Thomas Müller-Späth
- manuscript draft -
2/16/2016
Two Step Capture and Purification of IgG2 using Multicolumn
Countercurrent Solvent Gradient Purification (MCSGP)
a,b
a,b
a,b
c
d
T. Müller-Späth , L. Aumann , G. Ströhlein , H. Kornmann , P. Valax ,
5
L. Delegranged, E. Charbautc, G. Baerd, A. Lamproyed, M. Jöhncke, M. Schultee, M.
Morbidellia
aInstitute for Chemical and Bioengineering, ETH Zurich, CH-8093 Zurich, Switzerland
bChromaCon AG, Technoparkstr. 1, CH-8005 Zurich, Switzerland
10
cMerck Serono S.A., Zone Industrielle l’Ouriettaz, CH-1170 Aubonne, Switzerland
dMerck Serono S.A., Zone Industrielle B, CH-1809 Fenil-sur-Corsier, Switzerland
eMerck KGaA, Performance & Life Science Chemicals Research & Development - Life
Science, Frankfurter Str. 250, D-64293 Darmstadt, Germany
15
20
corresponding author:
Prof Massimo Morbidelli
25
Institute for Chemical and Bioengineering
ETH Zurich
Wolfgang-Pauli-Str. 10 / HCI F 129
8093 Zurich, Switzerland
email: massimo.morbidelli@chem.ethz.ch
30
phone ++ 41.44.6323034 // ++ 41.44.6323033
fax ++ 41.44.6321082
http://www.morbidelli.ethz.ch
35
1/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
Supplemental Information
Schematic and description of the four-column MCSGP process
5
10
Figure-S 1: Schematic of the MCSGP process using four columns shown in interconnected
state (CC) and batch state (BL). Arrows indicate mobile phase flows. The numbered
rectangles indicate positions and tasks of the columns that correspond to sections of the
schematic batch chromatogram shown in the lower part of the figure. The section borders
of the batch chromatogram are indicated by vertical thin dotted lines. Dashed sections and
15
flows correspond to tasks in the batch state, full flows and grey sections correspond to
tasks in the interconnected state. The dashed line in the chromatogram represents the
linear batch gradient that is transferred to the MCSGP process; the crosses (x) indicate the
gradient concentrations at the section borders.
20
The four-column MCSGP process includes fully continuous loading and CIP. It is divided
into an interconnected state, where the product is internally recycled (i.e. there is only
one outlet stream), and into a batch state, where product and impurities are eluted from
25
the system. Interconnected and batch states alternate in time. The columns remain in
the interconnected state for the time period tCC and in the batch state for the time period
tBL. The columns switch from high to low section numbers in the order (6, 5c, 5b, 5a, 4,
2/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
3, 2, 1) and return to section 6 after having completed the task of section 1.
Summarizing, the tasks consist of loading (batch and interconnected state, tasks 5c, 5b),
washing (batch state, task 5a), recycling of product overlapping with weakly adsorbing
impurities (interconnected state, task 4), product elution (batch state, task 3), recycling of
5
product overlapping with strongly adsorbing impurities (interconnected state, task 2),
stripping, CIP, column re-equilibration (batch state, task 1), and reception of the eluate
from section 4 (task 6). As batch and interconnected state tasks are performed in an
alternating manner, the process requires 4 pumps. In contrast to previous work (MullerSpath et al. 2009), the process used here does not have a section that is dedicated to
10
column sanitization. The CIP step is incorporated in section 1 after stripping of the
strongly adsorbing impurities.
Performance parameter definition
In order to compare MCSGP and batch process performance, performance parameters
15
are defined here.
