Identification of Performance Drivers for an Antibody Producing CHO-S Cell Line Culture
in the Allegro™ XRS 20 Single-Use Bioreactor Utilizing Historical Data
Todd Sanderson, Alex Martino, Ing-Kae Wang, Amanda Rose, Byron Rees, Lisa Bradbury, Pall Life Sciences, 20 Walkup Drive, Westborough, MA, USA 01581
Figure 2
Data from Six Fed Batch CHO Cultures Set Up as Three Paired Runs
A
B
30
EXPERIMENTAL APPROACH
25
Two different Allegro XRS 20 bioreactors were used for the 20 L fed batch runs. A monoclonal antibody
producing CHO-S cell line was grown in CD FortiCHOu media fed with EfficientFeedu C on five consecutive
days and glucose as needed. In all cases the final expansion was done in the XRS 20 biocontainer.
An initial evaluation of CHO fed batch cell culture in the Allegro XRS 20 bioreactor in terms of growth curve and
viability, antibody production, and contaminants demonstrated excellent reproducibility. This provides an
opportunity to identify relatively small output differences due to process variation.
For the analysis of cell culture performance drivers, the following input variables were considered: passage
number; initial and final post expansion seed density; VCD at first feed; agitation; viability at harvest; and total
days in culture (Table 2).
In addition to almost daily sampling for cell count, viability, metabolite/nutrient, gases and pH measures, the
final harvest was assayed for mAb, host cell protein (HCP), and DNA concentrations. These outputs were used
in the analysis as well. The data from these runs was compiled and analyzed for trends linking the variations in
run parameters with the critical outputs of maximum viable cell density and mAb titer.
Summary of Bioreactor and Process Conditions
Table 1
Allegro XRS 20 Bioreactor Fed Batch Process Conditions
Cell line:
Industrial suspension CHO producing mAb biosimilar
Seed:
0.3 – 0.5 x 106 viable cells/mL in 5 L CDFortiCHO (Life Technologies)
Expansion:
Expand at 2 – 4 M viable cells/mL to 14 L
Feeding strategy:
CD EfficientFeed C (EFC, Life Technologies), initial feed at 8.0 ± 2.0 M viable cells/mL.
Additional bolus feeds every 24 hours for a total of 5 feeds (1.2 L/feed)
Carbon source:
Glucose kept above >2 g/L (bolus target >5g/L)
Antifoam C:
Added as needed
Aeration rate:
1.0 L/min
37 °C
Table 2
Fed Batch Culture Information for Eleven Allegro XRS 20 Bioreactor Runs
December
CPM
VCD @ EFC
Addition
1
30
26
7.3
December
2
30
26
6.6
March
1
30
26
5.9
March
2
35
26
5.2
April
1
35
19
8.5
April
2
30
19
8.0
July
1
30
9
11.5
September
1
30
9
10.8
October #1
2
30
26
8.8
October #2
2
30
28
9.0
October #3
1
30
26
9.4
Data Color
RESULTS
4
2
mAb (mg/mL)
Dec 30 rpm
Mar 30 rpm
0
HCP (mg/mL)
Mar 35 rpm
Apr 35 rpm
dsDNA (ug/mL)
Apr 30 rpm
The analysis of the first six Allegro XRS 20 fed batch cultures at our site did not clearly identify the performance drivers
for high mAb yield or high VCD, although these are obviously linked. Thus, additional data was incorporated into the
analysis as it was generated.
The growth and viability plots from five additional Allegro XRS 20 bioreactor fed batch cultures are shown in Figure 3.
Overall the curves from different run dates are similar. They have, on average, higher VCD than the earliest cultures (Figure 1).
There were minor variations in the culture set up, either allowed by the process ranges or outside those ranges but agreed
to based on intended use of the final harvest. Additionally the metabolite and glucose consumption data are very similar
across all eleven runs (data not shown).
