Purification Design

and

Scale up
for mAbs

Daniel Karrer

Scale up Issue: Change in Priorities

Development Pipeline
Preclinical

Phase 1

Phase 2

Phase 3

Speed to Market

Fast and simple

Clinical Batches

Cost is secondary

Robust / Simple

Purity
Validateable

Scale up

Supply Market

Develop Process

approved

Priority
Change

GMP Consistent
Generic Process Model

Economic

Flexibility / Quick
changeover
High throughput / yield
Low running cost / risk

"Got a few problems with our linear scale up."

Robust mAb Process:

Generic Purification Template

mAbs are sufficiently

similar to permit a
template approach

Reduces development
time

sufficiently different that

there always has to be
optimization

Process parameters,
harvest, stability, dosage

Impurity
Culture Fluid Clarification Capture Purification Polishing Virus
Cell Debris
Colloids
DNA
HCP
Virus
Isoforms
Ligand

UF

Culture
Fluid

Minimum of steps for

high overall yield !
Final Filling

Key for Economic Design/Scale-up in DSP:

Low production cost
Throughput

Production Qty
$/gram

Process/Cycle time

Equipment size

Process Parameters

Drives utilization of equipment and facilities

needed
Drives fixed cost and headcount
Optimal process design ( min volumes,
surfaces and quantities) to reduce cycle time
Drive running/expandable cost

Cost

Total costs

Fixed costs
Variable costs

Purity/Quality/Yield

Drives batches needed /annum

Throughput

Increasing throughput leverages fixed cost

Key for Economic Design/Scale-up in DSP:

Know Your Expandable cost
Expendables/batch

Cost in k$
10.0

One time use

2.0

One time use

Buffer Filtration

10.0

One time use

Intermediate bulk sterile filtration

4.0

Air

0.8

40 k$/50 batches

4.5

240 m2 Prostak / yr

4
2
4
11

8x 16Millistak / batch
Polysep 9 x 30
Durapore 9 x 30

Media filtration (non serum free)

Capsules (ferm additives)

Prim Sep. /Perf.

Clarification

MF (Prostak)

Depth
Clarification
Sterile

Chromo AF

Ultrafiltration

3.5
8
2

Final aseptic filtration

0.3

Chromo IEX
Virus

TOTAL

66.1 k$

Base :
10 m3 batch
100 batches/yr
0.2g/l

100 l over 9 mts (4 x 75

cycles)
2 x 400 l/yr
4 x 10 NFP
75 m2 over 1 yrs

Media and Buffer Preparation Scale-up

Media Prep
Buffer Prep

Media and Buffer Filtration Scale-up:

Performance impacting Economy

High flow / high capacity filters needed

Superiority of PES membranes for large volume filtrations

Double flux due to composite asymetric membrane

Larger capacity due to second layer for plugging streams
Sterility assurance, low prot binding and caustic stability

Multiple use (buffer filtration)

Balance expandable cost with utility cost (WFI, Steam)

Relative Flux of PES membranes

Throughput (Filtr. Capacity) of Filters

Cell Culture and Clarification Scale-up

Media prep

Cell
culture

Prim Sep
Clarification

Clarification Scale-up / Design Strategy

SeparationTechnology :
Determines at which scale process can work

Cell Culture Type:

Determines which clarific.
technologies will work

Scale up
Strategy

Manufacturing
Challenges:
Determine which
technologies are economical

Mammalian Cell Culture Types

Affecting downstream separation design
Culture
Type

Batch
3-7 days

Fed-batch
7-15 days

Perfusion
> 20 days

Solids

Low
<1%

Medium
2-3%

High
>3%

Low

Medium to high

High

<3 x 10 cells/ml

5-15 x 10 cells/ml

>10 x 10 cells/ml

Cell Viability

High
>90%

Medium to High
20-90%

Medium to Low
<50%

Colloids

Low

Medium to High

High

Turbidity

Low
<200 NTU

High
>1000 NTU

Very High
>>1000 NTU

Ease of
Clarification

EASY

MEDIUM

DIFFICULT

Cell Density

Clarification Train Technologies:

