Downstream Cap 05 abstracts - … Downstream – Cap ’05 abstracts Clariﬁcation of biomass harvest from a perfusion bioreactor Fudu Miao, Dan Borginis, David Jen, and Andrew Ramelmeier

GE Healthcare

Downstream

Cap ’05 abstractsExtended Reports from the International Conference on Initial Recovery and Capture Techniques

Phoenix, Arizona, USA, May 22nd–25th, 2005

Downstream – Cap ’05 abstracts 3

5 From the chairman

Oral presentations

6 Clarification of biomass harvest from a perfusion bioreactor

9 Implementation of novel affinity ligands for biotherapeutic purification

12 Development of new affinity adsorbents for primary capture and purification of therapeutic proteins

14 Integrated capture process for purification of plasmid DNA based on aqueous two-phase separation

17 Metal affinity systems for the integrated purification and refolding of recombinant D5-3-ketosteroid isomerase

20 Development of a multivariate statistical model for accurate scale-up predictions of pressure/flow properties

22 Primary recovery – assessment of monoclonal antibody containing harvests to predict clarification requirements

25 Hydrodynamic characterization, comparison, and scaling-up of a cleaning protocol for severe fouling of an expanded bed column following the application of Escherichia coli homogenate

27 Establishing large-scale expanded bed adsorption technology in the first step purification of industrial enzymes

32 List of posters presented

34 Author index

In this issue

4 Downstream – Cap ’05 abstracts


From the chairman

The Cap ´05 meeting was a follow-on from the successful EBA conference series, but with a much broader focus on technologies at the interface between cell culture and downstream purification. Both oral and poster presenters contributed to a fascinating meeting, addressing challenges and recent advances. The meeting generated stimulating discussion and great enthusiasm around the critical capture phase of a downstream purification process.

Cap ´05 achieved its goal of bringing together industry and academia, balancing theoretical with application, as well as introducing new products and new uses for old products – all over an intensive three-day period.

This booklet contains some examples of the exciting presentations given during the conference that made Cap ´05 a much-appreciated event and a meeting to remember.

Rhona O´Leary Conference Chairperson


Clarification of biomass harvest from a perfusion bioreactorFudu Miao, Dan Borginis, David Jen, and Andrew Ramelmeier

Centocor, 260 Great Valley Parkway, Malvern, PA 19355.

e-mail: [email protected]

Extended AbstractPerfusion cell culture is a highly productive approach of producing biologics for therapeutic use. In perfusion cell culture, cells are separated from medium with a cell retention device. The cells return to the bioreactor, while the medium containing the product is harvested. The cell retention device is a key element for the success of the perfusion cell culture. Tangential flow filtration (TFF) with hollow fiber membranes is an effective cell retention approach. The most suitable hollow fiber membranes have rated pore sizes of 0.22 µm. The membranes have a complete retention of mammalian cells, thus achieving a high cell density rapidly. In addition, the permeate from the 0.22 µm membranes is cell-free; therefore no clarification is required. However, the bioreactor has to be “bled” to control cell density and to reduce the cell debris level in the bioreactor.

The biomass bleed stream in the TFF perfusion bioreactor usually accounts for 5–15% of the overall harvest. A lower bleed rate results in a higher cell density and a higher level of cell debris (due to a higher residence time of cells in the bioreactor). Disposing of this stream is usually not economical in commercial production. However, clarifying the biomass bleed stream is a significant separation challenge, because the stream has a cell density and cell debris level much higher than those in other types of cell culture, such as batch and fed-batch cell culture.

Three clarification strategies are proposed in this presentation: clarify the mixed stream of biomass and TFF permeate, clarify the biomass bleed stream inline, and clarify the biomass stream offline. The clarification performance in these strategies was evaluated using Cuno depth filters. The impact of biomass settling, feed rate, and filter charge on clarification performance was examined.

Four Zeta Plus depth filters from Cuno were used for the clarification study. Among them, 60M05 (30/60SP), 60M03 (10/60SP), and 30M03 (10/30SP) are double layer filters, while 10SP is a single layer filter. Cuno 60M05 has the same bottom layer as 60M03 while a tighter top layer than 60M03. Cuno 30M03 has the same top layer as 60M03 while a more open bottom layer than 60M03. Table 1 shows the approximate pore sizes of the Cuno depth filters.

The capacity of the Cuno depth filters was evaluated by monitoring the variation of the pressure drop with filtrate volume. Figure 1a illustrates that Cuno 10SP, 10/30SP and 10/60SP have a capacity of over 500 l/m2 in the clarification of mixed harvest of biomass and TFF permeate. The capacity of the Cuno 30/60SP is substantially lower than the other Cuno depth filters because this filter is the tightest. Figure 1b shows the overall turbidity of the Cuno filtrate and the maximum capacity (Vmax) of filtering the Cuno filtrate with a 0.22 µm membrane polishing filter (Durapore™, Millipore). A tighter depth filter results in lower filtrate turbidity thus a higher Vmax on the membrane polishing filter. Figure 1 indicates that Cuno10/60SP is the most suitable because it has turbidity lower than and a capacity as high as 10SP and 10/30SP.

Table 1. The approximate pore sizes of the Cuno depth filters used in this study

Cuno depth filters Filter type Approximate pore size, µm

60M05 (30/60SP) Double layer 0.3 to 3



10SP Single layer 1 to 5


0

2

4

6

8

10

12

14

16

0 100 200 300 400 500 600 700

Normalized Filtrate Volume (l/m2)

Pres

sure

(psi

)Pr

essu

re (p

si)

10SP10/30SP10/60SP30/60SP

0

50

100

150

200

250

300

350

400

450

0 5 10 15 20

Turbidity of filtrate (NTU)

V max

on

0.22

um

filte

r (l/m

2 )

10/60SP

10/30SP

30/60SP

10SP

0

2

4

6

8

10

12

14

0 50 100 150

Normalized Filtration Volume (l/m2)

10/30SP

10/60SP

0

10

20

30

40

50

60

70

80

10/30SP 10/60SP

Filters

Ove

rall

turb

idity

of t

he fi

ltrat

e (N

TU)

Vmax = 25 l/m2

Vmax=127 l/m2

0

2

4

6

8

10

12

14

16

0 100 200 300 400 500 600 700


Pres

sure

(psi

)Pr

essu

re (p

si)

10SP10/30SP10/60SP30/60SP

0

50

100

150

200

250

300

350

400

450

0 5 10 15 20


V max

on

0.22

um

filte

r (l/m

2 )

10/60SP

10/30SP

30/60SP

10SP

0

2

4

6

8

10

12

14

0 50 100 150


10/30SP

10/60SP

0

10

20

30

40

50

60

70

80

10/30SP 10/60SP

Filters

Ove

rall

turb

idity

of t

he fi

ltrat

e (N

TU)

Vmax = 25 l/m2

Vmax=127 l/m2

0

2

4

6

8

10

12

14

16

0 100 200 300 400 500 600 700


Pres

sure

(psi

)Pr

essu

re (p

si)

10SP10/30SP10/60SP30/60SP

0

50

100

150

200

250

300

350

400

450

0 5 10 15 20


V max

on

0.22

um

filte

r (l/m

2 )

10/60SP

10/30SP

30/60SP

10SP

0

2

4

6

8

10

12

14

0 50 100 150


10/30SP

10/60SP

0

10

20

30

40

50

60

70

80

10/30SP 10/60SP

Filters

Ove

rall

turb

idity

of t

he fi

ltrat

e (N

TU)

Vmax = 25 l/m2

Vmax=127 l/m2

0

2

4

6

8

10

12

14

16

0 100 200 300 400 500 600 700


Pres

sure

(psi

)Pr

essu

re (p

si)

10SP10/30SP10/60SP30/60SP

0

50

100

150

200

250

300

350

400

450

0 5 10 15 20


V max

on

0.22

um

filte

r (l/m

2 )

10/60SP

10/30SP

30/60SP

10SP

0

2

4

6

8

10

12

14

0 50 100 150


10/30SP

10/60SP

0

10

20

30

40

50

60

70

80

10/30SP 10/60SP

Filters

Ove

rall

turb

idity

of t

he fi

ltrat

e (N

TU)

Vmax = 25 l/m2

Vmax=127 l/m2

Figure 3a shows that the capacity of Cuno 10/60SP was approximately doubled when the biomass harvest was settled before clarification. Moreover, the capacity of a filter increases with decreasing feed rate, as shown in Figure 3b. A low flow rate should be used in the clarification of biomass harvest for the TFF aided perfusion cell culture. The clarification at a low feed rate is feasible because the harvesting rate of biomass is very small.

Figure 2 shows the capacity of the depth filters, Cuno 10/60SP and 10/30SP, in the clarification of biomass. Cuno 10/30SP has a higher capacity than 10/60SP because it has a more open top layer. As a result, the filtrate from Cuno 10/30SP has a higher turbidity (a lower capacity on 0.22 µm membrane polishing filter) than 10/60SP. Comparison of Figure 2 with Figure 1 reveals that there is no advantage of clarifying the mixed harvest of biomass and TFF permeate in terms of total amount of biomass filtered.

Fig 1. Clarification of mixed harvest of 15% biomass with 85% TFF permeate with Cuno depth filters.

a. Capacity of depth filter. b. Capacity of the filtrate on polishing filter.

a. Capacity in the filtration of biomass only. b. Overall turbidity of the filtrate.

