Immobilized metal affinity chromatography optimization for poly-histidine tagged proteins

Journal of Chromatography A 1629 (2020) 461505

Contents lists available at ScienceDirect

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

Immobilized metal affinity chromatography optimization for
poly-histidine tagged proteins
Valeria Riguero a,∗, Robert Clifford a, Michael Dawley e, Matthew Dickson d,
Benjamin Gastfriend f, Christopher Thompson c, Sheau-Chiann Wang b, Ellen O’Connor a,∗
a

Purification Process Sciences, AstraZeneca, One MedImmune Way, Gaithersburg, MD 20878, USA
Analytical Biotechnology, AstraZeneca, One MedImmune Way, Gaithersburg, MD 20878, USA
Data Science and Modelling, AstraZeneca, One MedImmune Way, Gaithersburg, MD, 20878, USA
d
Texcell North America, 4991 New Design Road, Suite 100, Frederick, MD, 21703, USA
e
Quality Engineering and Validation, Genentech, 1000 New Horizons Way, Vacaville, CA, 95688, USA
f
Department of Chemical and Biological Engineering, University of Wisconsin-Madison, Madison, WI, USA
b
c

a r t i c l e

i n f o

Article history:
Received 15 July 2020

Revised 18 August 2020
Accepted 20 August 2020
Available online 21 August 2020
Keywords:
Poly-histidine tag
Protein purification
IMAC
Resin/metal screening
Process optimization

a b s t r a c t
Immobilized metal affinity chromatography (IMAC) is a technique primarily used in research and development laboratories to purify proteins containing engineered histidine tags. Although this type of chromatography is commonly used, it can be problematic as differing combinations of resins and metal chelators can result in highly variable chromatographic performance and product quality results. To generate a
robust IMAC purification process, the binding differences of resin and metal chelator combinations were
studied by generating breakthrough curves with a poly-histidine tagged bispecific protein. The optimal
binding combination was statistically analyzed to determine the impact of chromatographic parameters
on the operation. Additionally, equilibrium uptake isotherms were created to further elucidate the impact of chromatographic parameters on the binding of protein. It was found that for protein expressed
in CHO cells, Millipore Sigma’s Fractogel EMD Chelate (M) charged with Zn2+ and GE’s pre-charged Ni
Sepharose Excel displayed the highest binding capacities. When the protein was expressed in HEK-293,
GE’s IMAC Sepharose 6 Fast Flow charged with either Co2+ or Zn2+ bound the greatest amount of protein.
The study further identified the metal binding capacity of the resin lot, the protein capacity to which the
resin is loaded, and the ratio of poly-histidine tag residues on the protein all impacted the chromatographic performance and product quality. These findings enabled the development of a robust and scalable process. The CHO expressed cell culture product was directly loaded at a high capacity onto variable
metal binding affinity Fractogel EMD Chelate (M). A 250 mM imidazole elution condition ensured the
product contained monomeric 4 and 6-histidine tagged bispecific proteins. The optimized IMAC process
conditions determined in this study can be applied to a wide variety of poly-histidine tagged proteins
in research and development laboratories as various poly-histidine tagged proteins of differing molecular
weights and formats expressed in either HEK-293 or CHO cells were successfully purified.
© 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY license. ( />
1. Introduction
Affinity chromatography is a preferred method to capture proteins from cell culture harvest. It utilizes a highly selective ligand

to bind a protein of interest. For example, Fc-containing proteins,
such as monoclonal antibodies, are easily captured using Protein

∗
Corresponding authors at: Purification Process Sciences, BioPharmaceutical Development, AstraZeneca, One MedImmune Way, MD 20878, Gaithersburg, USA.
E-mail addresses: (V. Riguero),
ellen.o’ (E. O’Connor).

A resin. Many other protein formats lack readily available affinity capture modality and, therefore, require tags for affinity capture. One such modification, the histidine tag, is a key tool used in
biopharmaceutical research and development. For this approach, a
poly-histidine tag is incorporated on a protein during protein engineering. The resulting tagged protein can be captured from the
cell culture harvest using immobilized metal affinity chromatography (IMAC) through specific binding of the poly-histidine tag to the
metal chelates present in the IMAC media. While IMAC is a widely
used technology in research and development laboratories, process
optimization and manufacturing operational challenges exist which

/>0021-9673/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license. ( />

2

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

can limit its use. One challenge is the need for clarification and diafiltration of the cell culture media prior to purification. This additional unit operation to enable chromatography increases the overall process time and decreases yield [1]. IMAC resins also have relatively low binding capacities compared to other chromatography
resins and variability in resin selectivity, both resulting in additional cycles required to complete purifications [2,3]. Additionally,
the need to strip metal and recharge the resin after each cycle adds
processing time. These challenges all contribute to IMAC being a
lower yielding and slow purification operation.
Many innovative immobilized metal binding methods are in development to overcome the challenges with the goal of streamlining purification of poly-histidine tagged molecules. One current development area focuses on directly applying a cell culture suspension containing the tagged protein of interest to the
IMAC resin without the need for clarification. Studies have found
that chromatography performed with large diameter IMAC agarose

beads allowed for the passage of cells while capturing the protein
[4]. Other studies have used charged and chelated detergent micelles to extract tagged proteins directly from cell culture supernatant [5]. Nanoparticle technologies are also being explored to enable direct purification from cell lysate. One group has developed
nanoparticles with a core functionalized with pentadentate chelate
affinity ligand that can be chelated with a variety of metals. The
charged particles, which can be loaded to a high capacity with cell
lysate containing protein, were magnetically separated and protein
was eluted at high purity and yield [6,7]. Another group examined
silica nanospheres containing dual chelating groups, which separated the proteins from cell lysate [8].
Another development area focuses on identifying novel chelating ligands. Groups have developed 1,4,7-Triazacyclononane (TACN)
and 5,5-dithiobis-(2-nitrobenzoic) acid (DTNP) ligands that have
been coordinated with Cu2+ or Ni2+ [2,9–13]. The development
of a new generation of chelating ligands is advantageous for several reasons. Compared to currently available ligands, these ligands
can increase the resin binding capacity to make the purification
more efficient. The novel ligands can also increase the selectivity
of the resin to elute a product of increased purity. The ligands also
have high ion stability so that the new generation resins do not
need to be recharged every cycle. Novel chelating ligands are also
being applied to non-column separation methods. Polyelectrolytes
that chelate metal ions have been applied to membrane adsorbers,
binding proteins at equal capacity to traditional resin beads, with
the added benefit of a 15 min purification time [14]. Monolith convective interaction adsorbents have also been developed, enabling
proteins purified to high purity levels at high binding capacities
[3,15,16]. The advantage of the monolith technology is the fast flow
rate compared to traditional media. Liquid-liquid extraction techniques are also being explored. A TACN liquid sorbent was developed that can coordinate with Cu2+ , Ni2+ , or Zn2+ to create an
ionic liquid sorbent [17]. Aptamer technology has also been developed to separate histidine tagged proteins. The aptamer complex
approach is advantageous because imidazole is not necessary for
purification and the purified protein is more pure compared to traditional resin based technology [18].
Although new IMAC technologies are being developed to overcome the challenges of currently available purification methods,
academic and industrial purification still relies on traditional resinbased processes. Therefore, it is useful to understand and better
control this widely used technology. Studies have contributed to

the current knowledge of resin based IMAC. In one study, IMAC
resins containing different chelators, iminodiacetic acid (IDA) and
nitrilotriacetic acid (NTA) were coordinated with nickel and evaluated for purification of 6-histidine tagged cytotoxin associated
gene A (CagA). Results were equivalent on all resins regardless of
the chelator that was used [19]. Another study observed a dif-

ference in the purification product resulting from IDA and NTA
chelators charged with different metals. It was found that Cu2+
charged IDA yielded higher purity 6-histidine tagged viral coat protein product compared to other metal-chelator combinations [20].
In a closer examination of IDA ligand mediating purification, four
different sorbents; HiTrap Chelating HP, TSK Cheltate-5PW, Poros
20 MC, and monolithic glycidyl methacrylate–ethylene dimethacrylate were studied. TSK Chelate-5PW bound the strongest while
Poros matrix had a high degree of non-specific binding. Agarosebased columns showed high selectivity and specificity [17].
In this study, a robust and scalable IMAC process was developed for the purification of a poly-histidine tagged bispecific protein. First, a resin and metal chelator screening study was performed to select conditions that would allow for direct protein
binding from the cell culture harvest onto a charged resin at a
both a high dynamic binding capacity and flow rate. In the presence of HEK-293 medium, the optimal resin and metal combination was IMAC 6 Sepharose FF charged with either Zn2+ or Co2+ .
Optimal combinations in the presence of CHO conditioned medium
occurred on Fractogel EMD Chelate (M) charged with Zn2+ and
Ni Sepharose Excel. Second, a statistical study was performed to
understand the impact of operational parameters on IMAC performance and product quality. It was found that the metal binding
capacity of the resin lot and the extent to which the resin was
loaded heavily impacted the chromatographic performance and resulting product quality. Finally, equilibrium isotherms were generated to understand the binding behavior of the 4 and 6-histidine
tagged materials comprising the poly-histidine tagged bispecific
protein. Although equivalent maximum adsorption capacities were
independently achieved, when combined, 6-histidine tagged bispecific protein competed with and displaced 4-histidine tagged bispecific protein. The three studies resulted in the development of a
controlled and scalable IMAC purification operation. The developed
process conditions described herein were successfully applied to
other poly-histidine tagged molecules.
2. Materials and methods
2.1. Materials and equipment

2.1.1. Materials
FreeStyleTM MAX 293 and FreeStyleTM MAX CHO Expression
systems were purchased from Thermo Fisher (Waltham, MA, USA).
Fractogel EMD Chelate (M) was purchased from MilliporeSigma
(Billerica, MA, USA). IMAC Sepharose 6 Fast Flow, Ni Sepharose
Excel, and Capto Q were purchased from GE Healthcare (Uppsala,
Sweden). Profinity IMAC, Profinity IMAC Ni-Charged, and CHT Type
II were purchased from Bio-Rad (Hercules, CA, USA). Ni-NTA Superflow was purchased from Qiagen (Hilden, Germany). TOSOH TSK
gel 30 0 0 was purchased from Tosoh (Minato, Japan). The ProPacTM
WCX-10 BioLCTM column, NuncTM 96-Well Cap Mats, Plate Sealers
and 3.5 kDa MWCO Slide-A-LyzerTM Cassettes were obtained from
Thermo Scientific (Waltham, MA, USA). The reverse phase column
Intrada, WP-RP, 3 μm (4.6 × 250mm) was purchased from Imtakt,
USA (Portland, OR, USA). AcroPrepTM Advance 96-well filter-plates
with 0.45 μm polypropylene membranes were purchased from
Pall (Port Washington, NY, USA). iCE280 Analyzer and ChromPerfect software were purchased from ProteinSimple (San Jose, CA,
USA). All other chemicals including metal salts and imidazole were
purchased from Avantor Performance Chemicals (Center Valley, PA,
USA).
2.1.2. Equipment
For the null cell culture growth, shake flasks were grown using the Multitron incubation shaker from Infors HT (Bottmingen,
Switzerland). For bench scale purification experiments, resins were


