“development application and single use a6 aber talk for umas 2_tcm18-148843.pdf• single use...
TRANSCRIPT
“Development application and single use probes for building Design Space of live cell concentration by dielectric spectroscopy”
BY: JOHN CARVELL
ABER INSTRUMENTS, Wales, UK
ABER Instruments Ltd – originally a spin out company from the University of ABERystwyth, Wales.
Wales also has the longest station name in the world!
Fortunately it is shortened to Llanfair PG
Scope of Talk
• Principle of Dielectric Spectroscopy (DS) –emphasis on scanning
• Correlation with offline readings
• Case study Gedeon Richter
• Case study GSK
• Case study Fujifilm Diosynth
• Single use probe applications
• Conclusions
Industry Initiative
“Systems that promote greater product and process understanding can provide a high assurance of quality on every batch and provide alternative, effective mechanisms to demonstrate validation” FDA Guidance for Industry: PAT—A Framework for Innovative Pharmaceutical Manufacturing and Quality Assurance, 2004.
Four types of tools for generation and application of process understanding:
1. Multivariate tools for design, data acquisition and analysis
2. Process Analyzers
3. Process Control tools
4. Continuous improvements and knowledge management tools
PAT perspective
Principle of Dielectric Spectroscopy (DS)
•When an alternating electrical field is applied to an ionic solution of cells, a charge separation or polarization effect develops at the cell membrane
•Cells behave like a capacitor under an electric field
•The magnitude of this charge separation can then be measured by the capacitance of the solution at varying radio-frequencies in the range 0.1-20MHz
•The resulting capacitance is generally proportional to the total enclosed bio-volume, which itself is directly correlated with cell density or cell weight
Dielectric Spectroscopy (DS) in bioreactors
Dielectric Spectroscopy Key Outputs: • Capacitance Data Over Time • Correlates with Viable Cell Density
++ +
++
++
+ +
+
+ ++
++
+
- - -
--
-
---
-
--
- --
-
++ +
+
++ +
+
-- - -
---
-
DS Probe
DS Probe
Cell Populations
Single Cell
Bioreactor
Aber DS Probe
Population of Cells
Single Cell
0
5
10
15
20
25
0 50 100 150 200 250 300
Pe
rmit
tivit
y (
or
Ca
pa
cit
an
ce
)
Time (hrs)
Total Permittivity (or Capacitance) vs. Time
+
+
+ +
+ +
- - - -
- - - - -
+ + +
0.E+00
1.E-09
2.E-09
3.E-09
4.E-09
1.E+05 1.E+06 1.E+07
Pe
rmit
tivi
ty (
F/m
) αC
apac
itan
ce
Frequency (Hz)
Capacitace/Permittivity vs Frequency
Day 3-66 hr
Day 6-137 hr
Day 8- 194 hr
Day 11-252 hr
Dielectric Spectroscopy Key Outputs: • Frequency Scanning Data Over Time • Tells us about Cell Populations
Graphics courtesy of Bend Research,Oregon
Correlation of online capacitance data with offline readings
• Plot of capacitance at specific single frequency versus VCD should produce straight line correlation with an R2 value close to 1
• Correlation during log phase of cell growth is well maintained
• Divergence observed as cells enter stationary/death phase at which point cell viability decreases and LDH levels increase proportionally
• Published data on models for correcting this divergence available eg Bend, UMASS, University of Manitoba
Correlation of Viable Cell Density v Capacitance
Viable Cell Density
0 2 4 6 8 10 12 14 16
Cap
acita
nce
0
5
10
15
20
25
R2 = 0.9787
CHO Cell Density Monitoring in Bioreactor Culture – Comparison of four methods (data courtesy) University of Manitoba )
10
Fed-Batch HEK cells (data courtesy of CNRC, Canada)
0.6 MHz 2.2 MHz 10 MHz
Change in shape of Spectrum- fundamentals of frequency scanning
101
102
103
104
105
-0.2
0
0.2
0.4
0.6
0.8
1
Frequency (kHz)
Norm
aliz
ed C
apacitance
Timeseries of Normalized Capacitance vs. Frequency Timeseries of Normalized Capacitance vs. Frequency Timeseries of Normalized Capacitance vs. Frequency
No
rma
lize
d C
ap
acita
nce
No
rma
lize
d C
ap
acita
nce
Frequency (kHz) Frequency (kHz)
Curves move from
Blue – 0hrs
Red – 72 hrs
Control CHO cell undergoing induced apoptosis
Details of study at Gedeon Richter
• 6 runs with 1L bioreactors and ATF2 technology (Refine,USA) with CHO expressing Mab
• High cell densities up to 75 x10E6 cells/ml
• Utilized Aber Biomass Monitor 220 with AberScan software
• Scans at 25 pre-defined frequencies every 8 minutes