Fig 2. Clarification of biomass directly at a feed flux of 3 L/(m2min)


16

0

10

20

30

40

50

60

70

60ZA05A 60M05

Clarification filter

Ove

rall

filtr

ate

turb

idity

(NTU

)

0

2

4

6

8

10

12

14

0 50 100 150

Normalized filtrate volume (l/m2)

0 50 100 150 200


Pres

sure

dro

p (p

si)


Pres

sure

(psi

)

settledunsettled

Feed flux = 3 l/(m2min)

0

2

4

6

8

10

12

14

16

Pres

sure

dro

p (p

si) 3 l/(m2min)

1 l/(m2min)

0

5

10

15

20

25

0 20 40 60 80 100 120

60ZA05A

60M05

16

0

10

20

30

40

50

60

70

60ZA05A 60M05


Ove

rall

filtr

ate

turb

idity

(NTU

)

0

2

4

6

8

10

12

14

0 50 100 150


0 50 100 150 200


Pres

sure

dro

p (p

si)


Pres

sure

(psi

)

settledunsettled


0

2

4

6

8

10

12

14

16

Pres

sure

dro

p (p

si) 3 l/(m2min)

1 l/(m2min)

0

5

10

15

20

25

0 20 40 60 80 100 120

60ZA05A

60M05

16

0

10

20

30

40

50

60

70

60ZA05A 60M05


Ove

rall

filtr

ate

turb

idity

(NTU

)

0

2

4

6

8

10

12

14

0 50 100 150


0 50 100 150 200


Pres

sure

dro

p (p

si)


Pres

sure

(psi

)

settledunsettled


0

2

4

6

8

10

12

14

16

Pres

sure

dro

p (p

si) 3 l/(m2min)

1 l/(m2min)

0

5

10

15

20

25

0 20 40 60 80 100 120

60ZA05A

60M05

16

0

10

20

30

40

50

60

70

60ZA05A 60M05


Ove

rall

filtr

ate

turb

idity

(NTU

)

0

2

4

6

8

10

12

14

0 50 100 150


0 50 100 150 200


Pres

sure

dro

p (p

si)


Pres

sure

(psi

)

settledunsettled


0

2

4

6

8

10

12

14

16

Pres

sure

dro

p (p

si) 3 l/(m2min)

1 l/(m2min)

0

5

10

15

20

25

0 20 40 60 80 100 120

60ZA05A

60M05

Another important factor affecting the clarification performance is the charge intensity on the filter. The depth filter has a positive charge, so it can remove negatively charged cellular components. Figure 4 shows the impact of charge on clarification for performance. In this experiment, Cuno 60M05 (30/60SP) has the same pore sizes as but less charge than 60ZA05A. The pressure drop pattern is very similar for both filters. However, the filtrate from 60ZA05A has much lower turbidity than 60M05, as shown in Figure 4b.

The clarification with TFF was also evaluated in this study. The clarification with TFF was performed using Millipore 0.65 µm and 0.22 µm PVDF membranes at a controlled permeate flux of 12 l/(m2hr). After the concentration, the biomass harvest was diafiltered for 3 retentate volumes with a diafiltration buffer to achieve a yield of 90%. The comparison of depth filtration with TFF illustrates that depth filtration has a higher product yield and greater ease of operation relative to TFF.

Fig 3. The impact of biomass settling and feed rate on the capacity of the Cuno depth filter.

Fig 4. Impact of the charge on the clarification performance of the Cuno depth filters.

a. Impact of biomass settling on filter capacity. b. Impact of feed rate on filter capacity.

a. Pressure profile for similar filters. b. Overall turbidity of the filtrate from the similar filters.


Implementation of novel affinity ligands for biotherapeutic purificationPim Hermans* and Mark ten Haaft

BAC BV, Huizerstraatweg 28, NL-1411 GP Naarden, The Netherlands.


www.captureselect.com

The burgeoning fields of healthcare and recombinant biotherapeutics require new purification methods to address the technical and economic limitations of protein A, dye mimetics, and immunochromatographic methods. In the field of affinity chromatography, we have developed a technology for the generation of affinity ligands (CaptureSelect) that can be used for purification of products, as well as scavenging of impurities from complex source materials. Our ligand technology is based on a special class of Camelid antibodies that are devoid of light chains. The variable domains of these so-called heavy chain antibodies have evolved to compensate for the lack of light chains. For instance, in order to maintain antibody diversity, the CDR regions of these single domain antibody fragments show a higher variability in amino acid composition compared to classical ones. In addition, a number of conserved hydrophobic residues, normally interacting with light chains, are replaced by charged amino acids, which prevents aggregation and increases solubility. On the basis of these unique features, we have identified single domain antibody fragments as ideal molecules to serve as affinity ligands in chromatography. These ligands of about Mr 12 000 have very compact and rigid structures, are highly soluble and show binding specificities and affinities that are comparable to classical two-chain antibody fragments.

The ligand technology is applicable to a wide variety of target molecules, ranging from complex antigen structures like bacteria and viruses, to proteins, carbohydrates, and even very small molecules like haptens and peptide tags. Target-specific ligands are isolated from expression libraries that represent the single domain antibody repertoire of a llama previously immunized with a target molecule of interest. These so-called immune

libraries are screened at the monoclonal level for ligands that specifically bind to the target molecule. The robustness of the screening process allows incorporation of specific requirements that closely relate to the final chromatography process, such as binding and elution conditions, contamination in the source material, and/or stability of the affinity ligand against cleaning agents like NaOH. Subsequently, ligands are selected and tested for affinity, selectivity, and stability in chromatography experiments. Finally the ligands are recombinantly produced at any scale in Saccharomyces cerevisiae (Baker’s yeast).

The advantages of these ligands as compared with other affinity technologies is the fact that this technology combines tuneable specificity and affinity, stability, short development times, and ease of non-animal derived production. Moreover, these affinity ligands can be very easily immobilized on different solid supports.

CaptureSelect affinity ligands for purification of human IgG antibodiesA llama was immunized with human IgG Fc fragments and the resulting affinity ligands were carefully screened for their ability to bind to all subclasses of IgG (IgG1–4). By incorporating stability towards NaOH in the screening phase we were able to select a human IgG affinity ligand which binds to all IgG subclasses and additionally, is stable towards caustic cleaning. The affinity ligand does not cross-react with IgG from other species. Subsequently the affinity ligand was immobilized on NHS Sepharose™ and we demonstrated repeated cycling with intermediate caustic cleaning for over 120 cycles when using 0.1 M NaOH.


1. Cow milk

2. Cow milk spiked with Human IgG

3. Tissue culture medium

4. Tissue culture medium spiked with Human IgG

5. Elution fraction from Cow milk

6. Elution fraction from Tissue culture medium

1 2 3 4 5 6

To demonstrate selectivity, this immobilized affinity ligand was used to purify human IgG antibodies from different source materials. Figure 1 shows that the ligand is capable of selective capture of human IgG from both cow milk and tissue culture samples in just a single step.

Fig 1. One-step purification of Human IgG from different feedstock.SDS-PAGE, Novex Gel Tricine 10–20%, CBB stained.

To demonstrate its efficacy in the field of plasma products purification, this affinity resin was used in collaboration with the Baxter company (USA) to purify intravenous Ig from fibrinogen-free plasma (see Fig 2). These data show the excellent yield of the purification step (87.6%) as well as the extremely good reduction in contaminants like IgA and human serum albumin. After the second wash step, levels of these contaminants dropped below the level of detection. Extremely important for functional recovery of IVIG is of course that the purified material should have a matching sub-class distribution with respect to the starting material. Figure 3 shows that there is a matching subclass distribution for two different pools of IVIG.

Fig 2. Purification of IVIG from plasma.NHS Sepharose™, immobilized human IgG ligand. Source material: fibrinogen free plasma. Process: 25 mg/ml load, 100 cm/h, wash 5CV, eluate I, 5CV, eluate 2, 5CV. Eq buffer PBS pH 7.4. Elution buffer 1; 0.1 M HCl/Glyc, pH 3.0. Elution buffer 2; 0.1 M HCl/Glyc, pH 2.0.

Albumin mg/ml

IgG mg/ml

IgA mg/ml

Albumin recovery (%)

IgG recovery (%)

IgA recovery (%)

Load 15.35 2.58 1.24 100 100 100

Total FT 14.65 0.19 1.13 95.4 7.4 91.1

Wash 1 2.85 0.18 0.2 9.7 3.6 8.4

Wash 2 <0.01 0.08 <0.01 0 1.6 <0.4

Elution, pH 3 <0.01 4.15 <0.01 0 87.6 0.2

Cryo Rich Plasma (1)

Eluate Pool (1)

Cryo Rich Plasma (2)

Eluate Pool (2)

IgG1 41.1% 42.7% 37.4% 42.3%

IgG2 49.8% 52.2% 56.4% 50.9%

IgG3 3.1% 2.1% 2.1% 2.1%

IgG4 5.9% 3.1% 4.2% 4.8%

Fig 3. Subclass distribution of purified IVIG as compared with starting material.

IgG subclass, percent of starting material

IgG subclass, percent in elution fraction

IgG1 57.1 1.7

IgG2 34.2 0

IgG3 4.0 0.6

IgG4 4.7 97.7

Fig 4. Purification of IgG4 from human serum.Medium: NHS-Sepharose. Elution buffer: Glycine/citric acid/HCl, pH 3


1. Cell culture media with FCS

2. Cell culture media with FCS spiked with Fab-kappa

3. Flowthrough fraction

4. Elution fraction

5. Fab-kappa sample

250 –

148 –

60 –

42 –

30 –22 –17 –

6 –

3 –

M 1 2 3 4 5 M 1 2 3 4 5

A B

As a second example for purification of human IgG antibodies we wanted to demonstrate that besides broad binding (that is, all IgG subclasses) this technology has the ability to purify single subclass IgG as well. Therefore we immunized a llama with IgG4 only, and carefully screened for affinity ligands that bind to human IgG4 only and not to other IgG subclasses and species. Figure 4 shows the excellent purification of IgG4 from serum. Besides the high recovery levels of IgG4, the level of the other IgG subclasses are lowered significantly. This work was done in collaboration with the Sanquin company (The Netherlands).