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

packed in Tricorn columns purchased from GE Healthcare (Uppsala,
Sweden) or Vantage columns purchased from MilliporeSigma (Billerica, MA, USA). Purification was controlled by Unicorn 6.0 software on an Äkta Avant purchased from GE Healthcare (Uppsala,
Sweden). For scale up chromatography, Fractogel EMD Chelate (M)
was packed in a BPG column purchased from GE Healthcare (Uppsala, Sweden). Purification was controlled by the GE Akta Prime

(Uppsala, Sweden). For statistical analysis, Jmp version 15 software
was purchased from SAS (Cary, NC, USA). The analog rocker was
from VWR (Radnor, PA, USA). The SorvallTM LegendTM RT+ Centrifuge and Cimarec+ TM stir plates were purchased from Thermo
Scientific (Waltham, MA, USA). Protein quantification was determined using NanoDropTM 20 0 0 from Thermo Scientific (Wilmington, DE, USA). High performance liquid chromatography (HPLC)
systems were purchased from Agilent Technologies (Santa Clara,
CA, USA).
2.2. Purification methods
2.2.1. Null cell culture
In order to evaluate the cell culture expression systems typically
found in research and early development settings, transient transfections for both HEK-293 and CHO cell lines were performed using Thermo Fisher’s FreeStyleTM MAX Expression System. The cells
were thawed, expanded, and grown per the FreeStyleTM MAX protocols. At transfection, the cells were transfected with non-coding
plasmid DNA at 1 mg of DNA/L of cells. The transfected cells were
grown in 5L fed batch shake flasks for 10 days, allowing for typical growth conditions and host cell by-products to be present in
the cell culture harvest media. The shake flasks were harvested by
centrifugation at 40 0 0× g for 30 min. The null cell culture media
was then 0.2 μm filtered.
2.2.2. IMAC resin screening
Resin and metal combination screening of the poly-histidine
tagged bispecific protein was performed for both purified and null
cell media (from Section 2.2.1) spiked materials. Tricorn columns
of 0.5 cm diameter were packed to a 10 cm bed height, 2 mL column volume (CV), with each resin listed in Table 1. For each binding experiment, resin was rinsed with 5 CVs of water, followed by
4 CVs of 250 mM metal charge solution. After charging, the resin
was rinsed with 5 CVs of water and then 5 CVs of 500 mM sodium
chloride. The resin was next equilibrated with 5 CVs of 20 mM
tris, 150 mM sodium chloride, pH 7.5 prior to loading purified protein, of 99% monomer, at a 1 mg/mL initial concentration either in
the equilibration buffer or in the null cell culture media. As the
protein was loaded, the flow-through material was tested for protein concentration by RPLC. The loaded resin was re-equilibrated
for 5 CVs prior to a step elution with 20 mM tris, 150 mM sodium
chloride, 500 mM imidazole, pH 7.5. Between cycles, the resin was
stripped with 50 mM EDTA, 500 mM sodium chloride. All chro-

3

matographic steps were performed at 300 cm/h (1 mL/min), which
corresponded to a residence time of 2 min. The dynamic binding
capacity was determined by plotting the protein concentration as a
fraction of the inlet feed concentration and normalized to the CV.
Linear interpolation of the resulting protein breakthrough data was
used to find the 10% breakthrough point.
2.2.3. Initial process development
Two different poly-histidine tagged bispecific protein cell culture lots were purified using two different Fractogel EMD Chelate
(M) resin lots. Cell culture lot 1 contained 85.0% 4-histidine tag and
12.4% 6-histidine tag, while lot 2 contained 76.3% 4-histidine tag
and 20.5% 6-histidine tag. Resin lots 1 and 2 respectively had metal
binding affinities of 71 and 97 μmol/mL for Cu2+ . Both resin lots
were packed in 1.1 cm Millipore Vantage L columns to bed heights
of 20 cm (19 mL CV). For each experiment, resin was rinsed with
5 CVs of water prior to charging with 4 CVs of 250 mM ZnCl2 .
The charged resin was then washed with 5 CVs of water prior to 5
CVs of 500 mM sodium chloride. The resin was then equilibrated
with 20 mM tris, 150 mM sodium chloride, pH 7.5 prior to loading the 1 mg/mL concentration conditioned media to 20 g/L capacity. After the load was complete, the protein loaded resin was
re-equilibrated for 5 CVs prior to a 20 CV elution gradient to 20
mM tris, 150 mM sodium chloride, 500 mM imidazole, pH 7.5. The
resin was regenerated by a 5 CV 50 mM EDTA, 500 mM sodium
chloride wash. All chromatography steps were performed at 300
cm/h (4.7 mL/min), corresponding to a 4 min residence time. Elution fractions were collected for analysis of the quantity of protein, the ratio of 4 and 6- histidine tag by cIEF, and the levels of
monomer and aggregate by HPSEC.
2.2.4. Statistical study
Jmp software was used to outline a full factorial experimental
design for the statistical study. The study tested the impact of 4

factors on IMAC performance: the Cu2+ binding capacity of the
Fractogel EMD Chelate resin (from 71 μmol/mL to 97 μmol/mL),
the amount of ZnCl2 charged onto the resin (4 CVs of 18 mM to
250 mM), the amount of protein loaded per volume of resin (10 to
30 g/L), and the ratio of 4 to 6-histidine tagged bispecific protein
present in the cell culture material (4-histidine 76.3%, 6-histidine
20.5% and 4-histidine 85.0%, 6-histidine 12.4%). Each resin lot was
packed in a 1.1 cm diameter Millipore Vantage L column to a bed
height of 20 cm (19 mL CV). For each of 18 experiments, including two center points, resin was rinsed with 5 CVs of water followed by charging with 4 CVs of either 18 or 250 mM ZnCl2 . The
resin was rinsed with 5 CVs of water followed by 5 CVs of 500
mM sodium chloride. After equilibration with 5 CVs of 20 mM tris,
150 mM sodium chloride, pH 7.5, the resin was loaded with to a
targeted capacity with either cell culture lot (both with titers of
1 mg/mL). After re-equilibration the protein was eluted using a 20
CV gradient to 20 mM tris, 150 mM sodium chloride, 500 mM imi-

Table 1
Commercially available IMAC resins. IMAC resins tested in the screening study are listed along with their backbone composition and ligand molecule. The particle sizes, pore
diameters and metal binding capacity of the resins are also shown.

Resin

Backbone

Ligand

Particle
Size (μm)

Pore Diameter

(nm)

Metal Affinity Capacity
(μmol/mL)

Fractogel EMD Chelate (M) (Merck-Millipore)
IMAC Sepharose 6 Fast Flow (GE)
Profinity IMAC (Bio-Rad)
Ni-NTA Superflow (Qiagen)
Ni Sepharose Excel (GE)
Profinity IMAC Ni-Charged (Bio-Rad)

Cross-linked polymeth-acrylate
Agarose
UNOsphere
Cross-linked agarose
Agarose
UNOsphere

IDA
IDA
IDA
NTA
NTA
IDA

40–90
45–165
45–90
60–160

90
45–90

80a
30b
130b
23c
30b
130b

60–100
25
12–30
12
54–70
12–30

a
b
c

MilliporeSigma product information.
Carta, Giorgio and Jungbauer, Alois. Protein Chromatography. Weinheim: Wiley-VCH, 2010.
Kastner, Michael. Protein Liquid Chromatography. New York: Elsevier, 20 0 0.


4

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

dazole, pH 7.5. Between cycles, the resin was stripped with 50 mM
EDTA, 500 mM sodium chloride. All chromatography steps were
performed at 300 cm/h (4.7 mL/min). Elution fractions were collected for analysis of the quantity of protein, the ratio of 4 and 6histidine tag by isoelectric chromatography (IEC), and the levels of
monomer and aggregate by HPSEC. Data was analyzed using Jmp
software. Least squares fit analyses were used to create models for
the prominent elution peaks based on the combination of significant factors.
2.2.5. Equilibrium binding experiments
Separate monomeric 4-histidine and 6-histidine tagged bispecific protein solutions were formulated at 5 mg/mL in equilibration buffer, 20 mM tris, 150 mM sodium chloride, pH 7.5 containing either 5, 100, or 170 mM imidazole. For each imidazole concentration, the two stock solutions were mixed together to create
5 protein solutions with different ratios of the 4 and 6-histidine
tagged bispecific proteins. Each of the 5 different ratio solutions
were then diluted using either equilibration buffer containing imidazole or null CHO cell culture medium (from Section 2.2.1) into 4
protein concentrations ranging from 2 mg/mL to 0.5 mg/mL.
Three lots of Fractogel EMD Chelate (M) resin with Cu2+ binding capacities of 71, 83, and 94 μmol/mL were obtained. Slurries
of the three resin lots were prepared in suspension. Preparation in
suspension proceeded as follows. First, resin was rinsed with 5 volumes of water with gentle mixing. Centrifugation was performed
at 500× g for 5 min to remove the rinse. Four volumes of 250 mM
ZnCl2 were next applied to the resin with gentle mixing. Centrifugation was used to remove the charge solution. Five CVs of water and 500 mM sodium chloride were then respectively applied,
mixed, and removed from the resin by centrifugation. Each slurry
was then prepared at a 10 % concentration in 20 mM tris, 150 mM
sodium chloride, pH 7.5.
Each well of a 96-well 0.45 μm filter plate was loaded with
100 μl of charged equilibrated slurry of the desired Cu2+ binding
capacity. After gentle mixing the equilibration buffer solution was
removed from each well via centrifugation of the plate at 500×
g for 5 min. 200 μL of equilibration buffer containing either 5,
100, or 170 mM imidazole was then added to the charged resin.
After gentle mixing, the new equilibration buffer solution was removed via centrifugation of the plate at 500× g for 5 min. After
the resin was prepared, 200 μL aliquots of histidine tagged bispecific protein solutions varying in protein concentration (load capacity), imidazole concentration, and histidine tag ratio were applied
to desired wells. The filter plate was then incubated overnight with
agitation to ensure protein binding. Once incubation was complete, the filter plate was centrifuged at 500× g, and the filtrate

of each well was collected. 50 μL of the equilibration buffer containing either 5, 100, or 170 mM imidazole was then added to each
well and the plate was incubated for 1 hour. After incubation, the
plate was centrifuged at 500× g and the centrifugal filtrate was
collected. The two filtrate solutions containing non-bound protein
were then pooled together. 200 μL elution buffer, 20 mM tris, 150
mM sodium chloride, 500 mM imidazole, pH 7.5 was then added
to each well and the plate was incubated for 4 hours. After incubation in elution buffer, the plate was centrifuged at 500× g and the
centrifugal filtrate (the eluate) was collected in a microplate. Each
filtrate containing non-bound or eluted materials was analyzed for
protein concentration and quantity of 4 and 6-histidine tagged bispecific proteins using IEC.
2.2.6. Impact of competition on binding capacity
A 0.5 cm diameter tricorn column was packed to a 10 cm bed
height, 2 mL column volume (CV), with Sepharose 6 Fast Flow
resin. For each experiment, the column was rinsed with 5 CVs of