The process yield Y generally refers to the product outlet and is defined as follows:
Y
mmAb, Product
mmAb, Feed
for MCSGP chromatography (1a), and
and
20
Y
mmAb, Product
mmAb, Feed
for batch chromatography (5b)
where mmAb , Product and mmAb , Feed in eq. 5a denote the average mass flow of mAb exiting
the MCSGP process through the product outlet and entering the system through the
feed inlet, respectively. In the yield definition for batch experiments (eq. 5b) the mass
25
flows are replaced by the total mass of mAb in the product pool, mmAb , Product and in the
feed, mmAb , Feed , respectively. These masses are calculated from the feed and product
pool volumes and the corresponding concentrations.
The measure for the impurity content X with respect to HCPs and DNA is given by:
X
cimp
cmAb
(2a)
3/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
where cimp and cmAb stand for concentrations of impurity and mAb in the product pool,
respectively. The impurity content is given in the units parts per million [ppm] (= ng
impurity/mg mAb).
With respect to impurities with concentrations large enough to be determined by
5
analytical chromatography, such as fragments and aggregates, the purity is defined by
the area ratio of the mAb peak and the total peak area and is given in percent [%].
P
AmAb
Atotal
(6b)
Such two purity definitions differ in the kind of non-mAb components they refer to, and
therefore eventually depend upon the adopted analytical technique as discussed in the
10
analytics section.
It is worth pointing out that the purity P, defined by eq. (6b) refers to the sum of all the
fragments (W1, W2 and W3) measured as discussed above. For judging the mAb purity
with respect to aggregates, no differentiation was made between dimers, trimers etc.
The time-specific load L of the MCSGP unit with continuous feeding is independent of
15
the switch times and defined as follows:

mmAb, Feed
L
VC
(3a)
For batch processes, the load per time is defined as:
L
mmAb, Feed
VC

1
trun
(7b)
where trun is the total time duration of the batch run and Vc indicates the total stationary
20
phase volume of the MCSGP unit or the batch column, respectively. The productivity
Prod is defined as the product of L and Y for both MCSGP and batch processes, i.e.:
Prod  L  Y
(4)
.
25
The buffer consumption BC is defined as the buffer stream, V buff required per mAb
product stream (MCSGP, eq. 9a), or as total volume of buffers, Vbuff used per total mass
of mAb produced (batch processes, eq. 9b):
4/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
.
V buff
BC 
mmAb , P
BC 
Vbuff
mmAb , P
(equation 5a)
(equation 9b)
Polishing experiments – detailed screening conditions
5
The polishing materials were packed according to the manufacturer’s instructions.
The anion-exchange (AIEX) resins Fractogel TMAE HiCap, Q-Sepharose FF and the
multi-modal anion-exchanger Capto Adhere were tested in flow-through mode. In order
to prevent breakthrough of impurities the ionic strength of the feed (i.e. the MCSGP
10
product pool) was decreased while keeping the pH high, but below mAb binding
conditions. This approach is also recommended by the manufacturer of Capto Adhere to
clear HCP. Since the salt concentration of the MCSGP product was ca. 0.1 M, dilution
was required. Due to the limited amount of MCSGP material available, no detailed
screening was performed regarding the impact of the loading pH.
15
For screening of AIEX and multimodal AIEX materials in flow-through mode, the
equilibration buffer was 20 mM phosphate, pH 6.0. The feed (MCSGP-derived material)
was diluted 1:5 with deionized water and the pH was 6.0-6.1.
Evaluation of the flow-through with the analytical protein A analysis was used as a first
indicator for the purity with respect to fragments. HCP-ELISAs were made to complete
20
the investigations.
The hydrophobic interaction (HIC) stationary phases Resource Phenyl and Resource
Isobutyl, the multi-modal anion exchange resin Capto Adhere and Ceramic Flouroapatite
Type I were screened in bind / elute mode. The eluate pools were evaluated for yield
and purity with the protein A analysis and tested for HCP content by NS0-ELISA.