Impact of Cell Culture Type
Primary
Separation

Sterilizing Filter
Secondary
(bioburden and colloids)
Depth/Prefiltration
(debris and colloids)

Perfusion Spin Filter

(whole cells)

Bioreactor
PRODUCT
Bioburden
Particles
Colloids
Cells
Debris

Fed Batch
Batch

AF
Chrom.
Centrif/Clarif.
(whole cells)

Clarification
(whole cells/cell
debris)

Clarification Technology

Manufacturing Challenge: Economy

Batch Size and Frequency

High
Disposables
low capital $

Manufacturing

Lab Scale
NFF
TFF

Low
Disposables
High capital $

Centrifuge
< 10 batches/yr,
<1,000 Liters

>> 10 batches/yr,
>>1,000 Liters

Primary Clarification Step

Perfusion Technology: Large scale

MF 60 m2 media perfusion system

-

SIP/CIP integrated
Low shear forces
Scalable to large scale

TFF Clarification : Prostak MF

Micro filtration unit: 280 m2, for 12m3 Batch

excellent filtrate quality

up to 99.5 % cell removal
up to 800 g/l wet weight
possibility of cell washing
easy scaling up

reduced final filtration needs

Secondary Clarification Methods

NFF clarification:

Adsorptive Depth Filters ( Pad

Filters) with high dirt load capacity
Pleated Prefilters

Large amounts of colloids limit

pleated filter capacity
Staged approach required

Typically two or three stages

Filtration Centre:
Flexibility is key

For multitechnology / multiproduct facility

2 Depth
Filter Steps

2 Pre- Filter
Steps

2 Sterile
Filter Steps

Filtration Centre :Design Concept

Adsorptive depth filters

Current issues
Premature particulate
breakthrough after 1st step

20

NTU

25

Increase of SF surface

No automated control
Difficult, costly and time
consuming handling

15
10
5
0
0

Sterile
0.22 micron
as a function vs
of Feed
turbidity
Sterile
0.2
umforCapacity
capacity
Turbidity
Mammalian Cells

10000
l/m^2 = 1322.1(NTU)-1.5022
R2 = 0.8301
1000
L/m^2

Turbidity breakthrough

100

10
0.1

1.0

NTU

10.0

100.0

100

200

L.M-2

300

Secondary Prefiltration:

Filter Train Optimization: Elimination of filtr.

step
Primary
Separation

Millistak HC
High Capacity
Depth Filter

Secondary
Prefiltration

Sterilizing Filter
(bioburden and colloids)

(debris and colloids)

Open grade
Tighter grade
Cellulosic Spin Filter
Membrane (whole cells)

AF
Chrom.

Flow

0.025

Centrifugation

0.22 Durapore Resistance

(PSID/LMH)

0.020

Standard Millistak
Technology
Multilayer Millistak
Technology

0.015
0.010

(whole cells)

0.005

Clarification

0.000
0

200
400
600
800
1000
0.22 Durapore Throughput (L/m2)

1200

1400

(whole cells)

Guarded Sterilizing Filter

(bioburden and colloids)

Sterile Membrane Capacity

Optimization of filtration parameters

Combined flow/pressure control

Increase of throughput

Karl Schick/SciLog Middleton/WI

Filtration Centre :

Flexible Design Concept

Autom. Filter Train
2-3 stage filtration
Flexible design (serial/parallel op)
CIP/SIP automated
Integrated Integrity Testing
Fully automated
Algorithms for pressure/flow/turbidity
control
Fast change over

Multipurpose / Multiproduct design

Clarification Step:

Scale up Recommendations

Know important feedstock variables

Optimize

Turbidity, Packed cell volume, Viability, Filterability (Vmax/Pmax/Tmax)

As multistep integrated process

Scale with

Consistent culture age and hold time windows

Consistent centrate residence times
Consistent feed volume/filter area
Feed representative for large scale or worst case operation

Good clarification guarantees long lifetime of capture media and

keeps media cost under control

Down stream processing Scale-up

Media prep

Purification
Polish

Capture

Capture Step: Affinity Protein A

Benefits

High purity in one

generic step (>98%)
High capacity - low
media and buffer usage
High throughput
Rapid equilibration min.
sample losses
Lifetime: 300 cycles
Potential for viral
reduction/inactivation