water, followed by 4 CVs of 250 mM Nickel Chloride charge solution. After charging, the resin was rinsed with 5 CVs of water,
followed by 5 CVs of 500 mM sodium chloride, and equilibrated
with 5CVs of 20 mM tris, 150 mM sodium chloride, pH 7.5.
For the first experiment, the column was loaded with purified
monomeric poly-histidine tagged bispecific protein at a 1 mg/mL
concentration in 20 mM tris, 150 mM sodium chloride, pH 7.5.
In the second experiment, the column was loaded with equivalent
protein spiked to 1 mg/mL concentration in null CHO medium. For
the third experiment, the column was washed with 100 CV of null
CHO medium and then re-equilibrated with 5 CV of 20 mM tris,
150 mM sodium chloride, pH 7.5. The purified monomeric polyhistidine tagged bispecific protein was then loaded onto the column at a 1 mg/mL concentration in the equilibration buffer. In the
fourth experiment, the purified protein was spiked into null CHO
medium and the resulting product was then dialyzed (40 times the
product volume) against 20 mM tris, 150 mM sodium chloride, pH
7.5. The dialysis buffer was replaced 4 times with new buffer after

the material was dialyzed with mixing for a minimum of 4 hours
at 2-8ºC. The resulting dialyzed product was then loaded onto the
charged Sepharose 6 Fast Flow column at a 1 mg/mL concentration.
For all experiments, as the protein was loaded, the flow-through
material was tested for protein concentration by RPLC. The loaded
resin was re-equilibrated for 5 CVs prior to a step elution with 20
mM tris, 150 mM sodium chloride, 500 mM imidazole, pH 7.5. Between cycles, the resin was stripped with 50 mM EDTA, 500 mM
sodium chloride. All chromatographic steps were performed at 300
cm/h (1 mL/min), which corresponded to a residence time of 2
min. The dynamic binding capacity was determined by plotting
the protein concentration as a fraction of the inlet feed concentration and normalized to the CV. Linear interpolation of the resulting
protein breakthrough data was used to find the 10% breakthrough
point.
2.2.7. Process scale-up
Fractogel EMD Chelate (M) resin was packed in a 20 cm diameter column to a bed height of 20 cm resulting in a 6.3 liter CV.
The resin was rinsed with 5 CVs of water, prior to application of
the 4 CVs of 250 mM ZnCl2 . After charging, the resin was rinsed
with 5 CVs of water and then 5 CVs of 500 mM sodium chloride.
The resin was next equilibrated with 5 CVs of 20 mM tris, 150 mM
sodium chloride, pH 7.5, prior to loading conditioned media at a
1 g/L initial concentration spiked with 5 mM imidazole to a load
of 18 g/L. The loaded resin was re-equilibrated with 5 CVs of 20
mM tris, 150 mM sodium chloride, pH 7.5. The column was then
washed with 5 CVs of 20 mM tris, 150 mM sodium chloride, 75
mM imidazole, pH 7.5. Product was then eluted with 20 mM tris,
150 mM sodium chloride, 250 mM imidazole, pH 7.5. The resin was
stripped with 5 CVs of 20 mM tris, 150 mM sodium chloride, 500
mM imidazole followed by 5 CVs of 50 mM EDTA, 500 mM sodium
chloride. All chromatographic steps were performed at 300 cm/h
(1570 mL/min). This operation was cycled four times to process the

500L reactor product.
2.2.8. Application to other poly-histidine tagged proteins
The optimized IMAC operation was applied to a variety of polyhistidine tagged molecules. Proteins expressed in HEK-293 cells
were purified on Sepharose 6 Fast Flow charged with 250 mM
CoCl2 . 5H2 0 or 250 mM ZnCl2. Proteins expressed in CHO cells were
purified using Fractogel EMD Chelate (M) charged with 250 mM
ZnCl2 or pre-charged Ni Sepharose Excel. For each purification, regardless of scale, a column was packed to a height of 20 cm and
operated at 300 cm/hr. Each purification included rinsing the column with 5 CVs of water prior to application of 4 CVs of the metal
charge solution. The charged column was then rinsed with 5 CVs
of water followed by 5 CVs of 500 mM sodium chloride. Prior to


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

5

Fig. 1. Protein quantitation by RPLC. A typical RPLC profile is shown. The UV trace of the protein is displayed in blue. The red line corresponds to the percentage of ACN/0.1%
TFA elution buffer. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

loading the column was equilibrated with 5 CVs of 20 mM tris, 150
mM sodium chloride, pH 7.5 buffer. The cell culture harvest containing the poly-histidine tagged protein was then loaded on the
charged and equilibrated column at capacities between 10 and 20
g/L. After re-equilibration, the protein was eluted with 20 mM tris,
150 mM sodium chloride, 250 mM imidazole. Column regeneration
proceeded with 5 CVs of 20 mM tris, 150 mM sodium chloride, 500
mM imidazole pH 7.5 and 5 CVs of 50 mM EDTA, 500 mM sodium
chloride.
2.3. Analytical methods
2.3.1. RPLC titer assay
A protein concentration standard curve was generated using

pure material by injecting samples onto an Intrada, WP-RP, 3μm
column (4.6 × 250 mm). Each sample was eluted with ACN/0.1%
TFA at a flow rate of 0.75 mL/min. Eluted protein was detected using UV absorbance at 280 nm. The protein concentration of each
sample was determined using regression analysis from the standard curve of the area under the product peak. An example of an
RPLC profile is shown in Fig. 1.
2.3.2. HPSEC purity assay
Each sample was injected onto a TSKgel G30 0 0SWXL column
(7.8 mm × 300 mm, 5 μm) and eluted with 0.1 M disodium phosphate containing 0.1 M sodium sulfate, pH 6.8, at a flow rate of
1.0 mL/min. Eluted protein was detected using UV absorbance at
280 nm. The results were reported as the area percent of the
product monomer peak compared to all other peaks excluding
the buffer-related peak observed in the blank run. Peaks eluting
earlier than the monomer peak were recorded as percent aggregate; high molecular weight (HMW) or dimer. Peaks eluting after
the monomer peak were recorded as percent fragment. A typical
HPSEC profile is shown in Fig. 2.
2.3.3. Capillary isoelectric focusing (cIEF) assay
Samples were concentrated to 2.0 mg/mL with Ultra Pure water using a Microcon 10,0 0 0 filter. Samples were adjusted to 0.25
mg/mL with Ultra Pure water, 1% methylcellulose solution, Pharmalyte pH 3-10, Pharmalyte pH 8-10.5, pI Marker 9.77, and pI
Marker 5.85. The samples were loaded onto an iCE280 Analyzer

and focused at 1500 V for 1 min, followed by 30 0 0 V for 6 min.
The resulting electropherograms were analyzed using ChromPerfect software and compared to a reference standard. A representative electropherogram showing the migration of the 4 and 6histidine tagged bispecific proteins is shown in Fig. 3.

2.3.4. Ion exchange chromatography (IEC) assay
To quantify the levels of 4 and 6-histidine tagged bispecific proteins in samples, analytical ion exchange chromatography was performed. One milligram of each sample was loaded at 1 mL/min
onto a ProPac WCX-10 (4 × 250 mm) analytical high-performance
weak cation exchange chromatography column equilibrated with
20 mM phosphate pH 7. The protein was then eluted with an increasing salt gradient of 20 mM phosphate, 100 mM sodium chloride pH 7 buffer. Eluted protein was detected using fluorescence
detector set to an excitation at 280 nm and an emission of 350
nm. The identity of each peak was determined by comparison to a

known reference standard containing 4,5, and 6-histidine species.
The percent peak area for each histidine tagged species was calculated by dividing the peak area by the total peak area. An example
of the IEC separation of 4 and 6-histidine tagged bispecific proteins
is shown in Fig. 4.

Fig. 2. Protein purity by HPSEC. A typical HPSEC profile is shown. The UV trace
of the protein is displayed in blue. The high molecular weight species elute first,
followed by dimer, and then monomer. (For interpretation of the references to color
in this figure legend, the reader is referred to the web version of this article.)


6

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

Fig. 3. cIEF profiles of histidine tagged variants. A typical isoelectric focusing profile is shown. The acidic and basic pI markers are labeled with their corresponding pI values.
The 4 and 6-histidine tagged bispecific protein peaks are also labeled with their pI values.

Fig. 4. High throughput quantitation of histidine tagged variants by IEC. A typical analytical ion exchange chromatography profile is shown. The fluorescence trace of the
protein is displayed in blue and the elution gradient is displayed in red. The 4-histidine tagged bispecific protein elutes first followed by the 6-histidine tagged bispecific
protein. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3. Results and discussion
3.1. Protein description
In this study, we sought to understand and optimize IMAC
to enable a controlled large-scale purification of a poly-histidine
tagged bispecific protein. The 52 kDa protein, shown in Fig. 5, consisted of two binding domains. The antigen binding domain was
engineered to bind to an oncology target, while the T cell binding region was designed to interact with the glycoprotein surface
complex. To enable IMAC purification, a poly-histidine tag was engineered into the terminus of the T cell binding domain. Prior to
large-scale purification of the molecule, several studies were performed and are described within.

3.2. IMAC resin screening of purified protein
A screening study was performed to select IMAC materials
that optimized the dynamic binding capacity of the poly-histidine
tagged bispecific protein. The study consisted of generating breakthrough curves using resin and metal chelator combinations of the
commercially available materials shown in Tables 1 and 2. As can
be seen in Table 1, studied resins included those having a variety of backbones and both IDA and NTA ligands. The particle
sizes composing each resin are also shown and ranged between
40 to 165 μm which enabled the superficial velocity of 300 cm/h
without the generation of pressure challenges at bench and large

Fig. 5. Diagram of the poly-histidine tagged bispecific protein. The protein consists
of two single chain variable fragment domains (scFv). The antigen binding and T
cell binding domains are shown for the 52 kDa protein. The 4-6 residues of the
terminal histidine tag are also depicted.