CaptureSelect affinity ligand for purification of human Ig Fab fragmentsIn addition to ligands that can be used to purify whole human antibodies, we wanted to demonstrate the ability of generating ligands that can be applied to efficiently purify fragments of antibodies, like Fab domains. For this purpose, we immunized a llama with polyclonal human IgM antibodies and then screened for affinity ligands that bind to human IgG1-kappa and not to human IgG1-lambda. This resulted in an affinity ligand that specifically binds to human light chains of the kappa family. Figure 5 clearly demonstrates the capability of this ligand to purify human Fab-kappa fragments from tissue culture medium.

Besides these examples, which relate to purification of human antibodies and fragments, we have also developed various ligands for affinity purification of other biotherapeutic products, like adeno-associated virus (AAV) and human serum albumin. The use of our technology could greatly enhance yields and purity of important therapeutic molecules and could even lead to new products isolated from valuable feed-stock like, for instance, human plasma.

Fig 5. Purification of human Fab-kappa from cell culture media.SDS-PAGE, Novex Gel Tricine 10–20 %:A) CBB stainedB) Western blot, detection by anti-kappa mouse Mab / anti-mouse-IgG-AP conjugate


Development of new affinity adsorbents for primary capture and purification of therapeutic proteinsJason Betley, Dev Baines, and Steven Burton

ProMetic BioSciences Ltd, 211 Cambridge Science Park, Milton road Cambridge CB4 OZA, UK.


IntroductionLibraries of ProMetic’s proprietary triazine-based Mimetic™ ligands are screened routinely for their ability to bind and elute specific proteins from a variety of feedstocks. The processes of ligand design, library synthesis, screening, and scale-up are discussed in the context of three screening projects carried out recently. Two of the target proteins were purified from human plasma, a highly challenging feedstock containing a high total protein concentration of a huge variety of proteins. The third target protein discussed was a therapeutic protein (amediplase) purified from a recombinant feedstock.

Ligand designWhen designing Mimetic ligands based on ProMetic’s proprietary triazine chemical scaffold, a number of approaches may be employed. If detailed structural information is available for the target protein and/or a previously identified ligand, it is possible to use computer-assisted rational design to instruct the combinatorial synthesis of candidate ligands on our Purabead™ matrix. In addition, sets of libraries spanning the entire diversity space of chemical structure are routinely screened. If no detailed structural information is available, only the latter approach is used in the first instance. Both approaches were used in the screens described.

Library synthesis and screening methodsLibraries are synthesized robotically in 96-well plate format, in fritted blocks, generating 8×8 arrays of candidate adsorbents occupying Rows A–H and Columns 1–8. The quality of each individual synthesis is monitored by appropriate in-process assays. Libraries are screened with extensive use of robotic liquid handling techniques. A schematic

diagram of the library synthesis process is shown (Fig 1). Each individual library member is treated as a miniaturized chromatography column with a packed bed volume of 0.25 ml; after equilibration of each well and application of feedstock, unbound, wash, elution, and sanitization fractions are collected by stacking the library block over 96-well deep-well fraction collection blocks. Fractions are then assayed using various high-throughput 96-well plate assays for estimation of [target protein] and [total protein]. Triplicate standard curves are generated for each fraction and each assay using columns 10 to 12 of the assay plate. Examples of assays used for these targets are Bradford Assay for Total Protein, ELISA assays, and 96-well SDS PAGE (E-PAGE™ from Invitrogen).

Specific screensTwo separate screens for candidate ligands that specifically bind and elute fibrinogen and plasminogen were carried out, using undepleted human plasma as feedstock. In a third screen, a recombinant feedstock was screened for ligands that specifically bind and elute ligands for amediplase

Fig 1. Schematic diagram of ligand library synthesis: Amines R1–R8 are added to activated triazine and arrayed in columns 1–8. Amines A–H are then added in rows A–H to generate a 64-component combinatorial library.


(a chimera of tissue plasminogen activator and a single chain urokinase plasminogen activator – The Menarini Group). For each assay, high-throughput primary screens were used to identify a small number (<100) of candidate adsorbents displaying the desired specificity of binding and elution for the target protein. A secondary screen was then employed using an expanded repertoire of higher precision analytical techniques to permit the recommendation of a number (4–6) of candidate adsorbents to scale up to the 50 ml scale. The analytical techniques used for the three screens are shown (Table 1). An example of an SDS-PAGE gel generated during primary screening for a plasminogen ligand is also shown (Fig 2).

Adsorbent developmentOnce a small number of candidate ligands (4–6) were synthesized for each target protein, their performance was tested chromatographically, and the most favorable candidate taken forward for full adsorbent development. Desirable properties for these adsorbents included specificity, binding capacity, purity of eluted protein, step yield, cost-of-goods, and ease of synthesis. After the selection of a single candidate, ligand loading and spacer arm combinations were explored, and the combination permitting the best combination of yield, binding capacity, elution capacity, and purity for the target protein was progressed through to technical transfer and manufacturing. The chromatographic performance of the best candidates identified during the three screening projects is shown (Table 2).

Table 1. Assays used in primary and secondary screening.

Screen Assays used in:Primary Screening Secondary Screening

Fibrinogen Bradford Total Protein Fibrinogen ELISA

Bradford Total Protein Fibrinogen ELISA

Nephelometry SDS PAGE

Plasminogen E-PAGE Nephelometry SDS PAGE

Amediplase Urokinase Activity Assay Bradford Total Protein

Urokinase Activity Assay Bradford Total Protein

SDS PAGE

Table 2. Performance parameters for adsorbents after adsorbent development

Preformance Fibrinogen Plasminogen Amediplase

Dynamic Binding Capacity (10% Breakthrough, g/l)

15

17

11

Step Recovery (%) > 90 > 90 92

Purity (%) > 85 > 90 > 99

Fig 2. E-PAGE 96 Sample Gel Electrophoresis on elution fractions from screening of library 254 for adsorbents specific for human plasminogen from human plasma feedstock. A1–H8: Elution samples from adsorbents A1–H8; Columns 9, 10: Blank; Columns 11, 12: Plasminogen Standard (D11, D12 highlighted in blue); AM, CM, EM, GM: Molecular Weight Markers. Promising candidates in row H are highlighted in red.

ConclusionsLibraries of specifically-designed and/or diverse Mimetic ligands were synthesized, and screened for their ability to purify specific proteins from human plasma and recombinant feedstocks. The screen combined specific and ‘Total Protein’ assays in a high-throughput fashion to generate leads for each protein. After secondary screening, verification chromatography, and adsorbent development, single candidates were progressed through to manufacturing. New projects can be progressed from inception to manufacturing in less than 12 months. All ligands discovered were robust and base stable. All three adsorbents displayed good levels of recovery and purity, and relatively low levels of non-specific binding of non-target protein.


Integrated capture process for purification of plasmid DNA based on aqueous two-phase separationAndreas Frerix1, Markus Müller2, Maria- Regina Kula3, and Jürgen Hubbuch1

1 Research Center Jülich, Institute of Biotechnology 2, BioSeparationsGroup, Jülich, Germany.2 QIAGEN GmbH, Max- Volmer- Strasse 4, Hilden, Germany.3 University of Düsseldorf, Institute of Enzyme Technology, Jülich, Germany.


IntroductionGene therapy and pDNA production

Plasmid DNA (pDNA) can be applied as a vector for gene therapy and DNA vaccination. In comparison to viral vectors, which are known to potentially cause immunological responses, pDNA is regarded as a safe method for administration of genetic material. However, the efficiency of the transfer operation lies often below one percent, while the duration of expression is relatively short. This results in the need for large quantities of pDNA often lying in the range of several milligrams per treatment [1].

Obviously, there is a high demand for economical and effective purification processes. Traditional approaches mainly based on chromatographic processes are hampered by a low capacity for pDNA and thus high costs. In contrast to this, two-phase extractions offer the advantage of high capacity, ease of scale, the use of simple technical equipment and cheap chemical materials. To date, two-phase extraction has been described for the purification of pDNA from cleared lysate using PEG/ salt [2] and thermoseparating polymers [3]. In this work, we propose an integrated process based on ATPS combining cell lysis, neutralization, solid-liquid separation, and initial purification in a single step [4].

Since pDNA is produced in E. coli, initial process streams are heavily contaminated with other nucleic acid forms such as RNA and genomic DNA (gDNA), as well as intracellular proteins, cell debris, and endotoxins. The method of choice for liberation of pDNA from E. coli cells is based on alkaline lysis followed by neutralization. As a result of this

procedure, the majority of proteins and gDNA are denatured and form a solid precipitate with cell debris while plasmids are recovered in the supernatant. In this cleared lysate, pDNA contains at least two different conformations, namely open circular and supercoiled. The latter is the actual target molecule for the purification process.

Results and discussionPartitioning studies in spiked systems

The main contaminant present in the cleared lysate is RNA. Thus the primary goal of an initial purification step is to achieve a significant reduction of RNA content. In a first screening attempt, individual partitioning of RNA and pDNA was examined by gradually changing concentrations of PEG and phosphate. For both process parameters—increasing PEG and phosphate concentrations—a shift of pDNA from the bottom to the top phase occurred while RNA either remained in the top phase or precipitated onto the interphase. This behavior clearly offers a tool for the separation of pDNA from RNA. The special partitioning behavior of pDNA was further detected when varying PEG molecular weight. Using different mixtures of PEG 600 and PEG 1000 in order to mimic different PEG molecular weights, an even more drastic shift of pDNA could be detected (Fig 1). Between PEG 650 and PEG 700, an almost complete change of pDNA partitioning from top to bottom phase was observed. By carefully optimizing the partitioning behavior, a system based on 15% PEG 800 and 20% potassium phosphate was identified leading to about 90% recovery of pDNA in the bottom phase and a 90% depletion of RNA in the top phase and interphase.