Capture process limitations

Case:

batch of MAb
dynamic binding capacity 20 g/L resin
10 L bed (25 cm diameter x 21cm bed height)

1993: 0.1 g/L Expression in 2000 L

Flow rate 600 cm/hr (PROSEP A): load time 6.8 hours

Process step is volume limited: need high flow

2003: 1.0 g/L Expression in 10 x 200 L

Flow rate 600cm/ hr (PROSEP A): Load time 40 minutes

Process step is capacity limited: need high capacity

Increasing capacity :

Increase resin surface area of CPG

700 Pore Diameter PG
Prosep Ultra

1000 Pore Diameter PG

Dynamic Capacity (10% BT)

(1.0 mg/cm3 hIgG feed

Dynamic Capacity (mg/cm )

Prosep-A
70

20 cm, 0.66 cm ID

60

17 cm, 1.1 cm ID

50

5 cm, 0.66 cm ID
5 cm, 1.1 cm ID

40

20 cm,0.66 cm ID

30

20 cm, 1.1 cm ID
10 cm, 1.1 cm ID

20
10
0
0

10

12

14

16

18

20

Residence Time (min)

5000

5000

McCue, J.T., et.al., J Chromatogr A, 989 (2003) 139

Protein A porous glass (PG) support

Narrow pore size distribution

Protein A immobilized onto PG support

5 cm, 0.66 cm ID
5 cm, 1.1 cm ID

Capacity and Throughput

Productivity vs Residence Time
25
PAHC
PAHC (optimised)

Productivity (g/hr)

20

Agarose
Agarose (optimised)

15

10

0
0

10

15

20

25

Residence time (min)

Consider throughput as well as capacity !!

at least 2 x capacity needed to compensate double residence time

Increasing Throughput (Flow):

Pressure-flow characteristics
Different bed diameters

0.30

0.25
0.20

0.20

10 cm

Agarose
CPG

0.10

44 cm

0.10

9 cm

0.05

25 cm

0.00
0

200

400
600
800
Mobile phase velocity (cm/h)

1000

Agarose has relatively higher back pressure

Increasing slope shows compressibility (wall

effects)
Back pressure varies with column diameter

CPG has relatively low back pressure

20 cm
30 cm
50 cm

0.15

0.15

0.05

Bed height

P re s s u re d ro p (M P a )

30 cm

0.25
Pressure drop (M Pa)

44 cm bed diameter CPG

Constant slope shows incompressibility

Back pressure insensitive to column diameter

0.00

1200

500
1000
Mobile phase velocity (cm/h)

CPG has a relatively low back pressure

even at 50 cm bed and 800 cm/h

Pressure-flow constraint not as limiting

Allows for greater flexibility in column
selection

1500

Scale-up :

Methods for more flexibility

Conventional scale-up

Residence Time scale-up

Increase diameter
Constant bed height and linear
flow rate
Linear increase of bedheight and
flow
Pressure-flow limitations due to
compressibility of conv. media

Combined scale-up

Combined
scale-up

Modern media are

incompressible and maintain a
low pressure drop
Permits scale up in any
dimension at constant residence
time
Greater flexibility for equipment

Residence time
scale-up

Conventional
scale-up

Scale up:

Constant residence time

Columns with same bed
volume and different bed
dimensions
Same dynamic capacities
Overlapping
chromatograms

3000

VL16, 19cm bed

VL22, 10cm bed

2500

VL44, 2.5cm bed

2000

AU (280nm)

1500
1000
500
0
0

50
100
mg Ab/ml media

Column
diameter

Bed Height
(mm)

Bed volume
(ml)

Residence
time (sec)

Flow velocity
(cm.h-1)

Dynamic capacity
(mg.ml-1)

44
22
16

25
100
190

38
38
38

90
91
91

100
395
750

17.8
17.7
16.2

Example of a combined column scale-up

Conventional

47 l

Alternative 1

470 l

Alternative 2

470 l

470 l

Scale Up Pilot Column by 10x to a Bed Volume of 470 litres

Initial
Chromatography
Equipment Cost
($ (Millions))