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

7

Fig. 6. Dynamic binding capacities of purified protein. The binding data of purified bispecific protein on commercially available resin/chelator combinations are shown. 6A
displays the protein’s breakthrough profiles when all resins were charged with Ni2+ . 6B shows the impact of the metal chelator on the breakthrough profiles. 6C includes
the dynamic binding capacities at 10% breakthrough for various resin and metal combination.

scales. The pore sizes and metal affinity capacities are also detailed.
Table 2 lists the studied metal chelators and the corresponding
salts used to generate the charged buffers.
The breakthrough curves of purified bispecific protein on resins
charged with Ni2+ are shown in Fig. 6A. The dynamic binding
capacities for each resin differ. The Profinity resins exhibited the

steepest binding profiles and the least capacities at 10% breakthrough. In contrast, shallow curves were observed using the GE
resins which also achieved greater than 50 mg/mL capacities at
10% breakthrough.
The dynamic binding capacities of the resins were next determined when charged with various metals. As shown in Fig. 6B,
regardless of the metal used, the resin binding order is consistent with that of Fig. 6A with IMAC Sepharose 6 FF > Fractogel
EMD Chelate (M) > Profinity IMAC. The GE resin, IMAC Sepharose
6 FF, displayed the highest capacity at 10% breakthrough for each
Table 2
Metal chelators. The metals and corresponding salts
used for charging the commercially available resins for
the screening study are shown.
Metal

Salt

Nickel
Zinc
Copper (II)
Cobalt
Iron (II)
Magnesium
Manganese

NiCl2 . 6H2 O
ZnCl2
CuSO4 . 5H2 O
CoCl2 . 5H2 O
FeCl2 . 4H2 O
MgCl2 . 6H2 O
MnCl2 . 4H2 O

charged metal. Charging of this resin with Ni2+ , Co2+ , Zn2+ , or
Cu2+ , each resulted in higher capacities at 10% breakthrough than
those of the Ni2+ pre-charged resins. Differences were also seen
in the binding capacity for all resins depending on the metal used
for charging. Resins charged with Mg2+ , Mn2+ , and Fe2+ , showed
lower binding capacity compared to resins charged with Ni2+ ,
Co2+ , Zn2+ , and Cu2+ , and suggested that these metals were not
as effective at coordinating with the chelating ligand and histidine tagged bispecific protein. An examination of the breakthrough
profiles on IMAC 6 Sepharose FF in Fig. 6C revealed that charging
with Co2+ , Zn2+ , and Cu2+ resulted in preferable binding, as breakthrough curves generated with the other metals showed steep
slopes or lower capacities at 10% breakthrough. Overall, the breakthrough curves of the purified poly-histidine tagged bispecific protein showed that IMAC Sepharose 6 FF resin charged with Cu2+
provided the most favorable binding profile and highest dynamic
binding capacity.
It is interesting to note that regardless of the metal chelator,
the dynamic binding capacities at 10% breakthrough were consistently ranked IMAC Sepharose 6 FF > Fractogel EMD Chelate (M)
> Profinity IMAC. Several factors govern dynamic binding capacity; residence time, particle size, and pore diffusivity [21]. As described in section 3.1, the molecule used in the study was a 52 kDa
protein with a calculated hydrodynamic radius of 3.22 nm [22]. A
closer examination of the resin specifications in Table 1 showed
that the pore sizes of the resins tested were theoretically adequate
to allow diffusion of the bispecific protein into the resin pores. Diffusion was likely as the protein was at least 5 times smaller than
the pore sizes of the Qiagen and GE resins, the resins with the


8

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

Fig. 7. Dynamic binding capacities of protein spiked in null cell media. The binding capacities of the protein spiked in null cell media is shown. 7A and 7B respectively
display the dynamic binding capacities at 10% breakthrough for bispecific protein purified on a variety of resin and metal combinations from HEK-293 and CHO media.

smallest pores. Therefore, for each resin, the protein could access
all the available pore volume for coordination.
Additionally, as Table 1 shows, there were differences in the
metal binding capacities of the resins as specified by their manufacturers. Fractogel EMD Chelate (M) possessed the greatest metal
capacity, 60–100 μmol metal/mL. The GE resins, Ni Sepharose Excel and IMAC 6 Sepharose FF, could be respectively charged to 54
–70 and 25 μmol metal/mL. Ni-NTA Superflow and Profinity IMAC
possessed the least metal binding capacity of 12–30 μmol/mL. The
data in Fig. 6 showed that the highest dynamic binding capacities were observed on the IMAC Sepharose 6 FF resin that was
composed of particles with small pores and a moderate level of
metal binding capacity. Ni-NTA resin was composed of particles
with similar pore size and decreased metal binding capacity and
demonstrated reduced protein dynamic binding capacity. The pore
size of Ni Sepharose Excel resin was also comparable to IMAC
Sepharose 6 FF and Ni-NTA. However, the metal binding capacity of Ni Sepharose Excel exceeded IMAC Sepharose 6 FF, but also
demonstrated reduced protein dynamic binding capacity. It is likely
that that protein binding created steric hindrance, effectively blocking additional proteins from accessing the additional coordination
sites. A level of optimal ligand density existed, between 25 to 54
μmol metal/mL, the metal binding capacities of IMAC Sepharose 6
FF and Ni Sepharose Excel resins.
3.3. IMAC resin screening of protein in CHO and HEK-293 media
IMAC is typically used as a capture chromatography operation.
Therefore, experiments were performed to identify the impact of
the cell culture expression material on dynamic binding capacity.
To mimic the capture step, breakthrough profiles were generated
for the poly-histidine tagged bispecific protein in null media. The
dynamic binding capacities at 10% breakthrough for the protein respectively spiked into HEK-293 and CHO null media are shown in
Fig. 7A and B. As expected, in both media the binding capacities
were decreased for all resin and metal combinations compared to
those previously generated with pure protein.

A closer examination of the dynamic binding capacity data
shows that compared to the purified protein values, the HEK-293
medium had a greater impact on binding than the CHO medium.
For example, both Ni Sepharose Excel and IMAC 6 Sepharose FF
charged with Ni2+ , Co2+ , Zn2+ , and Cu2+ had greater dynamic
binding capacities in CHO than HEK-293 null cell medium. Greater
capacities in CHO medium were also observed for Fractogel EMD
Chelate (M) when charged with Zn2+ and Cu2+ . Furthermore, this

data demonstrated the dynamic binding capacity results for the
purified protein did not extrapolate when in the presence of media, as the best overall binding capacities in media, exceeding 15
g/L, were on Fractogel EMD Chelate (M) charged with Zn2+ and
Cu2+ or pre-charged Ni Sepharose Excel.
The dynamic binding capacities in the presence of null conditioned media were decreased compared to those of the purified
protein as the protein competed with cell expression products or
media components capable of coordination to the charged metal.
The data indicated that there were components in HEK-293 null
conditioned medium that more effectively competed with the bispecific protein for the charged metal than components in the CHO
conditioned medium, as the capacities in CHO medium were generally greater. It is also interesting to note that unlike the case for
the pure protein, the highest capacities observed in CHO medium
were on resins containing the highest metal binding capacities,
Fractogel EMD Chelate (M) charged with Zn2+ and Ni Sepharose
Excel. The availability of the increased metal binding sites likely
increased the total quantity of both bound media components and
protein. The high levels of binding to these resins was not observed
in the HEK-293 medium. The differing components in the HEK-293
medium may have had increased affinity for Zn2+ and Ni2+ .
These experiments were impactful for purification development
of the poly-histidine tagged bispecific protein because they showed
that IMAC resin and metal combinations existed that achieved high

dynamic binding capacities at 10% breakthrough without requiring
diafiltration of the cell culture harvest media. The ideal combinations included IMAC Sepharose 6 FF charged with Co2+ , Zn2+ , or
Cu2+ when purifying a protein from HEK-293 medium and either
pre-charged Ni Sepharose Excel or Zn2+ charged Fractogel EMD
Chelate (M) when purifying a protein from CHO medium.
3.4. Initial process development
The screening study results were applied to the development
of the IMAC capture operation for the poly-histidine tagged bispecific protein. The expression system for the cell culture process
used CHO cells, therefore Ni Sepharose Excel and Zn2+ charged
Fractogel EMD Chelate (M) were considered as capture columns as
the protein could be directly loaded onto the resin at high capacity without the need for diafiltration. Fractogel EMD Chelate (M)
was selected for the process as its integration decreased purification costs.
Process development of the IMAC capture operation was initiated by performing gradient elutions to aide in the identification


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

9

Fig. 8. Variable IMAC gradient elution profiles. Variable IMAC elution profiles are shown for two different lots of Fractogel EMD Chelate (M) resin and two different cell
culture lots. The elution profiles of cell culture lot 1/resin lot 1, cell culture lot 2/resin lot 1, and cell culture lot 1/resin lot 2 are respectively shown in 8A, 8B, and 8C. The
blue traces indicate A280 levels, while the red trace displays the progress of the elution gradient. In each chromatogram, all three elution peaks are labeled along with their
corresponding imidazole elution concentrations. 8D, 8E, and 8F show the milligram levels of protein in terms of percent monomer purity and ratio of histidine tag in the
load, peak 1, peak 2, peak 3, and combined peaks. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

of step elution conditions. As described in Section 2.2.3, two lots
of Fractogel EMD Chelate (M) resin differing in Cu2+ binding abilities, 71 and 94 μmol/mL, were used to purify the poly-histidine
tagged bispecific protein expressed in two cell culture lots. Prior
to purification of both lots, the ratio of 4-histidine tagged and 6histidine tagged bispecific proteins was measured. Cell culture lot
1 contained 85.0% 4-histidine tagged and 12.4% 6-histidine tagged

bispecific proteins, while lot 2 contained 76.3% 4-histidine tagged
and 20.5% 6-histidine tagged bispecific proteins. Three experiments
of differing harvest and resin lot combinations were performed in
which 300 milligrams of harvested cell culture product was loaded
to a 20 g/L capacity on the IMAC resin prior to a 20 CV imidazole
gradient elution. Although the same procedures were used to perform the purifications, inconsistent chromatography profiles were
observed during the elution gradients. Examples of the protein elution profiles are shown in Fig. 8. All chromatographic profiles contained three prominent peaks, however the peaks had different
peak areas and eluted at different points of the imidazole gradient.
For example, as shown in Fig. 8A and B, protein from two different cell culture lots were purified using the same 71 μmol/mL lot
of IMAC resin. The first elution peak is more prominent in Fig. 8A
(cell culture lot 1) whereas the second peak is more prominent in
Fig. 8B (cell culture lot 2). Cell culture lot 1 was also purified on
the 94 μmol/mL IMAC resin lot. Three elution peaks were also ob-

served, however they eluted with increased imidazole concentrations. The findings of these experiments show that variation in cell
culture and resin lot properties impact protein elution behavior.
Due to the observed chromatographic inconsistencies, additional characterizations of the elution peak materials were performed. cIEF, HPSEC, and protein concentration analyses were performed for the three prominent peaks of each elution shown in
Fig. 8A, B, and C. As shown in Fig. 8D, E, and F, each elution
peak within a gradient was not homogeneous in terms of polyhistidine tag length and monomer purity. cIEF data revealed that
across the three purifications, peak 1 contained predominately
4-histidine tagged bispecific protein, peak 2 contained predominately 6-histidine tagged bispecific protein, and peak 3 contained
a mixture of poly-histidine tagged bispecific proteins, with the 6histidine tagged variant in greater proportion. For example, of the
255 mg of 4-histidine tagged bispecific protein loaded in Fig. 8A,
217 mg of the material eluted in peak 1, whereas 3 mg eluted in
both peaks 2 and 3. In the same experiment 37 mg of 6-histidine
tagged bispecific protein was loaded onto the resin. No significant amount of 6-histidine tagged material was found in peak 1,
whereas 19 and 15 mg were each respectively found in peaks 2
and 3. The same trends held true for the chromatography experiments shown in Fig. 8B and C. Additionally, a decrease in monomer
was observed across the gradients as peaks 1 and 2 contained