These findings conform with results shown in the agarose medium in Figure 2. Here, RNA was located in the top phase (Lane 1), while pDNA was found in bottom phase with RNA-contamination below the detection limit using ethidium bromide to stain (Lane 2). Due to high salt concentration in the bottom phase, pDNA shows a non-specific band, including scDNA and remaining ocDNA conformation. As expected, the precipitate contained RNA and gDNA (Lane 3), but also a certain loss of pDNA was observed, the latter suggesting that total pDNA yield might be increased by reducing product loss in the precipitate. Following the ATPS and recovery of the bottom phase, it is essential to reduce the salt load of the pDNA-containing phase. Desalting of plasmid DNA in the bottom phase was achieved by ultrafiltration using an Mr 100 000 cut-off membrane. After ultrafiltration, isolated pDNA shows a typical pattern of oc and scDNA forms with a slight reduction of ocDNA (Lane 5).

pDNA capture from whole biomassTaking into account the results of the partitioning studies above, an integrated three-step capture for a pDNA process was developed (Fig 3). Beginning with resuspended E. coli, biomass alkaline lysis was started by addition of NaOH and SDS containing lysis buffer. After gentle mixing and 7 min incubation, potassium phosphate buffer was added and the solution was gently inverted. In this step, potassium phosphate combined several tasks, namely neutralization of the solution to pH 7.5, precipitation of denatured proteins, gDNA, and cell debris together with the detergent and phase forming salt component for the two-phase system. The third step was the addition of a PEG 800 stock solution completing the phase-forming system. After mixing and centrifugation, two clear phases were obtained, in which the resulting cell debris containing precipitate was completely separated from the bottom phase and formed a distinct interphase.

Fig 1. Influence of PEG molecular weight on partitioning of pDNA and RNA in a system containing 15% PEG and 20% potassium phosphate (pH 7.4).

Fig 2. Agarose gel analysis of a PEG 800 (15% w/w) / potassium phosphate (20% w/w) aqueous two-phase system with biomass; combination of alkaline lysis and extraction.

lane 0: 1 Kb ladder; lane 1: top phase; lane 2: bottom phase; lane 3: precipitate resuspended in TE- buffer; lane 4: Ultra-permeate (Mr 100 000 membrane); lane 5: Ultra-retentate (Mr 100 000 membrane); lane 6: plasmid stock; lane 7: E. coli gDNA Standard (0.8% agarose gel, 10 µl per lane, 70 V).

Fig 3. Integrated Plasmid capturing in three steps using aqueous two phase systems.

Agarose gel analysis with desalted samples of top and bottom phase confirmed the results obtained from the spiked systems, showing pDNA recovered in the bottom phase while most of the RNA was found in the top phase.


For further analysis, analytical hydrophobic interaction chromatography [5] was performed by HPLC (Fig 4). The chromatogram of the cleared lysate is dominated by impurity peaks of proteins and RNA, while the analysis of the bottom phase shows a significant reduction of those impurities. Evaluating the performance, about 90% of pDNA were recovered in the bottom phase, while proteins were reduced to 95% and RNA to over 90%.

ConclusionsThe process for the purification of pDNA presented in this study integrates the steps of alkaline lysis, solid-liquid separation and initial purification into a single unit operation. It allowed an over 90% recovery of pDNA, while a significant removal of RNA, proteins and even oc pDNA could be obtained. Due to the fact that ATPS is characterized by a high product capacity and an ease of scale, this process might give an answer to the open question towards economic pDNA processing.

Acknowledgements

We would like to thank QIAGEN for the cooperation and support of this work.

References1. Ferreira G.N.M., et al. Trends in Biotechnology. 18(9): 380–388

(2000).

2. Ribeiro, S.C. et al Biotechnology and Bioengineering 78: 376–384 (2002).

3. Kepka, C. et al. Journal of Chromatography A 1024: 95–104 (2004).

4. Frerix, A., et al. Biotechnology & Applied Biochemistry 42(1): 57–66 (2005).

5. Diogo, M.M., Queiroz, J.A., Prazeres, D.M.F. Journal of Chromatography A 998:109–117 (2003).

Fig 4. HIC analysis using a SOURCE™ 15PHE column (GE Healthcare) using HPLC.

Finally, as a preparation for further downstream polishing steps, the bottom phase would need to be desalted in order to remove the rather high concentrations of potassium phosphate. The latter can easily performed using a conventional Mr 100 000 cut-off ultrafiltration membrane.


Metal affinity systems for the integrated purification and refolding of recombinant D5-3-ketosteroid isomeraseM. H. Hutchinson, A. P. J. Middelberg, and H. A. Chase

Department of Chemical Engineering, University of Cambridge, Pembroke Street, Cambridge, CB2 3RA, UK.


The expression of recombinant protein in Escherichia coli often results in the formation of inclusion bodies. A generic and cost effective process for obtaining native protein from cells containing inclusion bodies is required to take advantage of the high expression levels obtained with E. coli. The refolding step is particularly troublesome for scale-up, and is commonly characterized by low yields and high cost.

An integrated process for purifying and refolding his-tagged inclusion body proteins is demonstrated in this work. This process uses a chemical extraction technique to release inclusion body proteins from the cells, and is then followed by an expanded bed adsorption (EBA) step to purify the protein. This series of operations replaces the need for mechanical disruption equipment to release inclusion body proteins from the bacterial cells. In addition, clarification after cell disruption is not necessary in this process because the EBA column will not become blocked by the presence of particulate contaminants. The adsorbed his-tagged protein is refolded in the column by changing to a non-denaturing running buffer. This immobilized refolding technique is intended to improve the refolding yield due to spatial separation of the refolding intermediates (1), and is shown in this work to be successful at low concentrations of immobilized protein. However, when a higher mass of protein is immobilized on the column the yield of the refolding step is reduced,

likely due to increased interactions between the adsorbed proteins during refolding. After refolding, the purified and active protein is eluted from the column.

This work therefore demonstrates a simple process for obtaining a high yield of pure, active protein product from inclusion body containing cells. Only a single immobilized metal affinity chromatography (IMAC) purification step is needed in order to obtain a high purity of the target protein. This process is especially useful for applications where the presence of a his-tag in the final product is acceptable.

Chemical extractionD5-3-ketosteroid isomerase (KSI) from Pseudomonas testosteroni is used as a model protein (2), which was cloned with the addition of a six residue C-terminus his-tag. After expression of the target protein in E. coli, the inclusion body proteins were solubilized and released from the cells using the following chemical extraction conditions:

• 8 M urea• 3 mM EDTA• 100 mM HEPES buffer• spermine tetrahydrochloride• pH 9.0• shaking at 200 rpm, 37 °C


The chemical extraction was complete after one hour, and was shown to be as effective at releasing the inclusion body protein as high-pressure homogenization (1500 bar, 3 passes). Spermine tetrahydrochloride precipitates the released nucleic acids, and was added in proportion to the total cell mass in the extraction (3). This spermine-nucleic acid complex was removed with low-speed centrifugation following the extraction. 6mM CaCl2 was then added so that the EDTA in the extraction buffer would not strip nickel metal ions from the IMAC column.

Expanded Bed AdsorptionAn expanded bed adsorption column utilizing IMAC interactions was used as the initial capture step. A 1 cm diameter glass column containing a settled height of 15 cm IDA STREAMLINE™ adsorbent was charged with nickel ions, and was subsequently equilibrated with running buffer (8 M urea, 50 mM Tris, pH 8.0). The chemically extracted protein solution was applied directly to the expanded bed, with the flow rate adjusted to retain an expansion ratio of 2. A stable column was observed during the sample loading, and efficient capture of the his-tagged KSI was demonstrated by the slow break-through of the target protein (Fig. 1). After application of 120 ml of feedstock, the column was rinsed with the running buffer until a stable baseline was reached. The flow was reversed to operate the column in packed bed mode, and 45 mM imidazole was used to wash

loosely bound protein from the column. The protein was then eluted in denatured form with 500 mM imidazole, 8 M urea, 50 mM Tris, pH 8.0. The collected eluent fractions were analyzed with an SDS-PAGE gel, which showed a very pure target protein product (Fig. 2).

Protein refoldingAs a basis for comparison to the immobilized refolding technique, refolding of the protein was first attempted by batch dilution refolding. After optimization of this process, it was found that the maximum refolding yield using this technique is 70%.

An immobilized protein refolding step was then developed in an attempt to improve upon the dilution refolding process. The immobilized refolding step was initially performed in a 1 ml packed bed column so that the affect of various operating conditions on the refolding yield could be tested. It was found that the recovery of active protein was highest when the refolding step was performed with a step change to the non-denaturing buffer. The yield was also highest when the elution was performed with a shallow gradient to the buffer containing imidazole. The optimal buffer conditions for refolding were also investigated, and it was found that the highest refolding yield was achieved using 50 mM potassium phosphate buffer at pH 6.8 and 4° C.

Fig 1. Breakthrough curves for the IMAC-EBA purification of KSI protein using chemical extraction mixture as feedstock. Fractions of the flowthrough were collected for SDS-PAGE analysis (Fig. 2) as shown.

Fig 2. SDS-PAGE of fractions from the IMAC-EBA experiment reported in Fig. 1. Lanes: 1=feedstock (diluted 11X); 2–8=flowthrough fractions (diluted 11X); 9=wash (diluted 3.5X); 10=elution (diluted 33X).


A maximum refolding yield of 80% was achieved using immobilized refolding, which is an improvement over the highest yields obtained using dilution refolding (Fig. 3). However, the loaded mass on the column has a large affect on the refolding yield. We hypothesize that this is due to interaction between adsorbed proteins during the refolding step, which occurs in spite of the surface immobilization. When there is a low amount of adsorbed protein the refolding intermediates are successfully separated from each other as a result of being anchored to the surface of the adsorbent through the his-tag. At higher adsorbed protein concentrations the protein molecules are packed more closely on the adsorbent surface, and the refolding intermediates thus can interact and aggregate even while adsorbed to the matrix.

a large scale due to the high adsorbent and buffer volumes that would be required. Compared with the 70% refolding yields that were obtained with the dilution refolding method, the simplicity of the immobilized refolding method does not justify such a loss in yield.

