanced between pH and salt to establish reproducible and robust column processes.
Overall, the results obtained for Eshmuno® CMX indicate a
generic approach for a bind and elute mode with particular benefit for efficient purification of highly hydrophobic molecules and
especially bsAbs with complex formats. The single unit operation
step, moreover, offers the option of replacing a two-column procedure using CEX and HIC. As the mAbs under study range from
moderate up to extremely high hydrophobicity (relative hydrophobicity of 0.33- 0.74) the novel MMC resin appears suitable to cover
a wide range of antibodies of different hydrophobicity grade. This
may include standard antibodies with low hydrophobicity lacking
strong binding, so that the new MMC resin may be also applied in
a flow-through mode.

moderate hydrophobic properties like the novel MMC Eshmuno®
CMX. The combination of weak CEX and moderate hydrophobic interactions allows for robust operation under standard pH and salt
conditions during binding, washing, and final antibody elution, resulting in high product purity and yield. The successful purification of three antibodies with widely different formats, impurity
profiles, and hydrophobicities showed the high versatility of Eshmuno® CMX. In view of its favorable separation properties and

high selectivity, the resin can also provide an alternative approach
for further purification challenges, such as the separation of glycosylation variants and antibody drug conjugates with different drug
to antibody ratios. Moreover, by combining characteristics of CEX
and HIC in a single unit operation, implementation of the novel
resin provides the opportunity for simplification and cost-saving
during downstream processing. Integration into a platform purification approach for mAb manufacturing is therefore favored.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.
CRediT authorship contribution statement
Wolfgang Koehnlein: Conceptualization, Methodology, Formal
analysis, Writing – original draft. Annika Holzgreve: Investigation,
Writing – original draft. Klaus Schwendner: Investigation. Romas
Skudas: Supervision, Funding acquisition, Writing – review & editing. Florian Schelter: Conceptualization, Supervision, Funding acquisition, Writing – review & editing.

4. Conclusions
The present study demonstrated the effective purification of
highly hydrophobic complex antibody formats with a resin with
10


W. Koehnlein, A. Holzgreve, K. Schwendner et al.

Journal of Chromatography A 1687 (2023) 463696

Data availability

[8] M. Surowka, W. Schaefer, C. Klein, Ten years in the making: application of
CrossMAb technology for the development of therapeutic bispecific antibodies
and antibody fusion proteins, MAbs 13 (2021) 1967714, doi:10.1080/19420862.

2021.1967714.
[9] S.W. Chen, W. Zhang, Current trends and challenges in the downstream purification of bispecific antibodies, Antib. Ther. 4 (2021) 73–88, doi:10.1093/abt/
tbab007.
[10] A.M. Merchant, Z. Zhu, J.Q. Yuan, A. Goddard, C.W. Adams, L.G. Presta, P. Carter,
An efficient route to human bispecific IgG, Nat. Biotechnol. 16 (1998) 677–681,
doi:10.1038/nbt0798-677.
[11] J.B. Ridgway, L.G. Presta, P. Carter, Knobs-into-holes’ engineering of antibody
CH3 domains for heavy chain heterodimerization, Protein Eng. 9 (1996) 617–
621, doi:10.1093/protein/9.7.617.
[12] P. Gagnon, Technology trends in antibody purification, J. Chromatogr. A 20
(2012) 1221 57-70, doi:10.1016/j.chroma.2011.10.034.
[13] J. Tang, X. Zhang, T. Chen, Y. Wang, Y. Li, Removal of half antibody, hole-hole
homodimer and aggregates during bispecific antibody purification using MMC
ImpRes mixed-mode chromatography, Protein Expr. Purif. 167 (2020) 105529,
doi:10.1016/j.pep.2019.105529.
[14] V. Halan, S. Maity, R. Bhambure, A.S. Rathore, Multimodal chromatography for
purification of biotherapeutics - a review, Curr. Protein Pept. Sci. 20 (2019) 4–
13, doi:10.2174/1389203718666171020103559.
[15] R.A. Chmielowski, S. Meissner, D. Roush, T.O. Linden, E. Glowacki, J. Konietzko, J. Nti-Gyabaah, Resolution of heterogeneous charged antibody aggregates
via multimodal chromatography: a comparison to conventional approaches,
Biotechnol. Prog. 30 (2014) 636–645, doi:10.1002/btpr.1908.
[16] Y.F. Lee, S. Kluters, M. Hillmann, T. von Hirschheydt, C. Frech, Modeling of
bispecific antibody elution in mixed-mode cation-exchange chromatography, J.
Sep. Sci. 40 (2017) 3632–3645, doi:10.10 02/jssc.20170 0313.
[17] X. Santarelli, C. Cabanne, Mixed mode chromatography: a novel way toward new selectivity, Curr. Protein Pept. Sci. 20 (2019) 14–21, doi:10.2174/
1389203718666171024121137.
[18] P. McDonald, B. Tran, C.R. Williams, M. Wong, T. Zhao, B.D. Kelley, P. Lester,
The rapid identification of elution conditions for therapeutic antibodies from
cation-exchange chromatography resins using high-throughput screening, J.
Chromatogr. A 1433 (2016) 66–74, doi:10.1016/j.chroma.2015.12.071.