10

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505
Table 3
Statistical study parameters. Parameters were identified that impacted IMAC performance and product quality. The parameter ranges were studied
in a range finding study as shown. The products of each purification were analyzed for quality by the listed assays.
Parameters

Parameter ranges

Ratio of 4-histidine:6-histidine tagged bispecific protein in cell culture lot

76.3 : 20.5

85.0 : 12.4

Copper binding ability of Fractogel EMD Chelate (M) (μmol/mL)
Resin Charging (mM Zinc)
Load Capacity (g/L)

71
18
10

97
250
30

HPSEC Purity

Histidine Tagged Ratio

Outputs for Each Peak
Protein Concentration

mostly monomer, while peak 3 contained mostly aggregate. For example, analysis of the chromatography performed in Fig. 8A shows
that of the 228 mg of monomer loaded onto the resin, 199 mg of
monomer (87.3% yield) was recovered in elution peaks of 1 and
2. The monomer purity of peaks 1 and 2 were respectively 84.2%
(15.8% dimer) and 71.3% (28.7% dimer). Peak 3 was enriched high
molecular weight aggregate, with an aggregate level of 84.6%. The
same monomer purity level trends held true across the three chromatography experiments. The findings of these experiments show
that 6-histidine tagged and aggregated protein are likely to interact with more charged sites than 4-histidine tagged or monomeric
species as increased levels of imidazole are required to elute these
proteins. The same chromatographic trends were also observed
when the Fractogel EMD Chelate (M) resin was charged with other
metals, including Cu2+ , Co2+ , and Ni2+ (data not shown) as the
poly-histidine tagged bispecific protein interacted with the charged
ligand in a similar manner. The trends are likely to be similar on
other resins, however the separation would be impacted by the
metal binding capacity of the resin. These findings demonstrate
that variations in the metal binding capacity of an IMAC resin,
cell culture product quality, both monomeric purity and quantity of
differing histidine tagged levels, complicate development of IMAC
capture operations.
3.5. Statistical analysis of IMAC operation
The results of the three chromatography experiments performed during the initial purification development showed the
Cu2+ binding capacity of the IMAC resin as well as variation of the
histidine tag ratio resulted in inconsistent chromatographic performance and product quality. A statistical study was employed

to both determine if additional parameters affect the purification
performance, and to understand the impact of possible ranges of
the parameters. The operating parameters studied are highlighted
in Table 3. The poly-histidine bispecific protein materials used in
this study were the same as those used in the initial chromatography work. Cell culture lot 1 contained 85.0% 4-histidine tagged
and 12.4% 6-histidine tagged bispecific protein, while lot 2 contained 76.3% 4-histidine tagged and 20.5% 6-histidine tagged bispecific proteins. The resin used in the study also matched that used
in the initial development work and spanned the range of Cu2+
binding capacities produced by the resin manufacturer, 71 and 97
μmol/mL. In addition to the histidine tag ratio and the resin lot,
the extent of resin metal charging and the protein load capacity
were also explored. Resin metal charging solutions ranged from 18
mM to 250 mM as recommended by manufacturers’ protocols. Protein loading ranged from 10 to 30 g/L to observe the impact on
product quality as the resin is pushed in excess of 10% DBC. Each
chromatographic experiment in the statistical study was performed
as described in Section 2.2.4. Materials from each of the three characteristic gradient peaks were separately analyzed for protein concentration, HPSEC purity, and histidine tag ratio.

Required Imidazole
Concentration

The study identified the Cu2+ binding capacity of the resin and
the amount of protein loaded on the column were the two most
important factors that influenced the performance and product
quality. The contour plots displaying the impact of two parameters is shown in Fig. 9. As shown in Fig 9A and D, when protein
was loaded to 10 g/L onto 97 μmol/mL resin peaks 1 and 2 eluted
with higher concentrations of imidazole compared to when loaded
to 30 g/L on 71 μmol/mL resin. The earlier strong binding condition respectively required 145 mM and 230 mM of imidazole to
elute peaks 1 and 2. The later weak binding condition respectively
required 110 and 180 mM of imidazole to elute peaks 1 and 2.
The Cu2+ binding capacity and the amount of protein loaded on
the column also influenced the ratios of 4 and 6-histidine tagged

bispecific proteins in elution peaks 1 and 2. As shown in Fig. 9B
and E, when protein was loaded to 10 g/L onto 97 μmol/mL resin,
the histidine tagged ratio within each peak is more homogeneous.
For example, under the strong binding condition, the peak 1 elution was greater than 99% 4-histidine tagged and peak 2 contained
material greater than 87% 6-histidine tagged. This is in contrast to
weak binding conditions in which 98% of the peak 1 product was
4-histidine tagged and 72% was 6-histidine tagged.
Peaks 1 and 2 were also analyzed in the statistical study for
monomeric content. As shown in Fig. 9C and F, the monomer content in the elution products was influenced by both the Cu2+ binding capacity of the resin and the amount of protein loaded onto the
column. The monomer content of peak 1 material was as high as
88% when protein was loaded to 10 g/L on the 97 μmol/mL resin.
The purity decreased to 76% under the weak binding condition of
30 g/L binding on 71 μmol/mL resin. The same trends were observed for peak 2 material in which the monomer purity ranged
from 74% to 66%.
The trends observed in the study are logical. Fully charged
97 μmol/mL resin had the greatest possible number of coordination sites. Loading that lot of charged resin to the low 10 g/L capacity, allowed for the maximum interactions between the polyhistidine tagged bispecific protein with the resin. Elution from this
state required the maximum amount of imidazole and resulted
in relatively homogeneous histidine tagged material with a high
monomer percentage. Fully charged 71 μmol/mL resin loaded to
30 g/L had less coordination sites available. It is possible that these
weak binding conditions created an environment in which the 6histidine tagged bispecific protein competed with the 4-histidine
tagged material and the HMW aggregate competed with the 6histidine tagged bispecific protein. The elution products therefore
contained increased aggregate and an increased heterogeneity in
terms of 4 and 6-histidine tagged bispecific proteins.
The results of the statistical study showed that IMAC performance and product quality were influenced by process parameters.
For example, to ease the action of performing the purification, any
lot of Cu2+ binding capacity resin could be obtained from the manufacturer and fully charged with four CVs of 18 mM to 250 mM
metal chelator. The charged resin could be loaded to between 10

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

11

Fig. 9. Statistical study of IMAC gradient elution. Contour plots containing data from the statistical study are shown. The impact of Fractogel EMD Chelate (M) resin metal
charge capacity and protein loading on the imidazole required for elution of peaks 1 and 2 is shown in 9A and 9D. The impact of the same two parameters is shown on the
ratio of histidine tag in products of peaks 1 and 2 is shown in 9B and 9E. The monomer purities of the same materials are shown in 9C and 9F.

to 30 g/L prior to elution with 250 mM imidazole. As shown in
this study, the 250 mM imidazole elution allowed for the elution
of both 4 and 6-histidine tagged bispecific proteins that were enriched for monomer.
3.6. Modeling equilibrium isotherms
The statistical study revealed differences in the binding and elution patterns of the poly-histidine tagged bispecific protein which
resulted from the Cu2+ binding capacity of the resin and the
resin protein loading. The observations in the statistical study were
based on two cell culture lots. One lot was composed of 85.0%
4-histidine tagged and 12.4% 6-histidine tagged bispecific protein,
while the other lot contained 76.3% 4-histidine tagged and 20.5% 6histidine tagged bispecific proteins. Prior to finalizing the IMAC purification operation for the protein, it was desired to further study
the impact of a wider range of both 4 and 6-histidine tagged bispecific protein ratios. Confirmation of the binding behavior of pure
monomeric 4-histidine and 6-histidine tagged bispecific proteins as
well as a complete range of their ratios allowed for a thorough understanding of the operation so that any cell culture lot, regardless
of the histidine tagged ratio would be purified as expected.
Single-component equilibria binding experiments were first
performed with 100% 4-histidine tagged and 100% 6-histidine
tagged bispecific protein on 83 μmol/ml Cu2+ binding capacity
resin as described in Section 2.2.5. Adsorption isotherms were created for each protein solution in three mobile phases as determined based on the elution strength required to elute the two
variants in the statistical range finding experiments; 5 mM imidazole to prevent non-specific binding, 100 mM imidazole to elute
4-histidine tagged bispecific protein, and 170 mM imidazole to
elute 6-histidine tagged bispecific protein. The single-component
isotherms in Fig. 10A and B show both 6-histidine and 4-histidine

tagged bispecific proteins exhibit typical Langmuir isotherms. The
maximum adsorption capacities, qm , in the different mobile phases
are shown in Table 4. At 5 mM imidazole, the qm values for both
proteins were equivalent at 39.9 mg/mL. As imidazole increased in
the mobile phase to 100 mM, the adsorption equilibrium constant

(kL ) was decreased overall at 26.2 mg/mL. The adsorption equilibrium constant for 6-histidine tagged material was greater than the
4-histidine tagged material in both phases.
Adsorption isotherms were also created to study the impact
of a range of 6-histidine tagged and 4-histidine tagged ratios on
protein binding as the poly-histidine tagged bispecific protein harvest material contained an uncontrolled mixture of 4-histidine and
6-histidine tagged bispecific proteins. Shown in Fig. 10C and D,
equivalent qm values were observed for protein solutions composed of either 100% 4 or 6-histidine tagged materials, with kL6
greater than kL4 , respectively 7.99 and 3.28 mL/mg. As increasing percentages of the opposing molecule were introduced, lower
qm values were observed for the 4-histidine tagged material compared to the 6-histidine tagged material. For example, the qm of
4-histidine tagged bispecific protein loaded in a mixture of 25% 4histidine tagged and 75% 6-histidine tagged bispecific proteins approached 5 mg/mL, whereas that of the 6-histidine tagged material
loaded in a 25% 6-histidine tagged and 75% 4-histidine tagged mixture approached 12 mg/mL.
The results of the single-component and the multi-component
changing histidine tagged ratio experiments were used to generate a competitive Langmuir model that describes the competition of the 6-histidine and 4-histidine tagged bispecific proteins.
The model is outlined in Fig. 10E and the equation is shown below. qei represents milligrams of 4-histidine or 6-histidine tagged
bispecific protein (qe4 or qe6 ) adsorbed per milliliter of resin. cei
is equal to the concentration (mg/mL) of soluble 4-histidine (ce4 )
and 6-histidine tagged bispecific protein (ce6 ). qm represents the
maximum protein adsorbed per resin volume (mg/mL). kLi repreTable 4
Single component adsorption isotherm results. Isotherms were created for each protein solution in three mobile phases listed on 83 μmol/mL copper binding capacity
resin. The maximum adsorption capacities, qm , in the different mobile phases are
shown. Both the 4-histidine tagged and 6-histidine tagged bispecific proteins had
equivalent qm values.
Imidazole (mM)

5

100

170

qm (mg/mL)