25
For HIC, material from the MCSGP process was adjusted to pH 5.0 (Resource Phenyl)
and pH 5.8 (Resource Isobutyl) using acetic acid, respectively and brought to a sodium
chloride concentration of 1.5 M which was found to be sufficient for adsorption. The
buffers were 2.0 M ammonium sulfate, pH 5.8 (adsorption buffer) and 50 mM phosphate,
pH 5.8 (elution buffer). For elution from Resource Phenyl, a linear gradient from 60 to
30
100% elution buffer in 15 min was run at 90 cm/h. For Resource Isobutyl, a gradient
5/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
from 15 to 80% elution buffer was performed at 90 cm/h. The temperature for both
experiments was kept constant at 25°C.
For Ceramic Fluoroapatite (CFT), MCSGP-purified material served as feed, diluted 1:3
with deionized water and adjusted to the screening pH using either acetic acid or NaOH.
5
The screening was done with feed and buffers at pH 6.0, pH 7.0, pH 7.2 and pH 7.5. The
binding buffer concentration was 20 mM phosphate and the elution buffer concentration
was 400 mM phosphate. A linear gradient elution from 0 to 50% B in 20 min was run at
a linear flow velocity of 150 cm/h.
For Capto Adhere polishing in bind / elute mode, a one-dimensional screening with the
10
buffer pH of the elution buffer as screening parameter was performed. The salt
concentration was kept low throughout all experiments to maximize HCP clearance.
MCSGP-derived material as feed was diluted 1:3 with deionized water and its pH was
adjusted to pH 7.8-8.1 using NaOH. Only a marginal improvement of the 1% dynamic
breakthrough capacity of 35 g/L at pH 7.8 to pH 8.1 was observed (data not shown) at a
15
linear flow velocity of 920 cm/h, while at pH 7.0 early breakthrough was observed. For
this reason the feed pH was fixed at pH 7.8. The load was 15 g/L. The binding buffer
was 10 mM phosphate, pH 7.8. The elution buffer was a 10 mM phosphate/10 mM
citrate buffer intended to produce a smooth linear gradient from pH 7.8 to 4.0. The
elution was run from 0-100% elution buffer in 15 min. The loading flow velocity was 920
20
cm/h and the washing and elution flow rate was 300 cm/h.
Analytics – detailed conditions
Analytical protein A chromatography
MAb concentrations were determined using analytical protein A chromatography with a 2.1
25
x 30 mm Poros A/20 column (Applied Biosystems, Foster City, CA, USA).
The binding buffer was 20 mM phosphate, 0.1 M NaCl at a pH of 7.0 (buffer A). The
elution buffer (buffer B) was a 0.1 M citric acid solution adjusted to pH 2.5 with NaOH/HCl.
The flow rate was 1 mL/min. A step gradient elution (100% B) was done after a wash of
0.5 min with binding buffer. The protein A method was calibrated using cCCS with a
30
known concentration of cmAb= 2 g/L.
6/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
Size exclusion chromatography (SEC)
SEC was carried out using a Superdex 200 10/300 GL column (GE Healthcare, Uppsala,
5
Sweden) and a 25 mM phosphate, 0.2 M Na2SO4 buffer of pH 7.0 at a flow rate of 0.6
mL/min.
SDS-polyacrylamide gel electrophoresis (SDS-PAGE)
10
SDS-PAGE was carried out using NuPage Novex 4-12% Bis-Tris gels, MES SDS
running buffer and Novex Sharp Pre-Stained Protein Standards (Invitrogen, Carlsbad,
CA, USA). The non-reduced samples were prepared using NuPage LDS Sample buffer
(4x), (Invitrogen), according to the manufacturer’s protocol. The reduced samples used
the same sample buffer but included NuPage Sample reducing agent (Invitrogen) and
15
NuPage antioxidant (Invitrogen) was added to the running buffer. The gels were run for
40 min at a constant voltage of 200V. For staining, the Silver Stain Plus kit, (Bio-Rad,
Hercules, CA, USA) was used.