ConclusionsAn integrated process for the purification and refolding of a recombinant his-tagged protein has been demonstrated. Chemical extraction utilizing urea and EDTA has been proven to be as effective as homogenization for the recovery of inclusion body proteins. Expanded bed adsorption was then used to purify the his-tagged protein from the chemical extract, yielding a product with >95% purity using a single purification step. This generic and robust process can be used to recover and purify other his-tagged proteins expressed as inclusion bodies.

In an attempt to further streamline the process, an immobilized refolding step was added after the purification step. A maximum refolding yield of 80% was attainable by refolding the protein while adsorbed to an IMAC surface, which is an improvement over the maximum refolding yield of 70% that was obtained using dilution refolding. However, the yield of the immobilized refolding step is significantly reduced when there is an increased concentration of adsorbed proteins on the surface. This is likely due to interaction of the refolding intermediates made possible by their close proximity on the adsorbent surface.

References1. Creighton, T. E., Folding of proteins adsorbed reversibly to ion

exchange resins. Protein Structure, Folding, and Design, 249–257 (1986).

2. Choi, K. Y. and Benisek, W. F., Nucleotide sequence of the gene for the delta 5-3-ketosteroid isomerase of Pseudomonas testosteroni. Gene, 69 (1), 121–9 (1988).

3. Choe, W. S. and Middelberg, A. P. J., Selective precipitation of DNA by spermine during the chemical extraction of insoluble cytoplasmic protein. Biotechnology Progress, 17 (6), 1107–1113 (2001).

Integrated processThe immobilized refolding step was then integrated with the extraction and IMAC-EBA process described above, with all steps being carried out in the expanded bed. The yield of active protein from the integrated purification and refolding process was 11%. Although a significantly higher refolding yield would have been possible with lower column loading, under-loading the column is undesirable for work on

Fig 3. The immobilized refolding yield as a function of the mass adsorbed in a 1 ml packed bed IMAC column. The two series show the initial and optimized buffer conditions used for refolding.


Development of a multivariate statistical model for accurate scale-up predictions of pressure/flow properties Joakim Svensson1, Mattias Landin2, and Gunnar Malmquist2

1GE Healthcare, 800 Centennial Avenue, Piscataway, NJ, 08855, USA2GE Healthcare, Bjorkgatan 30, SE-175 84, Uppsala, SWEDEN


One of the basic scale-up issues in packed bed chromatography is that a wide column has less wall support than a small one, resulting in lower obtainable velocities at a given pressure drop in a larger column. Figure 1 shows the reduction in flow when scaling up from a 3.5-cm to a 30-cm column.

If large-scale pressure/flow cannot be measured, one solution is to model it . By modeling, it is possible to calculate what flow rates to expect in the manufacturing column when the process is already being designed at lab scale. Another reason for manufacturers of chromatography media to use models is that when the media are developed, there is a need to know early on if flow targets in large-scale columns can be reached.

A few models exist, among them one by Stickel et al. [1] and one by Keener et al. [2]. The model by Stickel is entirely empirical and does a fair job of predicting large-scale flow, provided the “small” columns are 10 to 20 cm in diameter. None of these models suited the goals in the development of Capto™ Q. The wish was to reduce the number of experiments required, keeping the number of measurements to a minimum and have a simple evaluation procedure. Another requirement was to improve the accuracy compared

to other models, if possible. If these criteria were met, the model could be used during Quality Control of future manufacturing lots.

One of the main issues during resin development is that there is an intentionally large variation between prototypes at early stages of a project. This is because the entire property space has to be scanned to investigate how to set specification limits. At first, there is a large variation in pore size and particle size, which in turn will lead to variation in pressure/flow (P/F) properties. A given prototype with small particles and large pores, leading to lower rigidity, has lower flow in general.

The approach was to employ a multivariate evaluation using the data and information that could be present in the small scale pressure-flow curve to try to couple this to large-scale data. Looking at the different curve shapes in Figure 2, it turns out that some of the curve features could be explained by rigidity and particle size. As a simplification, a larger average particle size gives a higher level of flow and a steeper initial slope, as indicated by arrows. A rigid medium gives the bed more resistance to compression and can therefore be run up to higher pressure and flow. It also has higher critical pressure

Fig 1. Reduction in flow when scaling from a 3.5 cm column to a 30 cm column.

Fig 2. Rigidity and particle size account for some of the profiling of the pressure/flow curves.


than the softer one and therefore the curve is more linear. A more rigid medium will lose less flow on scale-up, while the particle size does not affect scale-up but mainly just the level of flow.

Two separate models have been constructed during the development of two new media. One model was made for MabSelect Xtra™, a protein A medium. The other resin is an anion exchanger called Capto™ Q. Data in a 3.5-cm column and a 30-cm column for about 10 different lots were collected, all having different particle and P/F properties. A 30-cm column is actually a fair representation of even larger columns since roughly 90% of the wall support has already vanished at this scale. The utilized basic data set deserves an explanation. The flow rates (in this case at every 0.1 bar) in the small column were defined as independent x variables. The dependent response variable y was defined as the packed bed operational velocity in the 30-cm column.

By applying a Partial Least Squares regression model (Wold et al. [3]), the relationship between small and large scale could be statistically calculated. The result is a weighted linear combination of flow rates in the 3.5-cm column, giving an estimated P/F performance in the 30-cm column. Figure 3 shows a graphical representation of the model where the green bars are the variable weights of the model describing how the independent variables are related to the dependent response variable. Based on the weights and the x variables, it is now possible to calculate the response variable value for each of the included batches and compare them to the measured values. This will enable some statistical analyses of the models. Figure 4 shows that the match is very good for the Capto™ Q resin, meaning the correlation between measured and calculated values is good. The RMSEE, standing for Root Mean Square Error of Estimation, is 36 cm/h. ± 2 RMSEE is roughly the equivalent of a 95% confidence interval, that is, the predictive uncertainty in this model is around ± 72 cm/h or ± 7% of the specification limit of 1000 cm/h. The results for the MabSelect Xtra model were even better.

The models were then used to predict velocity for batches not included in the original data set. Small-scale data was collected and used to calculate the 30-cm column velocity. The predictions were generally good, below 5% from the actual measured large-scale column value and well within the ± 2 RMSEE limit.

The conclusion is that the whole small scale P/F curve contains valuable information on large-scale column flow properties. Also, this methodology can provide valuable information on P/F properties for resin from future production for example as a quality control test. Estimates of large scale P/F properties (30 cm in this case) can be supplied with each lot. Fewer tests at small scale are required, meaning that once the model is constructed and validated, only one test is required.

References1. Stickel et. al., Biotechnol. Prog. 17, 744–755 (2001)

2. Keener et. al., Biotechnol. Prog. 18, 587–596 (2002)

3. Wold, S., C. Albano, et al. (1984). Multivariate Data Analysis in Chemistry. Chemometrics: Mathematics and Statistics in Chemistry. B. R. Kowalski, Reidel Publishing Co, Dordrecht, Holland, pp 17–95.

Fig 3. Graphical representation of the Partial Least Squares regression model. The green bars are the variable weights of the model and describe how the independent variables are related to the dependent response variable. The response variable value can be calculated for each of the included batches and compared to the measured values.

Fig 4. The correlation between measured and calculated values is very good for the Capto™ Q resin.


Primary recovery – assessment of monoclonal antibody containing harvests to predict clarification requirementsCarl Spicer*

Cambridge Antibody Technology, Milstein Building, Granta Park, Cambridge, CB1 6GH, UK.


IntroductionDue to restricted development times and high quantities of monoclonal antibody required for initial safety and efficacy studies, a standard purification system is often employed. This standard purification process is designed to require little or no development time, although often the most difficult part of this standard process is the clarification step. Hence the evaluation of culture harvest properties to predict the most appropriate method of clarification.

BackgroundClarification is the removal of particulates down to 0.2 µm prior to the next purification step.

Many methods are available for clarification in the pharmaceutical and biotechnology sectors; expanded bed adsorption chromatography (EBA), centrifugation, and filtration being the most commonly used.

This report concentrates on filtration studies conducted at Cambridge Antibody Technology (CAT).

FiltrationFiltration has two basic modes: normal flow filtration (NFF) and tangential flow filtration (TFF) both of which use a porous membrane to separate particulates from a soluble stream. NFF is the classical filtration mode where the fluid flow is perpendicular to the membrane. In TFF, the fluid flow is parallel to the membrane surface and recirculates; this retentate flow and the particles present scour the membrane

surface preventing blockage and increasing the life span of the membrane. Hollow fiber TFF is a variant of this where the retentate passes down the core of porous fibers.

The choice of method depends upon the product, the buffer, and how these interact with the contaminants in the feed stream. In addition, factors such as size of system, cleaning methods, reuse requirements, validation, cost, and optimization all influence the choice of system. Examples of how these can affect clarification can be seen below.

Example 1. Both Figures 1 and 2 show particle distributions in fermentation harvests from mammalian cell lines producing a monoclonal antibody at different times during fermentation. Figures 1 and 2 show three populations of cells: dead cells (5 to 10 µm), live cells (10 to 15 µm) and dividing cells (greater than 15 µm). The two systems described are similar, but require quite different clarification conditions.

Fig 1. NS0 Cell Diameter During Fermentation.

0 5 10 15 20 25 30 35 40

cell diameter

day 2day 8day 12day 22 4C storage


Example 2. Figure 4 shows that the passage of an antibody across a TFF membrane during a diafiltration step can be affected by the buffer species used during diafiltration.