39.95

26.18

0.65


12

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

Fig. 10. Equilibrium isotherms of 4 and 6-histidine tagged bispecific proteins. Isotherms were generated for purified 4 and 6-histidine tagged bispecific proteins in 5, 100,
and 170 mM imidazole mobile phases on 83 μmol/mL Cu2+ binding affinity Fractogel EMD Chelate (M) resin and are shown in 10A and 10B. The protein solutions were
also mixed into different ratios to examine the competition of the poly-histidine tagged bispecific proteins. The resulting isotherms are shown in 10C and 10D. Isotherms are
shown with experimental data points. The displacement of 4-histidine tagged bispecific protein by 6-histidine tagged bispecific protein is shown in 10E.

sents the Langmuir equilibrium coefficients for 4-histidine and 6histidine tagged bispecific protein, respectively kL4 and kL6 . The denominator is the sum of the k and c terms for the 4 and 6-histidine
tagged bispecific proteins.

qei =

1+

qm kLiCei
n
j=1 kL j ce j

The model predicts the displacement of 4-histidine tagged protein with increasing amounts of 6-histidine tagged protein. As
shown in Fig. 10, the model is consistent with the data points
generated from the single and multi-component isotherm experiments. The kL4 and kL6 values in conjunction with qm values from
both sets of experiments indicated that 4-histidine and 6-histidine
tagged materials competed for binding. In the 5 and 100 mM imidazole mobile phases, the 6-histidine tagged bispecific protein
bound more favorably than 4-histidine tagged bispecific protein,
achieving higher adsorbed protein concentrations at lower solution
protein concentrations. These finding showed that increased protein loading of material containing both 4-histidine and 6-histidine
tagged bispecific proteins can impact the histidine tag ratio in the
elution products. These observations are consistent with the sta-

tistical study data which showed displacements of the 4-histidine
and 6-histidine proteins at high loading capacities.
The model was implemented to generate equilibrium isotherms
to examine the influence of resins with different Cu2+ binding capacities. Fig. 11 displays the 4-histidine or 6-histidine tagged bispecific protein single-component isotherms along with overlaying
experimental points generated in mobile phases of 5, 100, and 170
mM imidazole on each of three resins with Cu2+ binding capacities of 71, 83, and 94 μmol/mL. As Table 5 details, in the 5 mM
imidazole environment which is most favorable for poly-histidine
tagged bispecific protein binding, both the 6 and 4-histidine tagged
material achieved similar qm values, 38, 40, and 39 mg/mL, irrespective of the binding ability of the resin. As the imidazole concentration was increased to 100 mM in the mobile phase, the qm
values respectively decreased to 25, 26, and 29 mg/mL on the 71,
83, and 94 μmol/mL Cu2+ binding capacity resins. In the 170 mM
imidazole mobile phase, the respective qm values were 0.2, 0.7, and
3.6 mg/mL for the same three resins. For each mobile phase, the
lowest qm values were seen for the 71 μmol/ml Cu2+ binding capacity resin. Greater qm values were observed as the resin Cu2+

binding capacities increased. Table 5 also includes kL4 and kL6 values from the set of experiments. The kL6 values were greater than


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

13

Fig. 11. Equilibrium isotherms on resins of different Cu2+ binding abilities. Equilibria isotherms were generated for 4 and 6-histidine tagged bispecific proteins on three
Fractogel EMD Chelate (M) resin lots in 5 mM, 100 mM, and 170 mM imidazole mobile phases respectively shown in 11A and 11B, 11C and 11D, and 11E and 11F. Grey, blue,
and orange respectively represent 94, 83, and 71 μmol/mL Cu2+ binding capacity resins. Experimental data points are shown. (For interpretation of the references to color
in this figure legend, the reader is referred to the web version of this article.)

the kL4 values under all mobile phase and resin conditions evaluated, consistent with the previous single component and multicomponent experimental observation used to construct the model.
Specifically, kL4 values were two to three times less than kL6 values
under conditions favorable for protein binding (5 mM imidazole)
and are more than ten times under protein eluting conditions (100
mM imidazole). This observation was consistent with the experimental column chromatography data that shows 4-histidine tagged
bispecific protein was more readily eluted from the resin than 6histidine tagged bispecific protein. It is interesting to note that the

kL4 appeared similar across the three resin lots tested, while kL6
appeared to increase with increased Cu2+ binding capacity under
low imidazole (high protein binding) conditions. This observation
is consistent with the trend seen in the Fig. 9 statistical design experiments where the highest protein resolution was seen with the
94 μmol/mL resin.
Additionally, experiments were performed to study the impact
of null cell conditioned media on the binding equilibria as IMAC is
typically used as a capture chromatography operation. Binding experiments were performed in the 5 mM imidazole mobile phase on


14

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

Table 5
Influence of resin on poly-histidine tagged bispecific protein binding. Single component isotherms were created for 4-histidine tagged and 6-histidine tagged bispecific
proteins in three different mobile phases: 5, 100, and 170 mM imidazole on three
resins of different copper binding ability: 71, 83, 94 μmol/mL. Corresponding qm
and k values are shown for both proteins.
Mobile phase: 5 mM Imidazole
Resin Cu2+ binding capacity (μmol/mL)

71

83

94

qm (mg/mL)
kL6 (mL/mg)
kL4 (mL/mg)

38.08
7.23
3.19

39.95
7.99
3.28

39.16

10.19
3.31

71
25.49
0.89
0.04

83
26.18
1.02
0.08

94
29.46
1.00
0.06

71
0.20
36.15
1.20

83
0.66
11.61
1.88

94
3.56

1.52
0.20

Mobile Phase: 100 mM Imidazole
Resin Cu2+ Binding Capacity (μmol/mL)
qm (mg/mL)
kL6 (mL/mg)
kL4 (mL/mg)
Mobile Phase: 170 mM Imidazole
Resin Cu2+ Binding Capacity (μmol/mL)
qm (mg/mL)
kL6 (mL/mg)
kL4 (mL/mg)

Table 6
Impact of conditioned media on protein binding. Single-component isotherms were
generated by performing binding experiments in the 5 mM imidazole mobile phase
on all three copper binding capacity resins both in the presence of and without null
conditioned medium. The resulting qm and k values are shown.
Resin: 71 μmol/mL Cu2+ binding capacity

qm (mg/mL)
kL6 (mL/mg)
kL4 (mL/mg)

Purified protein

CM Protein

38.08

7.23
3.19

32.90
6.46
2.24

Purified Protein
39.95
7.99
3.28

CM Protein
35.87
7.13
2.31

Purified Protein
39.16
10.19
3.31

CM Protein
36.40
10.20
3.02

Resin: 83 μmol/mL Cu2+ Binding Capacity
qm (mg/mL)
kL6 (mL/mg)

kL4 (mL/mg)
Resin: 94 μmol/mL Cu2+ Binding Capacity
qm (mg/mL)
kL6 (mL/mg)
kL4 (mL/mg)

all three Cu2+ binding capacity resins both in the presence of and
without null cell conditioned medium. Isotherms generated using
the model as well as overlaying experimental points are shown in
Fig. 12. The qm , kL4, and kL6 values generated from the isotherms
are shown in Table 6. The qm, kL4, and kL6 values were all decreased
in conditioned medium conditions when compared to the pure
protein conditions, consistent with the screening study trends in
Fig. 7. The decreased binding in the presence of medium was most
different on the lower affinity Cu2+ binding resin. The differences
between the kL4 , kL6, and qm values decreased for the purified proteins compared to the proteins in medium as the Cu2+ binding capacity increased. It is likely that a component in the medium or
byproduct of cell expression, such as host cell proteins, stripped
the metal charge or competed with the histidine tagged bispecific
proteins for binding to the charged ligands. As the 71 μmol/mL
Cu2+ affinity resin contained less binding sites than 83 and 94
μmol/mL Cu2+ affinity resin, the greatest difference between the
purified protein and medium spiked protein was observed.
Overall, the observation that the isotherms produced in this
study are consistent with the Langmuir model is significant. Previous studies by different groups created isotherms that found the

Langmuir model, in which one protein binds to one metal ion,
was not adequate. For example, while creating isotherms for cytochrome c binding to TSK-chelate resin, isotherms were consistent with the Langmuir model at low levels of copper coordinated
on the resin, as immobilized copper decreased the possibility for
multiple site interactions. As increasing metal was coordinated on
the resin, Langmuir-Freundlich isotherms were observed, indicating

simultaneous coordination to more than one ion [23]. The same
group later found that a single histidine residue was capable of coordinating to multiple charged sites [24]. Another group conducted
a study with a diverse set of proteins including lysozyme, ovalbumin, and pig albumin. They found that proteins were capable of coordinating to 13 ions on the resin [25]. Yet another group showed
extensive Langmuir-Freundlich isotherms in the study of lysozyme,
ovalbumin, BSA, conalbumin, and wheat germ agglutinin [26]. The
qm and kL values were impacted by ionic strength and pH [27].
A fourth group extensively studied IMAC binding and developed
their own metal affinity interaction model (MAIC) to account for
multiple coordination bonds between the protein and the charged
sites, low capacities on the IMAC adsorbents, and the concentration
of mobile phase modifiers. The MAIC model explained the binding
behavior of RNAse A, myoglobin, lysozyme, conalbumin, and ovalbumin [28,29]. These studies are consistent. The proteins used in
these studies contained natural histidine residues that exist across
the proteins. The studied molecules did not contain engineered
poly-histidine tags. Therefore, multiple surface exposed histidine
residues present across a protein can interact with several ions on
the highly charged resin. The resulting isotherms would therefore
exhibit non-Langmuir characteristics. These studies showed that
the qm and kL values of these curves are impacted by the concentration of imidazole in the mobile phases and the placement and
number of histidine residues across the molecules.
The bispecific protein used in this study contained an engineered poly-histidine tag. Other potential interacting residues were
removed. Therefore, only one site on the protein could interact
with the highly charged resin containing the maximum level of accessible ions. The generation of this one to one interaction explains
the Langmuir consistent isotherms of this study. Although the 4
and 6-histidine tagged bispecific proteins each had one region capable of interacting, the 6-histidine tagged bispecific protein bound
with a greater kL than the 4-histidine tagged bispecific protein. The
shift of the isotherms according to total charge and concentration
of imidazole in the mobile phase is consistent with those reported
in the discussed literature.
3.7. Impact of competition on binding capacity

The binding isotherm data from Section 3.6 was compared to
the resin screening data from Sections 3.2 and 3.3 to better understand binding behavior for the poly-histidine tagged bispecific
protein on Fractogel EMD Chelate (M) charged with zinc. Comparison of the qm values calculated from the isotherm experiments
to the maximum protein binding capacities observed during the
screening experiments revealed different binding values. The difference in maximum binding values between the purified protein
and the protein in the presence of null cell media was greater
in the screening experiments (Figs. 6 and 7) versus the isotherm
binding experiments (Fig. 12). In the screening study, the maximum binding capacity of the purified poly-histidine tagged bispecific protein was approximately 30 mg/mL greater than when in
the presence of null cell medium. In comparison, the difference
in maximum binding capacity between the purified protein and
that in the presence of null cell medium was significantly less, 7
mg/mL, for the isotherm binding experiments.
An examination of the experimental conditions revealed that
the differences in both the maximum binding values as well as