NS0 HCP ELISA
20
The HCP concentration was determined with a NS0 Host Cell Protein assay.
Absorbance for ELISA was measured using a Tecan GENios Pro (Tecan, Männedorf,
Switzerland).
Analytical cation exchange chromatography
25
The identification and quantification of mAb fragments was carried out with weak cation
exchange chromatography using a 4 x 250mm Propac wCX-10 column, (Dionex,
Sunnyvale, CA, USA). The binding buffer (buffer A) was 20 mM acetate, pH 5.0; the
elution buffer (buffer B) was 20 mM acetate, 1 M NaCl, pH 5.0. A linear gradient from 20
30
to 40% B in 60 min was run at a flow rate of 0.5 mL/min.
7/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
mAb fragment isolation
For isolation of the mAb fragments a semi-preparative Mono S 5/50 GL (GE, Uppsala,
Sweden) column was used. The binding buffer (buffer A) was 20 mM acetate, pH 5.0,
5
and the elution buffer (buffer B) was 20 mM acetate, 1 M NaCl, pH 5.0. After
equilibration with buffer A, protein A flow-through from mAb capture (cCCS 1:3 diluted
with deionized water and adjusted to pH 5.0 with acetic acid) was loaded for 20 min at a
flow rate of 1.0 mL/min. The elution was done using a linear gradient from 0 to 40% B in
60 min at a flow rate of 0.5 mL/min. The gradient elution was fractionated at 1.0
10
min/fraction. Based on analyses with the Propac method fractions MS A-D were
identified to contain mAb fragments.
For the isolation of the W-impurities, protein A chromatography flow-through obtained
from Merck-Serono was loaded onto the Mono S column, eluted with the gradient
described above and fractionated into four fractions A-D. The Mono S fractions were
15
analyzed with the Propac column and SDS-PAGE in order to determine the molecular
weight of the impurities (data not shown). From the Propac analysis it can be seen that
fraction A contains impurities W1 and W2, while fractions B and C consist mainly of W2
and W3, and fraction D contains mainly W3. Figure-S 2 shows the gels for the nonreduced (left) and the reduced samples (right).
20
Figure-S 2: Left Gel: Non-reducing SDS-PAGE: lanes 1&10: Marker, lane 2: CCCS, lane 3:
Protein A flow-through, lane 4: Mono S fraction A, lane 5: Mono S fraction B, lane 6: Mono
25
S fraction C, lane 7: Mono S fraction D, lane 8: Pure mAb (bulk, Merck-Serono). Right Gel:
reducing SDS-PAGE: The sample order was the same as on the non-reduced gel.
Fractions C and D have essentially the same band pattern. Moreover, the same samples
were run on SDS-PAGE under reducing conditions in order to solve disulfide bonds in
8/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
mAb-related impurities and to generate 25 kDa and 50 kDa weak and heavy chains in
prospective mAb fragments. The band intensity from the gel (Figure-S 2) and the peak
height from the Propac chromatograms of fraction A-D, both proportional to the
concentration, correlate well which allows for a mapping of the chromatogram peaks to
5
molecular weights. W1 was identified as 5 kDa impurity, W2 as 25 kDa impurity and W3
as 50 kDa impurity. The reduced gel, shown in Figure-S 2, indicates that W2 and W3 are
mAb-related since all fractions containing 25 and 50 kDa impurities are reduced to a
band at 25 kDa. The fact that the band is the same indicates that W2 is a single chain
Fab fragment and W3 is a complete Fab fragment. Upon reduction, W3 is split into two
10
25 kDa single chain fragments. The occurrence of free Fab-fragments in cell culture
supernatant has been observed also by other researchers (van Reis 2006). The bands
at 12.5 kDa and 5 kDa are not affected by the reduction. Thus, it cannot be concluded
that W1 is mAb-related. The 12.5 kDa impurity is not visible in the chromatogram shown
in Figure 1 and must have a very similar retention behavior as W2 or W3 or result from
15
degradation during the sample preparation.