• Spectral data exported for linear regression or multivariate data analysis
• Cells/ml derived from Aber using 3 different methods:
- Simple measurement at 580KHz (linear modelling)
- Cole- Cole
- PLS using Umetrics SIMCA software
• at-line measured VCD values throughout Run 1-4
were plotted against predicted VCD values and
depicted as triangles
• linear equation for prediction:
VCD = 414492 × C580 + 1.187 × 106
• linear model over-estimated VCD at the later phases
of the cultivation
Linear Modelling at 580KHz
Deriving VCD using Cole-Cole using frequency scanning between 100 and 20,000KHz
•
Off-line cell counts (VCD) for 6 runs with REFINE system
Score plots for 6 runs
• Multivariate analysis (PCA) of capacitance spectra resulted in detection of measurement disturbances
• Principle component t(1) captured 92% of variance in spectra and showed same trend for all 6 runs
• DIELECTRIC SPECTRA NOT INFLUENCED BY BATCH TO BATCH DIFFERENCES OF CRITICAL CULTIVATION PARAMETERS INDICATING ROBUSTNESS OF probe
• VCD prediction with multivariate PLS model developed
CHO with continuous feed-Comparison of 3 methods to predict VCD
0
200
400
600
800
1000
1200
0.0E+00
1.0E+07
2.0E+07
3.0E+07
4.0E+07
5.0E+07
6.0E+07
7.0E+07
0 2 4 6 8 10 12 14
fc [
kH
z]
VC
D [
ce
lls
/ml]
Process time [day]
VCD prediction Run2 VCD at-line
linear (C580)
Cole-cole
PLS
fc
CHO with continuous feed-Comparison of 3 methods to predict VCD
0
200
400
600
800
1000
1200
0.0E+00
1.0E+07
2.0E+07
3.0E+07
4.0E+07
5.0E+07
6.0E+07
7.0E+07
8.0E+07
0 2 4 6 8 10 12 14
fc [
kH
z]
VC
D [
ce
lls
/ml]
Process time [day]
VCD prediction Run1 VCD at-line
linear (C580)
Cole-cole
PLS
fc
18 February 2014 Dielectric Spectroscopy and its
application in process development
Dielectric Spectroscopy and its application in process development Andrew Heinrich Biopharm R&D, GSK Stevenage
PAT & QbD Forum, Feb 18th
Goettingen, Germany
Case study 2
GSK, Stevenage, UK
Use of capacitance in scale-up and technology transfer
• Background
– Cytokine produced in modified CHO cells
– Initial cell line selection carried out in shake flask studies
– Process scoping in 2L bioreactor system
• In-situ Aber FUTURA Probe used in 2L glass bioreactors with multiple frequency scanning
• What information can we glean from this data?
– Relatively similar process profiles
– Perturbations in capacitance profiles appear to correspond with feed points in process
– Can this data be used to help build a ‘process model’ for scale up (and scale-down)?
Single frequency measurements for various bioreactor runs
Run Time Hrs
0 50 100 150 200 250 300 350
Me
asu
red
Ca
pa
cita
nce
0
10
20
30
40
50
Scanning frequency analysis Curve fitting the data
• GSK developed an XL-fit macro for rapid and reproducible automatic analysis of scanning data sets
Use of capacitance in scale-up and tech transfer
• Analysis of various scanning data parameters highlights clearer differences between some of the processes
• Using this information one can eliminate runs that appeared previously to be consistent and therefore make more informed conclusions of the ‘predictability’ of the process conditions
• Further processing of data is still required to fully analyse these results and gain a greater confidence in the ‘process model’
Biovolume profiles for Bioreactor runs
Run Time Hrs
0 100 200 300 400
Bio
vo
lum
e P
0
10000
20000
30000
40000
50000
Multivariate analysis
• PCA modelling can then be used to further interpret the key components and specific characteristics of individual batches
• Pre-processing of the raw data is critical in this sort of analysis in order to ensure that inferences from the statistical tools are ‘genuine’
• This analysis then offers a method for investigating design space and process ‘limits’ suitable for development and scale up
Understanding the data
18 February 2014 Dielectric Spectroscopy and its
application in process development
Using DS in scale up/scale down experiments
Using conventional probes volumes
down to 100ml can be monitored
Samples down to 1ml can be taken from shake flasks for DS using an inverted probe
Scale up data from GSK
• Capacitance as a scale parameter from shake flask to 50L single use bioreactor
• Inverted probe technology from Aber Instruments provides real-time offline capacitance readings that correlate well with online data