Case studyThe aim was to develop a standard clarification process for a range of antibodies produced in NS0 cell lines in a culture media containing bovine serum albumin (BSA) through evaluation of four systems:

NFFThe standard process, used at 100-l scale, was NFF with 4.5 m2 of 1.2-µm, 1.5 m2 of 0.45-µm, and 0.7 m2 of 0.2-µm membrane. The run time was 4 h. NFF was simple to operate, but expensive (£1500 per batch), and at larger scales the filter footprint became prohibitory.

TFF flat sheetFlat bed TFF was investigated and a process developed for 100-l scale using 0.5 m2 of a 0.2-µm membrane. TFF was more labor-intensive and had a run time of 6 h. The cost was £800 per membrane, which could be reused for multiple batches. Additionally, the subsequent 0.2 µm sterilizing filtration requirements reduced costs and the footprint of the equipment used.

Fig 2. CHO Cell Diameter During Fermentation.

Fig 3. Fed Batch Fermentation.

Fig 4. Buffer Effects on Product Transfer Across Membrane.

However, during process evaluation, antibody titer peaked around day 11, so continuation of the fermentation to day 14 is a cost with no gain. Reduction of costs by moving harvest date from day 14 to 11 significantly increased cell viability from 10% to 50% (Fig 3). Unexpectedly, this caused fouling of the membrane and a reduction in retentate flow. This issue was consistent with all membranes tested, leading to the conclusion that flat sheet TFF was not suitable for harvests with high cell viability. This contradicts the general belief that mammalian cell harvests are easier to clarify as they increase in viability.

Charged depth filtersEvaluation of diatomatious earth filters – charged depth filters – was successful for different cultures as they reduced costs (£400 per batch) and area. However, high and low percentage cell viability harvests required different filter trains and significant 0.2 µm final filtration. This was not an elegant single process system and would result in large volumes of held stock.

Hollow fiber TFF Hollow fiber TFF was not affected by a range of cell viabilities. A 100-l scale process was developed using 1.8 m2 of a 0.45-µm membrane; although this took around 6 h and was labor-intensive. The membrane costs were £1600 per set, but they could be reused and there was little requirement for subsequent 0.2 µm sterilizing membrane filtration. This reduced the membrane cost and the footprint of the equipment used. Recovery with this clarification step was typically 98% and lead to the adoption of hollow fiber TFF as the standard clarification method for harvests of differing cell viability.

Further process improvements resulted in a culture media that was animal component free (ACF). The

day 2

day 8

day 12

day 22 4C storage

0 5 10 15 20 25 30 35 40

cell diameter

0 2 4 6 8 10 12 14 160

102030405060708090

100

Days

Titr

e

Cells Titre

01020304050607080

Buffer pH


recovery of antibody during clarification of this ACF media was reduced to 85%, although this was still considered acceptable.

Evaluation of clarification of CHO cell line harvests (which have a different particle profile to NS0) initially produced low antibody recovery. However, relatively simple process optimisation of the hollow fiber process for the CHO cell lines and ACF media increased the recoveries back to 98%.

Case study conclusionsIn this study, hollow fiber TFF was the most robust method of harvest clarification at 100-l pilot scale. Hollow fiber TFF is inexpensive to run, has a small footprint, and is also scaleable to much larger volumes if required.

NFF can stiil be resorted to if a clarification proves problematic or must work first time with high recovery.

The futureThis work generated data (e.g. cell line, cell viability, particle size distribution, TMP, retentate flow rates, recoveries) that can be evaluated in an attempt to discover relationships between the different factors. These relationships can then be tested using multifactorial experimental design on small-scale trials.

University College London have taken this further with their ultra scale down initiative and developed a 47-mm membrane disc TFF rig. In a collaboration with CAT, many trials on a very small harvest volume have been performed. More importantly, the software developed will identify the operational window for scale-up and optimization (Fig 5).

Conclusions• Find out what technology is available: it is always

changing and progressing.• Learn as much as possible about the start

material: not just the product (which is usually well characterized), but also the buffer, the contaminants, and the interaction between them.

• Try as many methods as practical: even when you find one that works keep looking – you may find a better one.

• Do not always expect the process to work as it should: the combination of product and method is unique.

• Collect the data and use it: it is not just an historical document but information that can be made to work for you.

Fig 5. Window of Operation.

0

25

50

75

0 1000 3000 4000 5000 60002000

Wall shear rate (1/s)

Flux too low (time or membrane area constraint)

Shear rate too high (cells or equipment constraint)

Jcrit exceeded (fouling constraint)

Scale-up design area


Hydrodynamic characterization, comparison, and scaling-up of a cleaning protocol for severe fouling of an expanded bed column following the application of Escherichia coli homogenateSorina Morar*, Phil Ropp, Jeff Schrimsher, and Christie Williams

Diosynth Biotechnology, 3000 Weston Parkway, Cary, NC 27513, USA.


In an expanded bed adsorption (EBA) chromatography step, biomass/adsorbent interactions reached unacceptable levels and led to column fouling and decreased performance of the STREAMLINE™ SP adsorbent. We found that the conventional cleaning protocol was not sufficient to completely restore the hydrodynamic properties of the resin after the homogenate application. To ensure successful reusability of STREAMLINE SP, a unique cleaning protocol was optimized for the particular Escherichia coli feedstock. We applied the bed expansion and residence time distribution (RTD) analyses to quantitate and compare the efficiency of various cleaning protocols. Only 6 M guanidine hydrochloride (GdnHCl) proved to ensure a complete cleaning of STREAMLINE SP. The cleaning protocol was successfully scaled up 10-, 100-, and 2000-fold.

Extended ReportThe benefits achieved with the use of EBA are sometimes offset by the challenges of developing optimized conditions for cleaning the resin and column hardware. Lysates from E. coli are typically turbid and viscous as they contain large amounts of cell debris, proteins, lipids, fatty acids, nucleic acids, and amino acids. The solid components can interact non-specifically with the EBA adsorbent, leading to bead agglomeration and column fouling. These adverse effects can impose complex and harsh cleaning-in-place (CIP) solutions to restore the performance of the EBA adsorbent and allow its reusability.

In our study, E. coli homogenate was applied to a STREAMLINE SP expanded bed cation exchanger. The

small-scale system was a 1.6-cm column. The bed height was 30 cm. The conventional CIP procedure consisted of three steps, all performed in the upflow direction: 0.5 M NaOH followed by a water rinse, then 30% isopropanol, and 25% acetic acid. After a single run and conventional cleaning severe flow channels, stagnant dead liquid zones, aggregates of adsorbent particles, breakdown of the bed, agglomeration of absorbent particles at the inlet, modification of flotation properties of the matrix at the top of the expanded matrix, and bed turbulence were observed. In addition, at the end of each run, the settled bed height was found to increase by at least 2 cm.

Besides visual examination of the column during performance, we used bed expansion and residence time distribution (RTD) analyses to evaluate reusability of STREAMLIE SP. The dimensionless number q@maxE with a value higher than 0.8 is here proven to be an appropriate acceptance criterion for cleaning efficiency. Thus, after one chromatography run with the crude lysate and a cleaning cycle following the manufacturer's recommended protocol, the decreased performance of STREAMLINE SP was characterized by poor bed expansion, distorted peak shapes and tailing, and a q@maxE value below 0.8 (see Fig 1 and Table 1).

These results triggered the search for a load-specific cleaning protocol. A variety of other cleaning solutions were tested, such as 1 M NaOH/2 M NaCl, hot water, 6 and 8 M urea, or 6 M guanidine hydrochloride (Gdn). Visual observations and hydrodynamic evaluations rendered the Gdn solution as the most effective in restoring performance of STREAMLINE SP following full 30 column volume loads.


q

0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00

(Eq )

0.0

0.5

1.0

1.5

2.0

2.5

3.0 preuseConvClean

q

0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00

(Eq )

0.0

0.5

1.0

1.5

2.0

2.5

3.0SP50_preuse postRUN1postRUN2

Table 1. Hydrodynamic properties of the Streamline SP medium before and after a conventional cleaning were compared using RTD analysis at small scale. After a full purification run, the used medium was cleaned with 0.5 M NaOH/1M NaCl, 30% isopropyl alcohol, and 25% acetic acid. As described by the Richardson-Zaki equation, n is the expansion bed index and ut is the terminal velocity. According to the dispersion model, A is the peak asymmetry, HETPEB = HETP · x, HETP is the height equivalent of a theoretical plate, x is the expansion factor, u is the linear velocity required to maintain a 3× expansion, and D is the dispersion coefficient.

Measured Parameter

Pre-Use Value Post-Cleaning Value

n 4.80 –

ut 798 –

A 1.95 4.56

HETPEB (cm) 2.50 20.8

Theoretical Plates (N/m)

120 14

u (cm/hr) 284 418

q@maxE 0.93 0.62

D (m2/s) 9 × 10-6 120 × 10-6

Note: The simplified equation used to determine HETP and the use of dispersion model to extract D assume ideal fluid distribution/symmetrical RTD. When q@maxE is less than 0.8, these assumptions are violated. Thus the HETP and D values for post-cleaning are only reported to show that the boundary conditions are violated.

Fig 2. Observed RTD generated on a 5-cm column using GdnHCl as cleaning solution between runs: (●) fresh STREAMLINE SP, (▲) STREAMLINESP used and cleaned once, (■) STREAMLINE SP used and cleaned twice.

Table 2. Two chromatography runs were performed on a 5-cm column with 6 M guanidine hydrochloride, cleaning after each run. The runs were designated as Run 1 and Run 2. The hydrodynamic properties of the Streamline SP before and after cleaning were compared using RTD analysis. See Table 2 for notations.

Measured Parameter Pre-Use Post-Run1 Post-Run2

n 4.61 4.87 5.06

ut 830 858 913

A 2.12 2.02 2.04

HETPEB (cm) 8.19 12.66 8.49

Theoretical Plates (N/m) 36 24 35

u (cm/h) 311 316 306

q@maxE 0.87 0.86 0.87

D (m2/s) 30 × 10-6 41 × 10-6 32 × 10-6

Fig 1. RTD curves showing the loss of STREAMLINE SP performance on a 1.6-cm column following a purification run with a 30 column volumes lysate load and the conventional cleaning protocol (0.5 M NaOH/1M NaCl, 30% isopropylalcohol, and 25% acetic acid): (▲) fresh STEAMLINE SP, (●) used and cleaned adsorbent.