Feed pH screening
Supernatant was diluted 1:3, adjusted to different pH values and run on the CIEX
material in linear batch gradient elutions at pH 6.0. Figure-S 3 shows the chromatograms
20
and the results obtained from offline Propac analyses for the mAb and the sum of the
mAb fragments W1, W2 and W3, for feed pH values of 5.5, 5.8, and 6.0, respectively.
The elution profiles of the mAb match very well indicating that the runs are comparable
in terms of mAb adsorption and desorption. However, the amount of fragments clearly
decreases from pH 5.5 to 6.0. At pH 5.5, the amount of fragments is 7-fold larger than at
25
pH 6.0; at pH 5.8 it is still 3.5-fold higher than at pH 6.0. The washing with the pH 6.0
buffer after loading did not allow bringing the purity to the same levels. It would be very
interesting to examine closer the retention behavior of the fragments as a function of the
pH on the one hand, and as a function of the salt concentration on the other. A
comparison with the retention behavior could lead to new separation options e.g. by pH
30
gradients.
9/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
Figure-S 3: Overlaid chromatograms of batch gradient elutions at pH 6.0 from Fractogel
SO3 (M), mAb and fragment concentrations (sum of W1, W2 and W3, determined by Propac
offline analyses). Identical amounts of supernatant diluted 1:3 and adjusted to pH 5.5, 5.8
5
and 6.0 were loaded. Closed symbols indicate the fragment concentration and open
symbols the mAb concentration, respectively, for loading at: pH 5.5 (,), pH 5.8 (,)
and pH 6.0 (,).
10/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
MCSGP operating conditions
Table-S I: Operating conditions for MCSGP using Fractogel SO3 (MCSGP 1 – left part) and
5
Poros HS50 (MCSGP 2 and 3 – right part). Parameters of Poros HS50 experiment MCSGP 3
that differ from the ones from MCSGP 2 are given in parentheses. The feed for MCSGP 1
and MCSGP 2 was cCCS diluted 1:4 and adjusted to pH 6.0. The feed for MCSGP 3 was
cCCS diluted 1:3 and adjusted to pH 5.8. All parameters are defined in Figure-S 1
Separation tasks
Separation tasks
tBL
[min]
sw. time
Q2
Q3
Q4
Qfeed
Q6
tCC
[min]
10
tBL
[min]
6
sw. time
BL phase CC phase grad start grad end
[mL/min] [mL/min] [% B]
[% B]
x
0.61
28.3
36.6
0.97 x
24.9
28.3
x
1.01
15.8
19.0
4.00
4.00 n.a.
n.a.
x
0.0
0
0
Strip and CIP tasks
Q1
QCIP
Qpre-equil
Qequil
Qwash
Q2
Q3
Q4
Qfeed
Q6
tCC
[min]
4
BL phase
[mL/min]
x
1.70
x
6.5 (4.9)
x
6
CC phase grad start grad end
[mL/min] [% B]
[% B]
0.19
28.3
36.6
x
24.9
28.3
0.35
15.8
19.0
6.5 (4.9) n.a.
n.a.
0.19
0
0
Strip and CIP tasks
BL phase time
[mL/min] [min]
3
1.5
1.5
2.2
3.5
1.8
1.8
3.0
2.8
grad start
[% B]
100
n.a.
n.a.
n.a.
grad end
[% B]
100
n.a.
n.a.
n.a.
1.7
0
0
Q1
QCIP
Qpre-equil
Qequil
Qwash
BL phase time
[mL/min] [min]
4
2
2
4
4
0.7
0.8
1.0
0.9
grad start
[% B]
100
n.a.
n.a.
n.a.
grad end
[% B]
100
n.a.
n.a.
n.a.