Preliminary data from shake flask studies suggested that passage number influenced final [mAb], with passage number
>22 associated with reduced productivity. Data from shake flasks and XRS 20 fed batch cultures showed some indications
that the VCD window for the first nutrient feed (EFC) should be more narrow (data not shown), as feeding too early resulted
in lower maximum VCD and thus lower [mAb]. To evaluate this more fully, these parameters along with several others
were analyzed in pairwise combinations.
Figure 3
Growth Curves from an Additional Five Separate
Fed Batch CHO Cultures
40
100
35
90
Figure 4
Pairwise Analysis of Final mAb Production vs. CHO
Clone Passage Number
80
70
25
60
20
50
15
40
30
10
July
Sept
Oct #1
Oct #2
Oct #3
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
5
10
15
20
Passage Number
25
30
Color coded to match run colors shown in previous figures.
Note that some data points overlap
20
10
0
3 4 5 6 7 8 9 10 11 12 13
Elapsed Time (Days)
Figure 4 shows data looking at passage number and final mAb
concentration. For this non-GMP process the passage number is
generally kept below 20, but for reasons of timing higher passage
ViCell data average of replicate samples (2 or 3)
number was allowed for some of these runs. This provided an
Agitation at 30 rpm for all runs
opportunity to use these outside values to revisit the importance
of passage number on mAb production. Note that some other parameters vary as well (Table 2). Interestingly, the highest
and lowest mAb yields were achieved with high passage number, suggesting that this is not a strong performance driver.
0
1
2
Figure 5
Pairwise Analysis of Final mAb Production vs. Maximum VCD (A) and VCD at First Feed (B)
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
10
15
20
25
30
Maximum Viable Cell Density (million/mL)
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
00
2
4
6
8
10
VCD at 1st EFC Feed (million/mL)
12
Color coded to match run colors shown in previous figures, 11 fed batch cultures in total
Excellent Reproducibility in Parallel Allegro XRS 20 Bioreactor CHO Runs
In the first analysis, the focus was on three separate pairs of fed batch cultures (6 runs total, Table 2). In each of these,
two parallel fed batch cultures were set up from the same inoculum. The first of these was a training run, the second and
third looked at agitation of 30 vs. 35 rpm.
An initial comparison of cell growth and viability plots (Figure 1) and metabolite and glucose consumption data (not
shown) from these three separate parallel culture runs show:
Parallel culture (same date) show extremely high level of reproducibility
Repeat cultures (different dates) are very similar
December training run was good, but minor process changes led to viability and VCD improvements in April run
Figure 1
Growth Curves from Six Fed Batch CHO Cultures Run as Three Paired Runs
40
100
35
90
ViCell data average of replicate samples (2 or 3). Agitation at
30 or 35 rpm as indicated, Dec runs were technical replicates
80
70
60
20
50
15
40
30
Cell V
Viability
iability (%)
VCD (million/mL)
6
Analysis of Historical Data from Eleven Fed Batch Cultures in the Allegro XRS 20 Bioreactor
0
December, March, and April runs were each a pair of parallel cultures from the same seed train
5
8
mAb assay using ForteBio Octet system with Protein A sensors, HCP using Cygnus CHO HCP ELISA kit 3G, Invitrogen DNA determination
with Life Technologies PicoGreen assay
5
CHO Passage
# @ Expansion
10
10
mAb conc (mg/mL)
Run Date
Allegro XRS 20
Bioreactor ID
Max VCD (mil/mL)
Dec 30 rpm
VCD (million/mL)
5 L: 25 RPM, 5°X, 5°Y; 14 – 20 L: 30 or 35 RPM, 15°X, 5°Y
Temperature:
25
5
0
14
C
12
30
7.2 ± 0.05
Agitation:
30
10
O2 control at 40%
pH set point:
Dec #1, 30 rpm
Dec #2, 30 rpm
Ma
r, 30 rpm
Mar,
Ma
r, 35 rpm
Mar,
Apr, 35 rpm
Apr,
Apr,
Ap
r, 30 rpm
15
mAb Conc (mg/mL)
DO set point:
20
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Conc (mg/mL)
During the development of processes for high production CHO fed batch cultures, there is usually insufficient
time for thorough process optimization. Once defined, these processes often have ranges rather than
specific values for many parameters, resulting in minor variations in the runs. These small process variations,
combined with resultant differences in performance, can be used to further optimize the process. Over the
course of one year, eleven 20 L fed batch cultures of an antibody producing CHO-S cell line were generated in
the Allegro XRS 20 bioreactor. In addition to normal minor process variation, small process changes occurred
due to method optimization efforts or culture timing requirements. An analysis of the combined historical data
was undertaken to better understand the specific culture performance drivers, leading to further improvement
in culture outputs, specifically final antibody concentration and maximum viable cell density (VCD).