To further prove the efficiency of the Gdn-based cleaning protocol, the EBA step was first scaled up 10-fold to a 5-cm diameter column (STREAMLINE 50). As shown in Fig. 2 and Table 2, Gdn reversed the biomass-EBA adsorbent interactions and q@maxE remained above 0.8 following the second cleaning cycle. We further scaled up to 20-cm and 60-cm Streamline columns. In both cases, two consecutive runs were performed successfully, with the hydrodynamic properties being completely restored after each Gdn cleaning.

In conclusion, a series of cleaning solutions were screened at small–scale for their efficiency in disrupting specific biomass-EBA adsorbent interactions. The dimensionless RTD number q@maxE value used as an acceptance criteria for cleaning showed that only 6 M GdnHCl solution was successful in restoring the hydrodynamic properties of the STREAMLINE SP adsorbent after a full load. The cleaning protocol was demonstrated to perform consistently between the small- and the large-scale operations. Thus, we showed that reusability of EBA adsorbents depends on the implementation of an efficient regeneration protocol that is specific for the feedstock. This is especially important as it increases the cost-effectiveness of manufacturing processes for protein drugs.


Establishing large-scale expanded bed adsorption technology in the first step purification of industrial enzymesWerner Hölke and Stephan Glaser

Roche Diagnostics GmbH, Roche Applied Science, Global Biotechnology Operations, Nonnenwald 2, 82372 Penzberg, Germany


IntroductionRoche was looking to develop a new procedure for purifying extracellular Pichia pastoris proteins. The new purification procedure should be more economical and with fewer steps.

Following successful experiments at small-scale, the decision was made to employ expanded bed adsorption (EBA) technology, replacing the traditional approach of time-consuming precipitation and concentration steps, space-consuming equipment, expensive solid/liquid phase separation, and standard column chromatography.

The Pichia expression system grows fast to a high cell density in a defined synthetic medium. The system is eucaryotic, producing glycosylated extra-cellular proteins. Complex enzyme proteins are correctly folded.

Pichia cells are small, 3–5 µm in diameter, and robust, which is advantageous for harvesting by separation or filtration.

CriteriaIn developing the new procedure the challenge was to define the optimal conditions at small-scale and then transfer and establish these at the final scale in EBA columns up to 1000 mm in diameter. For EBA to be used for the commercial purification of enzymes from Pichia fermentation the following criteria were to be met:• Scale: the EBA process must be able to process a

fermentation broth volume between 1 and 10 m3 directly from the fermentor.

• Process: it must be a reliable and robust processing operation

• Product stability: the product must be stable, defined as no proteolytic degradation at a defined pH over 4 to 5 days.

• Economy: the process must be economical, faster, and consume less space than the existing one.

• Quality: it must meet all regulatory demands.


noitatnemreFhtorb- etartnecnoC

1-lCaN

etartnecnoC dicA-cirtiN

-lorecylGetartnecnocetartnecnoC

2

-muidoS-edixordyH

noituloS

noitarbiliuqE-erP AnoitarbiliuqE B

noitprosdA CgnihsaW-erP D

gnihsaW-lorecylG E

gnihsaW-tsoP FnoitulE G

noitulE-tsoP HnoitarenegeR I

noitarenegeR-tsoP J

-gnihsaWnoituloS

-tcudorPnoituloS

-tcudorPnoituloS

tesaW

dezinoi-eDretaw

-ssecorP-oiBtnalP

nmuloc mm 0001/006

-600 -400 -200 0

Wasching

ml

0 500 1000 1500

Waste

Glycerinwaschung Elution

65432

Aequilibrieren

l

10.0 15.0 20.0

16.68

11.85

10.0 15.0 20.0

8.23 10.96 14.39 16.58 19.04

12.69

50 mm column

200 mm column

Elution profile Purity

Fig 1. Scale-up from 50 mm to 200 mm column.

Fig 2. Construction of our Automated Purification System.


E 0304

E 0305

E 0306

l35000300002500020000150001000050000

0.0

1.0

2.0

3.0

4.0

5.0

6.0AU

-600 -400 -200 0

Wasching

ml

0 500 1000 1500

Waste

Glycerinwaschung Elution

65432

Aequilibrieren

l

Scaling-upScale up was performed in three steps: The elution profile obtained from the 200 mm EBA column was identical with that from the 50 mm EBA column (Fig 1) the decision was then taken to scale up to a larger column 600 and then 1000 mm columns. At the same time the new, purification plant was under construction that would eventually house the new automated process. (Figs 2 & 3).

At the final scale, the 1000 mm column, there was good reproducibility, bound cells could be easily removed with 5–10% glycerol, and elution performed isocratically in expanded bed mode.

One issueBut there remained one main issue – cell deposits at the distribution and endplates. This was resolved by modifying the large EBA column and including a special construction between the distributor plate and the coarse net. A 50 µm stainless steel pre-filter was introduced to remove cell agglomerates, as were wash steps at the column inlet (Table 1). The column was also operated without a support net at the outlet and without a mobile adaptor. (Fig 4).

Fig 3. Results from final process scale, 1000 mm, EBA column.

Table 1. Measures to avoid cell deposits between the endplate and distributor plate.

During the process1) Pump diluted feedstock in upward flow for 3 hours

2) Stop

3) At the same flow velocity, pump diluted feedstock in downward flow for 3 min

4) STOP

5) Pump diluted feedstock in upward flow for 3 hours,

6) Repeat steps as necessary.

RegenerationEndplate with 2 inlets and 2 outlets

Regeneration completed with sodium hydroxide

Pump de-ionized water into the endplate

Reverse the stream every minute: inlets outlets; outlets inlets, etc

Rinse for 6 min at maximum speed

Elution Profile

Final dimension: Good reproducibility 1000 mm column

Removal of bound cells: 5–10% glycerol is sufficient

Elution: Isocratic elution in expanded bed mode3–4 exchanger volume

50 mm Column

Identical Elution Profile

200 mm Column


Cell agglomeration – preferred at low pH – result in expanded bedmode without distribution plate and mobile adapter

detnemideS 1deb

dezidiulF 2deb

noitule lamroN 4)deb dekcap(

.lppa elpmaS 3 gnihsaw dna

)deb dednapxe(

,.lppa elpmaS :laicepS 5noitule ,gnihsaw

)deb dednapxe(

Meeting the challengesThe EBA technology met our criteria as described earlier.• Scale: It was qualified to substitute the classical

steps in our process• Reliability: Three processes are in routine

production. Scalability was never an issue. Problems with cell deposition and antifoaming were resolved.

• Product stability: Enzyme stability was met according to our given criteria

• Economy: Space was reduced and the process was completely automated – which improved overall process economy.

• Quality: The new process using EBA technology are validated and fulfill all regulatory demands.

In conclusionThe well-known problem of cell agglomeration resulted in the use of two different types of EBA columns: with and without a mobile adapter/support net at the outlet. In addition, the use of different technical devices avoided cell deposit. Transfer of the process from small- to large-scale column presented no difficulties. By employing EBA we achieved our objective and developed highly automated and stable processes. The new plant fulfils all regulatory requirements. Two additional processes using EBA are in development.

Fig 4. EBA Column: No Mobile Adapter.



List of posters presented at Cap 2005, Phoenix, USA* denotes author for correspondence

1. A New Medium for Scale-Up Purification of Histidine Tagged ProteinsL. C. Andersson, A. Bergh, A. Heijbel*, H. Lindgren, J. Lundqvist, K. Torstensson, and K. ÖbergGE Healthcare, Björkgatan 30, SE-751 84 Uppsala, Sweden.

2. Scale-Up of Membrane Chromatography for Process-Scale Capture ApplicationsPeter R. Levison, Pall Life Sciences, Walton Road, Portsmouth, Hampshire, PO6 1TD, UK.

3. Precipitation – A Tool for Novel Product FormulationM. Golubovic*1, M. Ottens1, G. J. Witkamp2, and L. A. M. van der Wielen1

1Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628 BC Delft, The Netherlands. 2Delft University of Technology, Laboratory for Process Equipment, Leeghwaterstraat 44, 2628 CA Delft, The Netherlands.

4. Recovery of Malaria Transmission Blocking Vaccine Antigen from Pichia pastoris by Adsorption on Newly Design Expanded Bed (STREAMLINE Direct)Loc Trinh, Je Nie Phue, and Joseph Shiloach*Biotechnology Unit NIDDK NIH Bethesda MD 20892, USA.

5. Controlling Autodegradation and Yield During AIEX Capture of rhFVIIaDaniel E. Rasmussen and Janus C. Krarup*Department of Protein Separation, CMC Development, Novo Nordisk A/S, Hagedornsvej 1, DK-2820 Gentofte, Denmark.

6. Capture of GFP on the New Anion Exchanger Capto QKjell Eriksson*, Anna Åkerblom, and Mattias BryntessonGE Healthcare, Björkgatan 30, 751 84 Uppsala, Sweden.

7. Characterization of a New Separation Medium Aimed for Hydrophobic Interaction Chromatography – Butyl-S Sepharose 6 Fast FlowJan Gustavsson*, Makonnen Belew, John Clachan, Anders Gustafsson, Bo-Lennart Johansson, Emil Larsson, Malin Sparrman, and Marianne Sparrman, GE Healthcare, Björkgatan 30, SE-751 84, Uppsala, Sweden.

8. Optimization of Cutinase Purification by Expanded Bed Adsorption (EBA) and On-Line Monitoring of Target Enzyme Activity Almost in Real TimeC. F. Almeida, C. R. C. Calado, J. M. S. Cabral, and L. P. Fonseca*Instituto Superior Técnico, Centro de Engenharia Biológica e Química Av. Rovisco Pais, 1049–001 Lisboa, Portugal.