0.85
0
0
10
The experimental design procedure (Aumann and Morbidelli 2007) was used to provide
initial values for the simulation of the MCSGP operations. Based on the simulations the
three sets of operating conditions reported in Table-S I have been identified and
15
validated experimentally. It is worth noting that the initial volumetric flow rate values for
the runs using the Poros material were 2.2-fold lower than for the Fractogel runs which
corresponds to the ratio of the column cross section areas (Fractogel: 0.75 cm inner
diameter columns, 4.4mL column, Poros: 0.5 cm inner diameter, 2.0 mL columns).
11/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
MCSGP run 3
Figure-S 4: Start-up and steady state for run MCSGP 3 (Poros HS50). The thin grey solid
5
line indicates the A280 UV signal recorded at the product outlet as a function of time. The
symbols with horizontal bars denote the following parameters calculated from offline
measurements (by analytical protein A chromatography) averaged over one cycle: mAb
concentration (), yield (), purity (). The vertical lines indicate the sampling intervals
that correspond to one cycle.
10
12/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
Polishing step with batch processes
AIEX, multi-modal AIEX chromatography in flow-through mode
The material obtained from MCSGP was purified using AIEX chromatography in flow-
5
through mode using different stationary phases with comparable feed conditions. The
mAb was fully recovered with yields of > 99% and the HCP clearances of the single
resins were 6-fold (Fractogel TMAE HiCap (M)), 22-fold (Q-Sepharose FF) and 31-fold
(Capto Adhere). With the latter two materials, it was possible to reach 20 ppm HCP
content which is just 2-fold higher than the specification limit. Thus summarizing, in order
10
to be in specification, an HCP clearance about one order of magnitude larger than the
one achieved with these resins is needed. Accordingly, these resins would require
further optimization possibly including pH adjustment and dilution.
Ceramic Fluoroapatite and HIC chromatography in bind- elute mode
Ceramic Fluoroapatite (CFT) has been reported in the literature as a powerful tool for the
15
removal of impurities in polishing applications (Schubert and Freitag 2007). However,
because it may degrade through metal ion catalysis, it is not used as a capture material.
The screening with CFT was done at pH 6.0, pH 7.0, pH 7.2 and pH 7.5. For the run at
pH 7.2, yield and purity values as determined by Protein A analysis were the highest (>
99% and 98.1%, respectively). The HCP content of the pool fraction under these
20
conditions was 30 ppm which is three fold higher than the specification limit. As in AIEX
chromatography (flow-through mode), a clearance of one order of magnitude larger
would be desirable. We can then conclude that CFT under the tested conditions is not
sufficient as a polishing step for the 2-step process.
The HIC materials Resource Phenyl and Resource Isobutyl were tested for polishing in
25
bind-elute mode. The purified fractions of both runs tested with ELISA, showed that the
clearance of HCP was negligible (factor 1.4 and 2.0).
13/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
Robustness of multi-modal AIEX chromatography in bind/elute mode
Figure-S 5: HCP concentration from Capto Adhere polishing pools determined by ELISA
5
as a function of product yield in the Capto Adhere polishing runs
14/15
Thomas Müller-Späth
- manuscript draft -
2/16/2016
References
Aumann L, Morbidelli M. 2007. A continuous multicolumn countercurrent solvent
5
gradient purification (MCSGP) process. Biotechnology and Bioengineering 98(5):10431055.
Muller-Spath T, Aumann L, Morbidelli M. 2009. Role of Cleaning-in-Place in the
Purification of mAb Supernatants Using Continuous Cation Exchange Chromatography.
Separation Science and Technology 44(1):1-26.
10
Schubert S, Freitag R. 2007. Comparison of ceramic hydroxy- and fluoroapatite versus
Protein A/G-based resins in the isolation of a recombinant human antibody from cell
culture supernatant. Journal of Chromatography A 1142(1):106-113.
van Reis R. 2006. Membrane Processes in the Biotechnology Industry. AICHE. San
Francisco, USA.
15
15/15