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

15

Fig. 12. Equilibrium isotherms for proteins incubated in null CHO cell medium. Equilibria isotherms were generated for 4 and 6-histidine tagged bispecific proteins on three
Fractogel EMD Chelate (M) lots of 71, 83, and 94 μmol/mL Cu2+ binding capacities. Respective isotherms are shown in 12A and 12B, 12C and 12D, and 12E and 12F. Grey
and orange respectively represent isotherms generated when the proteins were bound in the presence of null media and without null media. Experimental data points are
shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

the magnitude of the differences were attributed to the presence
and interactions of imidazole and null cell media. The resin screening experiments were performed in the absence of spiked imidazole. In contrast, the isotherm binding experiments mimicked the
developed process and were performed in the presence of imidazole. Imidazole was present in the developed process to prevent
the non-specific binding of media components or undesired cellular expression products to allow for the elution of a product with

increased purity. Without imidazole in the screening experiments,
the purified protein bound to the charged resin, achieving the 48
mg/mL binding capacity. In the presence of null cell media and absence of imidazole, the maximum binding value dramatically decreased to 21 mg/mL. The decrease was attributed to the competition of the protein with media components or host cell proteins
for binding to the resin. The presence of imidazole in the isotherm
binding experiments prevented the high level of binding of the pu-


16

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

Fig. 13. Impact of null cell medium on binding capacity. Dynamic binding capacity profiles were generated for the poly-histidine tagged bispecific protein on Ni2+ charged
Sepharose 6 Fast Flow. Orange triangles and red squares respectively represent the pure protein and the pure protein in CHO null cell medium. The green diamond represents
binding of the pure protein to resin washed with CHO null cell medium. The blue circle shows the pure protein dialyzed from CHO null cell medium to tris buffered saline.
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

rified poly-histidine tagged bispecific protein that was observed in
the screening experiments as it may have competed with the protein for binding. In this case, a qm value of 39 mg/mL was determined. In the presence of both null cell medium and imidazole,
the qm value was found to range from 32 to 36 mg/mL. In addition
to null cell medium and imidazole competing with the protein for
binding, it is probable that imidazole interacted with null cell components. The interaction of imidazole and null cell medium components may have allowed for increased protein to bind when compared to the null cell spiked resin screening condition.
A DBC study was performed to study the impact of null cell
medium components on poly-histidine tagged bispecific protein
binding to IMAC resins. The results of the binding study on Ni2+
charged Sepharose 6 Fast Flow resin are shown in Fig. 13. DBC profiles, performed with equilibration buffers that did not contain imidazole, were generated for the purified molecule, the molecule in
the presence of CHO null cell medium, and the molecule buffer
exchanged from the medium into buffer. A DBC curve was also
created for the purified poly-histidine tagged bispecific protein applied to the charged resin that was washed with null cell culture
medium prior to loading. The purified poly-histidine tagged bispecific protein achieved a 10% DBC of 57 mg/mL as neither imidazole nor null cell medium were present. As expected, a dramatic decrease in binding, to a 10% DBC of 12 mg/mL, was seen
when the protein was loaded in the presence of null cell medium

as components competed with the poly-histidine bispecific protein for binding. Interestingly, this DBC profile matches that of the
purified molecule applied to the null cell medium washed column, indicating that medium components interacted with a significant amount of binding sites, effectively blocking the binding of
the poly-histidine tagged bispecific protein. Multiple media components are likely responsible for the interaction as the 10% DBC
value for the null cell spiked molecule dialyzed into buffer is an
intermediate level. The buffer exchange removed small molecules,
less than 3.5 kDa, and retained larger host cell protein molecules.
Several possibilities can account for the small molecules resulting in decreased binding values in the presence of null cell media.
First, the metal could have been stripped from the charged ligand
by EDTA which is frequently found in cell culture media. The scavenging of metal would eliminate binding sites and result in a de-

creased binding ability. Second, it is possible that the medium contained components that competed with the histidine tag for the
charged ligand. These medium components included amino acids
that exist in zwitterionic form with a deprotonated carboxyl group
such as; L-Cystine, L-Glycine, L-glutamine, L-Isoleucine, L-Leucine,
L-Lysine, L-Methionine, L-Phenylalanine, L-Serine, L-Threonine, LTryptophan, L-Tyrosine, and L-Valine. Media can also contain free
L-Histidine in its anionic state. In addition to amino acids, salts
were abundant in the medium. The negative ions comprising the
salts included chloride, nitrate, sulfate, bicarbonate, and phosphate
groups that could have coordinated with the positively charged
ligand and competed with the histidine tag. The coordination of
free amino acids or negative ions to the charged ligand would decrease the available binding sites for the histidine tagged bispecific protein. Finally, it is also important to note that the cell culture medium also contained positively charged metal ions including zinc, cadmium, selenium, copper, barium, tin, and silver. These
metals could have coordinated with the unbound histidine tagged
bispecific protein, blocking the protein from being able to interact
with the charged resin. Any or the combination of all three events
could potentially decrease the binding of histidine tagged bispecific
proteins in cell culture media to IMAC resins.
3.8. Scale-up chromatography
The enhanced understanding of IMAC performance gained from
the three studies described in this paper were applied to a gram
scale purification of the poly-histidine bispecific protein. The goal

of the purification was to use IMAC to capture the protein directly
from the CHO cell culture harvest that contained, as shown in
Fig. 14, mostly 6-histidine tagged material and 70% monomer. The
chromatography operation was to be high yielding and include a
step elution that contained both 4-histidine and 6-histidine tagged
monomeric proteins.
Fractogel EMD Chelate (M) was selected for the scale-up purification because, as determined in the described studies, it was able
to bind CHO expressed poly-histidine tagged bispecific protein directly from the harvested cell culture product in a cost effective
manner. One lot of this resin, with 83 μmol/mL Cu2+ binding capacity, was obtained and packed in a 20 cm diameter column to


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

17

Fig. 14. Scale-up IMAC purification. The bispecific protein was purifed using Fractogel EMD Chelate (M) at the gram scale. 14A shows the chromatographic profile of the
purification. The A280, pH, conductivity, and pressure measurments are displayed. 14B shows the mass values for the load and product in terms of histidine tag length and
monomer purity.
Table 7
Application of IMAC protocol to other molecules. Various histidine tagged molecules expressed in either CHO or HEK-293
cells were purified on either Fractogel EMD Chelate (M) or IMAC Sepharose 6 Fast Flow resin. The yields and purities of
the elution products for each purification are listed.
Histidine tagged protein

Molecular weight (KDa)

Expression cell line

Yield (%)

Monomer purity (%)

Fab #1
Fab #2
Fab #3
Fab #4
Fab #5
Fab #6
Fab #7
Fab #8
Hormone
Fusion #1
Fusion #2
Exotoxin
Receptor
Receptor
Bispecific T Cell Engager #1
Bispecific T Cell Engager #2
Bispecific T Cell Engager #2

50
50
50
50
50
50
50
50
6
120

71
39
136
150
52
52
52

HEK-293
CHO
CHO
CHO
CHO
CHO
CHO
CHO
HEK-293
CHO
HEK-293
CHO
CHO
HEK-293
HEK-293
HEK-293
HEK-293

72
79
85
70

75
72
68
73
60
78
80
75
80
82
85
84
78

95
82
76
81
91
82
79
83
86
75
78
89
74
73
92
88

74

a bed height of 20 cm. The column dimensions were consistent
with the geometry of the statistical study, resulting in a 6.5 liter
CV. Based on the size of the column and the quantity of protein
expressed, the load capacity was expected to be 18 g/L per cycle.
In order to ensure the step elution would elute the totality of
the 4-histidine and 6-histidine tagged bispecific monomeric proteins at the same ratio as the starting product, data from the statistical study and the equilibrium binding experiments were analyzed. The scale-up input parameters of 18 g/L load capacity and
83 μmol/mL Cu2+ binding resin capacity were overlaid on Fig. 9A
and D. The overlay showed that the 4 and 6-histidine tagged bispecific proteins would respectively elute with 129 mM and 211 mM
imidazole, while the aggregate would elute with 299 mM imidazole. The data also indicated that displacement of the 4-histidine
and 6-histidine proteins would not occur as the scale-up parameters overlaid in the homogenous design space for 4 and 6-histidine
tags on Fig. 9B and E.
Additionally, the scale-up load material contained a higher ratio of 6-histidine tagged bispecific protein compared to those used

in the screening and statistical studies. However, the equilibrium
binding study in Fig. 11 demonstrated that a load of 100% 6histidine tagged bispecific protein showed binding across the range
of Cu2+ binding capacity resins in the 5 mM mobile phase. In
the presence of the 170 mM mobile phase, only minor binding
was observed to the 97 μmol/mL resin. This indicated that imidazole concentrations greater than 170 mM would elute the pure
6-histidine tagged bispecific protein on the 83 μmol/L Cu2+ binding capacity resin. The imidazole concentration required for preventing the binding of the pure 6-histidine tagged material in the
binding study was consistent also with the imidazole levels necessary to elute peak 2 in the statistical study. Together, the statistical
and equilibrium binding studies indicated that 250 mM imidazole
step elution would elute both the 4 and 6-histidine tagged bispecific proteins while leaving the high molecular weight aggregate to
be stripped by the 500 mM imidazole buffer.
The scale-up purification also implemented the imidazole spiking of the load material and a wash prior to the elution. These
steps were included to decrease non-specific binding and allow for


18

V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

separation of host cell protein from the eluted protein. The 5 mM
imidazole spike was supported by the equilibrium binding data
generated in the 5 mM imidazole mobile phase. The 75 mM imidazole wash was supported by the statistical study in Fig. 9A that
showed 110 mM imidazole was the minimal concentration that
eluted protein from the weakest binding conditions.
The optimized IMAC process parameters were included in the
purification of harvested cell culture material, as described in
section 2.2.7, with successful results. Fig. 14 shows a representative
scale-up chromatogram as well product quality data. The product
quality of: 28.6% 4-histidine, 67.2% 6-histidine, 8 mg/mL, and 74.4%
monomer met the requirements for scale-up processing. The IMAC
product intermediate was successfully further polished to meet the
final product targets.
3.9. Application to other protein classes
The IMAC conditions developed in this study apply to the purification of molecules of various molecular formats. The IMAC
procedure described in Section 2.2.7 was applied to a diverse
poly-histidine tag containing molecules. Table 7 lists the diverse
molecules studied, molecular weights, cell lines used for expression, and resulting monomer purities and yields. All proteins were
directly purified from either CHO or HEK-293 cell harvests. Fractogel EMD Chelate (M) charged with 250 mM ZnCl2 or Ni Sepharose
Excel was used to capture products expressed in CHO cell culture. IMAC Sepharose 6 FF charged with either 250 mM ZnCl2
or CoCl2 . 5H2 0 was used to capture proteins expressed in HEK293 cell culture. The ease of operation, protein yields, and product
purities were consistent with the screening findings and showed
a broad application of the developed procedure. Polishing of the
IMAC products only required one additional aggregate removal operation to deliver material of acceptable purity.

was loaded. The results of the study showed that regardless of the
resin lot and protein loading, an elution buffer containing 250 mM

imidazole was adequate to elute the desired monomeric 4 and 6histidine species with separation from the high molecular weight
aggregate.
A third set of experiments assessed the binding of 4-histidine
and 6-histidine tagged bispecific proteins. As cell culture lots contained uncontrolled ratios of 4-histidine tagged and 6-histidine
tagged bispecific proteins, equilibria isotherms were generated for
the individual species as well as combined at different ratios.
Isotherms were also generated for different resin lots and in the
presence of null expression medium. It was observed that although
the 4-histidine and 6-histidine tagged bispecific proteins exhibited
different kL values, they obtained equivalent qm values. However,
when combined, 6-histidine tagged bispecific protein competed
with and displaced 4-histidine tagged bispecific protein. The presence of imidazole and null cell culture medium were also found to
decrease the qm values as they competed with the histidine-tagged
protein for binding.
The combination of the three sets of experiments enabled a
thorough understanding of the IMAC process for the histidine
tagged bispecific protein. The results of the resin-metal chelator
choice and performance experiments in conjunction with the binding experiments were utilized in the scale-up purification of the
poly-histidine tagged bispecific protein as well as proteins of a variety of other molecular classes; Fab, hormone, receptor, and exotoxin. For each class of poly-histidine tagged protein, the developed
process allowed for the elimination of pre-load conditioning diafiltration. The molecules also achieved high protein loading on the
IMAC resin. All purifications were high-yielding with acceptable
product quality to enable successful downstream polishing. Academic and industrial scientists can apply these findings and suggested process to quickly enable their purifications both at bench
and large scale.