Excellent Reproducibility in Parallel Allegro XRS 20 Bioreactor CHO Runs
The data summary in Figure 2, showing maximum VCD (panel A), concentration determinations for mAb and host cell
protein (panel B), and double stranded DNA (panel C) in the harvest sample indicate:
Parallel cultures give similar results, especially mAb titer & max viable cell density (VCD)
Cultures with highest maximum VCD also had highest mAb titer, as expected. The degree of mAb increase (~50%) was
surprising and the reason for the increased VCD was not obvious from this data
Comparison of 30 vs. 35 rpm data shows these agitation settings have limited or no impact, but further testing with
a broader range of agitation settings could be explored
Cell Viability (%)
BACKGROUND AND NOVELTY
20
10
0
0 1 2 3 4 5 6 7 8 9 10 11
11 12 13
3
Elapsed Time (Days) Post Final In-biocontainer Expansion
0
Contact: 800.717.7255 (USA and Canada) • 1.516.484.5400 (Outside USA and Canada) • www.pall.com/biopharm • E-mail: biopharm@pall.com
A plot of maximum VCD against the final mAb concentration (Figure 5A) shows a clear correlation between higher
mAb and higher VCD, which was not a surprise. However, this data does not indicate why higher VCD was achieved. An
analysis of the [mAb] against VCD at the time of the first addition of the efficient feed C supplement (Figure 5B) showed that
there is a relatively narrow optimal VCD for this step. Although the preliminary testing identified a VCD range for EFC addition,
this analysis suggested that the window was set too wide. By raising the lower limit from 6 to 8 x 106 cells/mL, the mAb
production was significantly increased. Setting the initial feed range to 8-10 x 106 cell/mL results in higher maximum VCD as
well (data not shown). If the initial feed is delivered before or after that VCD, both maximum VCD and final [mAb] are reduced.
Analysis of other variables did not show a significant relationship with mAb production or maximum VCD.
CONCLUSION
The Allegro XRS 20 single use bioreactor generated highly reproducible cell culture performance for this CHO mAb
production process, allowing for a holistic analysis of relatively small changes in process conditions and the resultant
differences in performance. In three separate paired runs, the first being identical and the second two differing by 30 vs.
35 rpm agitation, the parallel culture results are extremely similar, highlighting the overall reproducibility of the system.
This bioreactor reproducibility makes it possible to review historical data to identify performance drivers that were missed
in limited optimization testing done at smaller scale prior to the 20 L runs. The VCD at the time of the first feed addition
proved very important for overall mAb yield. Although a relationship was not surprising, the optimal range was tighter
than previously believed. Nutrient feed timing can be easily modified as there is very good understanding of the growth
curves for this CHO-S cell line in the Allegro XRS 20 bioreactor. This opportunistic analysis of historical data has some
limitations compared to experiments designed specifically to look at process controls. Despite these limitations the value
is clear, it can provide additional process understanding without additional experimentation. Thus, this approach could
further understanding of performance drivers for any production run, and can be expanded to other cell line / media
combinations in any culture system having sufficient reproducibility to distinguish relatively small changes in process
conditions and the resultant small but significant differences in performance.
© 2015, Pall Corporation. Pall,
, and Allegro are trademarks of Pall Corporation. ® indicates a trademark registered in the USA. uFortiCHO and EfficientFeed are trademarks of Life Technologies .
5/15, GN15.6279