[19] M. Haverick, S. Mengisen, M. Shameem, A. Ambrogelly, Separation of mAbs
molecular variants by analytical hydrophobic interaction chromatography
HPLC: overview and applications, MAbs 6 (2014) 852–858, doi:10.4161/mabs.
28693.
[20] J. Weaver, S.M. Husson, L. Murphy, SR. Wickramasinghe, Anion exchange membrane adsorbers for flow-through polishing steps: Part II. Virus, host cell protein, DNA clearance, and antibody recovery, Biotechnol. Bioeng. 110 (2013)
500–510, doi:10.1002/bit.24724.

The authors do not have permission to share data.
Acknowledgments
The authors thank Alisha Koch, Roxana Disela, Svetlana Melnik, Hakyoung Kim, Katja Schubert, Karin Reindl and Alexander
Kurtenbach for expert technical support. We also want to thank
Raena Morley for proof reading of the manuscript and Marlena
Surowka for providing the molecule pictograms. In addition, we
thank Heidemarie Peuker (BISs - Biomedical Investigation Services)
for diligent preparation of the manuscript. This research did not
receive any specific grant from funding agencies in the public,
commercial, or not-for-profit sectors.
References
[1] U. Brinkmann, R.E. Kontermann, The making of bispecific antibodies, MAbs 9
(2017) 182–212, doi:10.1080/19420862.2016.1268307.
[2] R.E. Kontermann, Dual targeting strategies with bispecific antibodies, MAbs 4
(2012) 182–197, doi:10.4161/mabs.4.2.190 0 0.
[3] S. Dengl, K. Mayer, F. Bormann, H. Duerr, E. Hoffmann, B. Nussbaum,
M. Tischler, M. Wagner, A. Kuglstatter, L. Leibrock, C. Buldun, G. Georges,
U. Brinkmann, Format chain exchange (FORCE) for high-throughput generation
of bispecific antibodies in combinatorial binder-format matrices, Nat. Commun.
11 (2020) 4974, doi:10.1038/s41467- 020- 18477- 7.
[4] A.F. Labrijn, M.L. Janmaat, J.M. Reichert, P. Parren, Bispecific antibodies: a
mechanistic review of the pipeline, Nat. Rev. Drug. Discov. 18 (2019) 585–608,
doi:10.1038/s41573- 019- 0028- 1.

[5] C. Klein, C. Sustmann, M. Thomas, K. Stubenrauch, R. Croasdale, J. Schanzer,
U. Brinkmann, H. Kettenberger, J.T. Regula, W. Schaefer, Progress in overcoming
the chain association issue in bispecific heterodimeric IgG antibodies, MAbs 4
(2012) 653–663, doi:10.4161/mabs.21379.
[6] W. Schaefer, J.T. Regula, M. Bähner, J. Schanzer, R. Croasdale, H. Dürr,
C. Gassner, G. Georges, H. Kettenberger, S. Imhof-Jung, M. Schwaiger,
K.G. Stubenrauch, C. Sustmann, M. Thomas, W. Scheuer, C. Klein, Immunoglobulin domain crossover as a generic approach for the production of bispecific IgG antibodies, Proc. Natl. Acad. Sci. U. S. A. 108 (2011) 11187–11192,
doi:10.1073/pnas.1019002108.
[7] C. Klein, W. Schaefer, J.T. Regula, The use of CrossMAb technology for the
generation of bi- and multispecific antibodies, MAbs 8 (2016) 1010–1020,
doi:10.1080/19420862.2016.1197457.

11