4. Conclusions
Funding
IMAC purification process performance is known to be variable,
resulting in unpredictable product quality. This study contributes
to the understanding of factors impacting its variability and how
to control those factors. This understanding is critical to providing

confidence prior to using IMAC in both bench and large-scale research and development settings.
The first study in this work revolved around choosing an appropriate resin-metal chelator combination for the purification of
a poly-histidine tagged bispecific protein. It was found that screening results from the purified protein could not be extrapolated to
the protein present in cell culture harvest media, as the greatest
dynamic binding capacity for the purified protein was achieved
on IMAC Sepharose 6 FF charged with Cu2+ . In the presence of
HEK-293 medium, the optimal resin and metal combination was
IMAC 6 Sepharose FF charged with either Zn2+ or Co2+ . The highest binding capacity in CHO conditioned medium occurred on Fractogel EMD Chelate (M) charged with Zn2+ and Ni Sepharose Excel.
The binding differences were attributed to media components or
host cell line by-products that competed for binding and the extent of resin metal binding capacity. These identified combinations
allowed for the direct capture of expressed protein from media at
high loading capacities.
A second study examined interacting factors that impacted the
purification performance of the chosen resin-metal chelator combination, Fractogel EMD Chelate (M) charged with Zn2+ . This statistical study showed that the amount of imidazole needed for elution
of both the 4 and 6- histidine tagged bispecific proteins as well as
the product yield, purity, and histidine-tagged protein ratio were
influenced by two main factors. Those factors included the metal
binding capacity of the resin lot and the extent to which the resin

This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
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
Valeria Riguero: Conceptualization, Methodology, Investigation,
Writing - original draft, Resources, Supervision. Robert Clifford:
Methodology, Investigation, Writing - original draft. Michael Dawley: Methodology, Investigation. Matthew Dickson: Methodology,
Writing - original draft. Benjamin Gastfriend: Methodology, Software, Investigation. Christopher Thompson: Data curation, Formal analysis. Sheau-Chiann Wang: Methodology, Resources. Ellen
O’Connor: Conceptualization, Methodology, Formal analysis, Investigation, Writing - original draft, Resources, Supervision.

Acknowledgements
We would like to thank Christopher Afdahl, Jenny Feng, Coral
Fulton, Jeffrey Gill, Gerard Lacourciere, and Eugene Sun for their
initial observations. Thanks also to Marcia Carlson for her encouragement to initiate the project. We would also like to thank
Alan Hunter and William Wang for their careful reviews of the
manuscript.


V. Riguero, R. Clifford and M. Dawley et al. / Journal of Chromatography A 1629 (2020) 461505

References
[1] C. Kittinger, A. Barnes, A. Hunter, L. Machiesky, S. Phipps, A. Shannon,
R. Stadelman, S. Wilson, E. O’Connor, A high yielding IFNAR1 ECD mammalian
expression process for use in autoimmune disease drug development, Protein
Expres Purif. 167 (2020) 1–12.
[2] M. Flickinger, Downstream industrial chromatography, Recovery and Purification, Wiley, Hoboken, New Jersey, 2013.
[3] M. Peterka, M. Jarc, M. Banjac, V. Frankovic, K. Bencina, M. Merhar, V. Gaberc–
Porekar, V. Menart, A. Strancar, A. Podgornik, Characterisation of metal-chelate
methacrylate monoliths, J. Chromatogr. A 1109 (2006) 80–85.
[4] A. Kinna, B. Tolner, E. Rota, N. Titchener-Hooker, D. Nesbeth, K. Chester, IMAC
capture of recombinant protein from unclarified mammalian cell feed streams,
Biotechnol. Bioeng. 113 (2016) 130–140.
[5] S. Wang, S. Bai, X. Dong, Y. Sun, Partitioning behavior of enhanced green fluorescent protein in nickel-chelated affinity-based aqueous two-phase micellar
system and its purification from cell lysate, Sep. Purif. Technol. 133 (2014)
149–154.
[6] P. Fraga Garcia, M. Freiherr von Roman, S. Reinlein, M. Wolf, S. Berensmeier,
Impact of nanoparticle aggregation on protein recovery through a pentadentate chelate ligand on magnetic carriers, Appl. Mater. Interfaces 16 (2014)
13607–13616.
[7] P. Fraga Garcia, M. Brammen, M. Wolf, S. Reinlein, M. Freiherr von Roman,
S. Berensmeier, High-gradient magnetic separation for technical scale protein

recovery using low cost magnetic nanoparticles, Sep. Purif. Technol. 150 (2015)
29–36.
[8] Y. Yin, G. Wei, X. Zou, Y. Zhao, Functionalized hollow silica nanospheres for
His-tagged protein purification, Sens. Actuators B 209 (2015) 701–705.
[9] N. Bortone, M. Fidaleo, Immobilization of the recombinant (His)6 -tagged
l-arabinose isomerase from Thermotoga maritima on epoxy and cupper-chelate epoxy supports, Food Bioprod. Process. 95 (2015) 155–162.
[10] Y. Li, X. Xiao, R. Zhao, K. Yu, C. Hou, J. Liang, Study of the fabrication and
characterization of porous Ni using polystyrene sphere template, Synth. React.
Inorg. Metal 46 (2016) 286–290.
[11] Y. Chou, C. Ko, L. Chen, C. Yen, J. Shaw, Purification and immobilization of the
recombinant Brassica oleracea chlorophyllase 1 (BoCLH1) on DIAION R CR11 as
potential biocatalyst for the production of chlorophyllide and phytol, Molecules
20 (2015) 3744–3757.
[12] C. Zhang, D. Fredericks, D. Longford, E. Campi, T. Sawford, M. Hearn, Changed
loading conditions and lysate composition improve the purity of tagged recombinant proteins with tacn-based IMAC adsorbents, Syst. Synth. Biol. Nano
Med. 10 (2015) 480–489.
[13] J. Mooney, D. Ferdericks, C. Zhang, T. Christensen, C. Jespergaard, C. Schiodt,
M. Hearn, Purification of a recombinant human growth hormone by an integrated IMAC procedure, Protein Expr. Purif. 94 (2014) 85–94.

19

[14] W. Ning, S. Wijeratne, J. Dong, M. Bruening, Immobilization of carboxymethylated polyethylenimine–metal-ion complexes in porous membranes to selectively capture his-tagged protein, Appl. Mater. Interfaces 7 (2015) 2575–2584.
[15] R Prasanna, M. Vijayalakshmi, Characterization of metal chelate methacrylate
monolithic disk for purification of polyclonal and monoclonal immunoglobulin
G, J. Chromatogr. A 1217 (2010) 3660–3667.
[16] M Cheeks, N. Kamal, A. Sorrell, D. Darling, F. Farzaneh, N. Slater, Immobilized metal affinity chromatography of histidine-tagged lentiviral vectors using
monolithic adsorbents, J. Chromatogr. A 1216 (2009) 2705–2711.
[17] D. Ren, N Penner, B. slentz, F. Regnier, Histidine-rich peptide selection and
quantification in targeted proteomics, J. Proteome Res. 3 (2004) 37–45.
[18] F. Bartnicki, E. Kowalska, K. Pels, W. Strzalka, Imidazole-free purification of

His3 -tagged recombinant proteins using ssDNA aptamer-based affinity chromatography, J. Chormatogr. A 1418 (2015) 130–139.
[19] C. Karakus, M. Uslu, D. Yazici, B. Salih, Evaluation of immobilized metal affinity
chromatography kits for the purification of histidine-tagged recombinant CagA
protein, J. Chromatogr. B 1021 (2016) 182–187.
[20] W. Choe, R. Clemmitt, H. Chase, A. Middelberg, Comparison of histidine-tag
capture chemistries for purification following chemical extraction, J. Chromatogr. A 953 (2002) 111–121.
[21] W. Kast, (in German), Adsorption aus der Gasphase: Ingenieurwissenschaftliche Grundlagen und technische Verfahren, VCH, Weinheim, 1988
ISBN 3527267190, OCLC 35459784.
[22] Hydrodynamic radius calculator, Fluidic Analytics 2020. https://www.fluidic.
com/resources/Toolkit/hydrodynamic- radius- Converter/.
[23] R. Todd, R. Johnson, F. Arnold, Multiple-site binding interactions in metal-affinity chromatography: I. Equilibrium binding of engineered histidine-containing
cytochromes C, J. Chromatogr. A 662 (1994) 13–26.
[24] R. Johnson, R. Todd, F. Arnold, Multipoint binding in metal-affinity chromatography II. Effect of pH and imidazole on chromatographic retention of
engineered histidine-containing cytochromes C, J. Chromatogr. A 725 (1996)
225–235.
[25] M. Belew, T. Yip, L. Andersson, J. Porath, Interaction of proteins with immobilized Cu2+ : Quantitation of adsorption capacity, adsorption isotherms and
equilibrium constants by frontal analysis, J. Chromatogr. A 403 (1987) 197–206.
[26] S. Sharma, G. Agarwal, Interactions of proteins with immobilized metal ions:
a comparative analysis using various isotherm models, Anal. Biochem. 288
(2001) 126–140.
[27] S. Sharma, G. Agarwal, Interactions of proteins with immobilized metal ions, J.
Colloid Interface Sci. 243 (2001) 61–72.
[28] S. Vunnum, S. Gallant, Y. Kim, S. Cramer, Immobilized metal affinity chromatography: Modeling of nonlinear multicomponent equilibrium, Chem. Eng.
Sci. 50 (1995) 1785–1803.
[29] S. Vunnum, S. Cramer, IMAC: nonlinear elution chromatography of proteins,
Biotechnol. Bioeng. 54 (1997) 373–390.