Scale Up
Method

Media
Volume
(Liters)

Bed
Height
(cm)

Bed
Diameter
(cm)

Production
Rate
(g MAB/hr)

Chromatography
Equipment
Depreciation
Costs ($/batch)

Conventional

470

15

200

726

1500

1.5

Alternative 1

470

30

140

701

800

0.8

Alternative 2

470

45

115

681

600

0.6

PG 700, 1.0 mg/ml MAb, 10 KL batch, 10 yr column life, 300 cycle resin life, 100 batch/yr

Capital Cost columns

Exponential cost increase with diameter

200.00
180.00

Cost (k $)

160.00
140.00
120.00
100.00
80.00
60.00
40.00
20.00
0.00
0

200

400

600

800

1000

1200

Column Diameter (mm)

Isopak column stainless steel

1400

Media Cost in columns

Exponential cost increase with diameter

2'500.00

Cost (k $)

2'000.00
1'500.00
1'000.00
500.00
0.00
0

200

400

600

800

1000

1200

Column Diameter (mm)

Prosep A affinity media

1400

1600

Example of a Column Scale Up:

Split batch

Alternative

Conventional

10 cycles

Scale Up Pilot Column by 10x to a Bed Volume of 470 litres

Media
Volume
(Liters)

Bed
Height
(cm)

Bed
Diameter
(cm)

Prod
Rate
(g MAB/hr)

Chromat
Media
Costs
($/column)

Conventional

470

15

200

726

3.76

12500

1.5

Alternative

47

15

63

73

0.37

12500

0.6

Scale Up
Method

Chromat
Media
Costs
($/batch)

Tradeof of time against cost and risk

PG 700, 1.0 mg/ml MAb, 10 KL batch, 10 yr column life, 300 cycle resin life, 100 batch/yr

Chromat
Equip Cost
($ (Millions))

Optimised Hardware Designs

Compact modern valve designs

replace 3 -way valves with

modular valve clusters

Combined bubble trap/filter

decreases dead volumes,

piping and improves residence
time and yields

Capture Step :

Scale up Recommendations

Dynamic Capacity

Combined Scale-up

Chose first media with high dynamic capacity at low residence time
(micro particulate packing)

Optimize for above media bed height at minimal residence time and
acceptable pressure drop
Calculate column diameter needed for full scale

Split batches

Splitting of batch volume to reduce capital cost and risk using one or
more columns of smaller diameter with optimized bed height

Down stream processing Scale-up

Media prep

Virus removal

Virus Clearance

Virus Filtration

Trend to Sgmall Virus Filltration

-

Reliable- Removal based on size

Parvo Virus removal

High LRVs ( 4-6) readily achievable

Robust

Viral clearance insensitive to

operating conditions (pressure,Temp)

Viresolve NFP( 72 L/M2, disks)

High Yield

Recovery ca 98%

Short Cycle time

Fluid
Pressure (psi)
Mab (2.5 g/l)
30
DMEM
30
Mab (2.5 g/l)
45
DMEM
45

PPV LRV
> 4.7
> 6.5
> 5.7
> 6.2

Virus Filter Sizing on Flow or Capacity ?

m2 =
L

Sizing is
determined by flux
& capacity
High Flux is key for
low process time
at comparable filter
surfaces

Vmax

1
J*t

Vmax

8.0

sqm/100 L

Flux J

7.0

Filter A: 1000 L/sqm, 15 LMH

6.0
5.0

Filter B: 200 L/sqm, 100 LMH

4.0
3.0
2.0
1.0
0.0
0

6
hrs process

10

12

Virus Filter :Sizing on Capacity

PhiX174 LRV

8.0
7.0
6.0
5.0
Blue Dextran
4.0
3.0
2.0
1 ppm
100 ppm
1.0
500 ppm
0.0
0 10 20 30 40 50 60 70 80 90 100
Percent Flow Decay

LRV correlated with flow decay

(Hirasahi, Kempf)