Technology comparison
Correlation of Off-line v In-situ Capacitance probes at scale
Run Time Hrs
0 100 200 300 400
Measure
d C
apacita
nce
0
5
10
15
20
25
Via
ble
Ce
ll D
ensity
0
2
4
6
8
10
12
2L Bioreactor
50L SUB
Offline Probe (1L S/Flask)
Viable Cell Density (Flask)
Case study 3
Fujifilm Diosynth, UK
y = 0.663x + 0.4493R² = 0.9949
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16 18
Via
ble
ce
ll C
on
cen
trat
ion
(x 1
06 )
ce
ll/m
L
Capacitance (pF)
Conventional Calibration showing online cell diameter from ABER SCADA software
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0
2
4
6
8
10
12
14
16
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 20.00
Ce
ll D
iam
ete
r (A
rbit
rary
un
its)
Via
ble
Ce
ll C
on
cen
trat
ion
(x
10
6/
mL
Elapsed Time (Days)
Online VCC
Offline VCC
Cell size
Offline vs on line Viable cell concentrationusing Aber calibration method
• PCA analysis of Raw B-dispersion Spectra from CHO Mammalian cell culture
• P1 – Strong Biomass concentration signal
• PC2 and 3 Remaining Orthogonal variance in signal
Principal Component Analysis Raw scores plot of B-Dispersion Spectra
-10
-8
-6
-4
-2
0
2
4
6
8
-40
-30
-20
-10
0
10
20
30
40
0 2 4 6 8 10 12 14 16 18 20
PC
2 a
nd
PC
3 S
core
s
PC
1 S
core
Elapsed time (days)
P1
P2
P3
VCD- Variable importance plots vs Frequency
32
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
Capaci5
0C
apaci6
4C
apaci8
2C
apaci1
06
Capaci1
36
Capaci1
74
Capaci2
24
Capaci2
87
Capaci3
68
Capaci4
73
Capaci6
07
Capaci7
79
Capaci1
000
Capaci1
284
Capaci1
648
Capaci2
115
Capaci2
714
Capaci3
484
Capaci4
472
Capaci5
740
Capaci7
368
Capaci9
457
Capaci1
2139
Capaci1
5582
Capaci2
0000
Conduc50
Conduc64
Conduc82
Conduc106
Conduc136
Conduc174
Conduc224
Conduc287
Conduc368
Conduc473
Conduc607
Conduc779
Conduc1000
Conduc1284
Conduc1648
Conduc2115
Conduc2714
Conduc3484
Conduc4472
Conduc5740
Conduc7368
Conduc9457
Conduc12139
Conduc15582
Conduc20000
VIP
[3]
Var ID (Primary)
131216 PLS predictionset.M2 (PLS), Viable cells only
VIP[Comp. 3]
SIMCA-P+ 11.5 - 09/01/2014 20:28:52
Capacitance 50-20K Hz
For VCD the lower
frequencies correlate more
strongly than the higher
frequencies as expected
Viability-Variable importance plots vs Frequency
33
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
2.1
Capaci5
0C
apaci6
4C
apaci8
2C
apaci1
06
Capaci1
36
Capaci1
74
Capaci2
24
Capaci2
87
Capaci3
68
Capaci4
73
Capaci6
07
Capaci7
79
Capaci1
000
Capaci1
284
Capaci1
648
Capaci2
115
Capaci2
714
Capaci3
484
Capaci4
472
Capaci5
740
Capaci7
368
Capaci9
457
Capaci1
213
Capaci1
558
Capaci2
000
Conduc50
Conduc64
Conduc82
Conduc106
Conduc136
Conduc174
Conduc224
Conduc287
Conduc368
Conduc473
Conduc607
Conduc779
Conduc1000
Conduc1284
Conduc1648
Conduc2115
Conduc2714
Conduc3484
Conduc4472
Conduc5740
Conduc7368
Conduc9457
Conduc1213
Conduc1558
Conduc2000
VIP
[6]
Var ID (Primary)
131216 PLS predictionset.M3 (PLS), % Viability Only
VIP[Comp. 6]
SIMCA-P+ 11.5 - 09/01/2014 20:29:14
-Two peaks in the VIP plot vs
frequency with a peak in the
mid range.
-change in the shape of the
peak in the mid range over
time probably due to a shift in
the cell size distribution.
Capacitance 50-20K Hz
PLS Calibration
0
20
40
60
80
100
120
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10 12 14 16 18 20
% C
ulture
Via
bili
ty
Cell
Concentr
ation
(x 1
0^6
cells
/mL)
Elapsed Time (Days)
Viable cell Conc prediction Total Cell Conc Prediction "offline VCC"
% Viability Prediction "offline % Viability"
Single use applications
Results from Aber single use probe in a Sartorius rocking bag
0 48 96 144 192 240 288
0
10
20
30
40
50
60
70
0
50
100
150
200
250
300
0
1
2
3
4
5
6
7
8
5
10
15
20
Ca
pa
cita
nce
[p
F/c
m]
Time [h]
Ce
ll D
ensity [
10^5/m
l]
Via
ble
Ce
ll V
olu
me
[%
]
Ce
ll D
iam
ete
r [µ
m]
Flow past cells for use on external loops with silicone tubing
First version of Flow past cell will be tested at Shire this month
Conclusions
• Spectral scanning of capacitance becoming increasingly used in PAT
• Software based on Cole-Cole or PLS can be used to correct VCD during death phase ,derive viability and in scale up studies
• Promising technology for studies on apoptosis on certain cell lines
• Development of single use probes has enabled technology to be applied to single use bioreactors