Dont exceed 50% flow decay

(75% Vmax) when scaling-up
Safety margin needed

Same behavior with other

plugging agents
Not expected for adsorption
High LRV maintained, using
clean feedstock's
(Prefiltration)

Virus Filter : Influence of Protein Concentration

Relative Membrane Area Requirements as a function of Protein
Concentration

Viresolve NFP
30 psi TMP
Relative Membrane Area Requirements
(m^2 /m^2 optimal)

Vmax ( l/m^2)

900
800
700
600
500
400
300
200
100
0
1

IgG 1
IgG 2
IgG 3
BSA 1
BSA 2

10

100

Concentration (g/l)
Capacity (Vmax) can decline
with protein concentration

1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
0

10

15

IgG Concentration ( g/l)

Area & Cost are minimal at

concentrations of 8-10 g/l

20

Down stream processing Scale-up

Media prep

UF / DF

Pellicon UF Cassettes Scale-up:

An example of true linear scalability

Process Scale Holder

1 -80 m
5 - 1000s liters

50cm
5ml - 1000 ml
Pellicon XL

0.1, 0.5, 2.5m

200ml - 100s liters
Pellicon Cassettes

Pellicon High Performance Membranes:

Reduction of Process Time and Holdup Volume
Automated UF/DF 15m2 System
Conv. Polysulfon
Biomax 10

10kd

Retention

>99.95%

99.9%

Flux

118 lmh

80 lmh

Recirc.

4 lpm

6 lpm

Line Size

1.5

2.5

Hold Up

8.4 l

20.8 l

Yield

Plus 2 3 %

UF/DF operating parameters:

Optimal TMP for process time reduction

Optimization of Trans
Membrane Pressure

Reduction of process
time
Reduction of
membrane surface

Flux [L mh]

Initial (low )Prot. Concentration

Optimum Operating
Point

Final (high)
Prot. Concentration

Transmembrane Pressure [bar]

Opt. TMP for Retained Products

Run at knee of flux curve at highest

concentration
Reduced aggregation, fouling, cleaning
issues

DF Strategy

Reduction of process time and membrane surface

Buffer
Feed

Concentrate

Diafilter

Permeate

Permeate

Concentrate

Flux

Retentate

Turbidity

Optimal Diafiltration Point

At Cdiaf = Cg/e ~ 80 - 100 g/L, Minimal Filter

Area/Process Time
At Higher Conc. : Lower Buffer Use, but low
Flux
At Lower Conc. : High Buffer use, high Flux,
less Aggregate Formation

Retention
Cg/e

ln(g/L protein)

Cg

UF/DF Step:

Scaling Difficulties HW / Design related

UF/DF Step Challenges :

Trend to very high protein endconcentrations

(200 g/l) for liquid formulations, requesting
lowest possible holdup volumes

Reachable Endconcentration / Purity :

depends on overall design/membrane
surface

DF volumes needed can be impacted by

system design (mixing)

Higher viscosities creating difficulties (150

CP) Elevated temperatures to be used to
lower viscosity.

Aggregation/foaming issues impotant for

tank/pump design

Product recovery is key (1% yield = 0.5 Mio

$ for 10 kg -batch)

UF/DF System : integrated designs for low

final volumes

Integrated tank/pump/holder design

reducing hold-up volume
Instrument blocks to reduce piping
length
Modular holders for Flexibility

Process Development Summary

Key elements for Scale-up:

Visualize full scale process and scale down first

Visualize all process steps : integrated approach

Dont forget productivity,yield and economy already at

development scale

10 m3 Batch value 40 Mio $ (10 kg protein)

Make your experiments at small scale and the

profits on large scale!!

Thank You

Thank You

Acknowledgments/Reference

Glen Kemp, Millipore

Herb Lutz, Millipore
Karl Schick, SciLog
Fred Mann, Millipore
Duncan Low, Amgen

Bio

Project GmbH
Project Management
at it's best

Process Design/Optimisation
Scale up
Process Reviews / Economy
Feasability Studies / Technology
assessements
Engineering Concepts
Technology Transfers / Outsourcing