9. An Experimental Determination of the Particle Density Distributions of STREAMLINE Expanded Bed AdsorbentsSally Hassan*1, Nigel Titchener-Hooker1,2, and Nik Willoughby2

1Department of Biochemical Engineering, University College London, Torrington Place, London WC1E 7JE, UK. 2EPSRC Life Sciences Innovative Manufacturing Research Centre for Bioprocessing, The Advanced Centre for Biochemical Engineering, University College London, Torrington Place, London WC1E 7JE, UK.

10. Development of Windows of Operation for Expanded Bed AdsorptionN. Willoughby*, S. Ngiam, H. Baldascini, P. Gardner, and N. Titchener-HookerEPSRC Life Sciences Innovative Manufacturing Research Centre for Bioprocessing, The Advanced Centre for Biochemical Engineering, University College London, Torrington Place, London WC1E 7JE, UK.

11. Experimental and Modeling Study of Protein Adsorption in Expanded Bed Adsorption ProcessPing Li, Guohua Xiu, and Alirio E. Rodrigues*Laboratory of Separation and Reaction Engineering, Department of Chemical Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, s/n 4200–465 Porto, Portugal.

12. Development of Innovative Process Technologies Addressing the Need for a Reliable and Economic Plasmid DNA Manufacturing ProcessLothar Breitkopf*, Peter Moritz, Nicola Scholle, Sandy Leifholz, Eveline Samol, Jörg Hucklenbroich, Manuela James, Astrid Breul, Wayne Tvrdik, Joachim Schorr, and Markus MüllerQIAGEN GmbH, Contract Manufacturing & Process Development, QIAGENstr. 1, 40724 Hilden, Germany.


13. Development and Scale-Up of a New High-Capacity Protein A ResinJoy Adiletta*1, Yan Brodsky1, Ganesh Vedantham1, Joanne Reeder1, Robert Hershberg1, Clayton Brooks1, Anders Ljunglöf2, Joakim Svensson2, Hans J. Johansson2, and Gunnar Malmquist2

1Purification Process Development, Amgen Inc., Seattle, WA 98119, USA. 2GE Healthcare, Björkgatan 30, SE-751 84 Uppsala, Sweden.

14. Tangential Flow Filtration of Influenza Virus as a Model SystemB. Kalbfuß1, M. W. Wolff, A. Zimmermann1, C. Best1, R. Wickramasinghe4, R. Morenweiser3, U. Reichl1,2, and N. Ross*3

1Bioprozesstechnik, Max-Planck-Institut Magdeburg, Germany. 2Lehrstuhl für Bioverfahrenstechnik, Otto-von-Guericke Universität, Germany. 3GE Healthcare, Fast Trak Services Europe, Müntzinger Strasse 9, Freiburg, Germany. 4Department of Chemical Engineering, Colorado State University, USA.

15. Optimization of an E. coli Homogenate Clarification Using ÄKTAcrossflowJozsef Vasi1 and Alisa D. Liten*1, and Anna Andersson2

1GE Healthcare, Björkgatan 30, SE-75184 Uppsala, Sweden. 2Department of Medicinal Chemistry, Uppsala University, Uppsala, Sweden.

16. New Applications of High-Gradient Magnetic Fishing in BioprocessingT. J. Hobley*1, H. Ferré1,2, C. S. G. Gomes1, D. B. Hansen1, T. L. Petersen1, S. Buus2, and O. R. T. Thomas1,3

1Center for Microbial Biotechnology, BioCentrum-DTU, Technical University of Denmark, Building 223, Søltofts Plads, DK-2800, Kgs. Lyngby, Denmark. 2Institute of Medical Microbiology and Immunology, The Panum Institute, University of Copenhagen, Blegdamsvej 3, DK-2200, Copenhagen, Denmark. 3Now at: Department of Chemical Engineering, School of Engineering, The University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK.

17. Further Advances in Bioprocessing with Magnetic AdsorbentsT. J. Hobley*1, M. Franzreb2, M. Siemann-Herzberg3, and O. R. T. Thomas4

1Center for Microbial Biotechnology, Technical University of Denmark, Denmark. 2Institute for Technical Chemistry, Forschungszentrum Karlsruhe, Germany. 3Institute of Biochemical Engineering, University of Stuttgart, Germany. 4Department of Chemical Engineering, University of Birmingham, Edgbaston, UK.

18. Performance Comparison of Solid-Phase Refolding in PBA and EBA ColumnWon Chan Choi1, Min Young Kim, Chang Woo Suh2, and E. K. Lee*Bioprocessing Research Laboratory, Department of Chemical Engineering, Hanyang University, Ansan 425–791, Korea. 1Present address: Plasma Derivatives Team, Korean Green Cross Corp., Shingal, Korea. 2Present address: Biotechnology Research Center, DaeWoong Corp., YongIn, Korea.

19. Lid Approaches Towards Cell Adhesion Free Expanded Bed AdsorptionPhilippe Busson, Martin Hall, Anders Ljunglöf, Jozsef Vasi, and Jean-Luc Maloisel*GE Healthcare, Björkgatan 30, SE-751 84 Uppsala, Sweden.

20. Development of a Capture Step for Recombinant Dalbergia cochinchinensis b-GlucosidaseKristina Nilsson-Välimaa*, Christine Sund-Lundström, Jozsef Vasi, and Robert MagnussonGE Healthcare, Björkgatan 30, SE-751 84 Uppsala.

21. Linkage of Hydrodynamic and Adsorption Models of Expanded Bed OperationSabin Maskey* and Nigel J. Titchener-HookerAdvanced Centre for Biochemical Engineering University College London, London WC1E 7JE, UK.

22. Recovery of a Biologic Bound to the Cell Surface of a Mammalian Cell Culture Suspension Using Depth Filtration Followed by a Buffer FlushRichard Weber*, Nels Pederson, John Pierraci, Yao-Ming Huang, Merna Sada, and Chistine PhamBiogen Idec, One Antibody Way, Oceanside CA 92056, USA.

23. EBA Column EvaluationKine A-K. Barnfield Frej*, Magnus Asplund, Åsa Lagerlöf, Mikael Lundberg, and Magnus RyggeGE Healtcare, Björkgatan 30, SE-751 84 Uppsala, Sweden.

24. The Impact of New High Capacity Protein A Matrices for an Economical mAb Purification ProcessFranz Nothelfer*, Jasmin Nehmer, and Dorothee AmbrosiusBoehringer Ingelheim Pharma GmbH & Co. KG Biopharmaceuticals, Process Science, Germany.

25. Use of Vibrating Membrane Microfiltration Technology for Recovery of Secreted ProteinsJ. M. Liddell* and A. D. Harland, Process Science Group, Avecia Biotechnology, Billingham TS23 1YN, UK.


Author index

Baines, D. 12 Betley, J. 12 Borginis, D. 6 Burton, S. 12

Chase, H. A. 17

Frerix, A. 14

Glaser, S. 27

Hermans, P. 9 Hubbuch, J. 14 Hutchinson, M. H. 17 Hölke, W. 27

Jen, D. 6

Kula, M-R. 14

Landin, M. 20

Malmquist, G. 20 Miao, F. 6 Middelberg, A. P. J. 17 Morar, S. 25 Müller, M. 14

Ramelmeier, A. 6 Ropp, P. 25

Schrimsher, J. 25 Spicer, C. 22 Svensson, J. 20

ten Haaft, M. 9

Williams, C. 25

imagination at work28-4094-51

www.gehealthcare.com

GE Healthcare Bio-Sciences AB

Björkgatan 30

751 84 Uppsala

Sweden

Asia Pacific Tel: +85 65 6 275 1830 Fax: +852 2811 5251 • Australasia Tel: +61 2 9899 0999 Fax: +61 2 9899 7511 • Austria Tel: 01/57606-1619 Fax: 01/57606-1627 • Belgium Tel: 0800 73 888 Fax: 03 272 1637 • Canada Tel: 800 463 5800 Fax: 800 567 1008 • Central, East, & South East Europe Tel: +43 1 982 3826 Fax: +43 1 985 8327 • Denmark Tel: 45 16 2400 Fax: 45 16 2424 • Finland & Baltics Tel: +358 (0)9 512 39 40 Fax: +358 (0)9 512 39 439 • France Tel: 01 69 35 67 00 Fax: 01 69 41 96 77 • Germany Tel: 0761/4903-490 Fax: 0761/4903-405 • Italy Tel: 02 27322 1 Fax: 02 27302 212 • Japan Tel: +81 3 5331 9336 Fax: +81 3 5331 9370 • Latin America Tel: +55 11 3933 7300 Fax: +55 11 3933 7304 • Middle East & Africa Tel: +30 210 9600 687 Fax: +30 210 9600 693 • Netherlands Tel: 0165 580 410 Fax: 0165 580 401 • Norway Tel: 815 65 555 Fax: 815 65 666 • Portugal Tel: 21 417 7035 Fax: 21 417 3184 • Russia & other C.I.S. & N.I.S Tel: +7 (095) 232 0250 Fax: +7 (095) 230 6377 • South East Asia Tel: 60 3 8024 2080 Fax: 60 3 8024 2090 • Spain Tel: 93 594 49 50 Fax: 93 594 49 55 • Sweden Tel: 018 612 1900 Fax: 018 612 1910 • Switzerland Tel: 0848 8028 12 Fax: 0848 8028 13 • UK Tel: 0800 616928 Fax: 0800 616927 • USA Tel: 800 526 3593 Fax: 877 295 8102

Capto, MabSelect Xtra, Sepharose, and STREAMLINE are trademarks of GE Healthcare companies.

GE, imagination at work, and GE monogram are trademarks of General Electric Company.

CaptureSelect is trademark of BAC B.V.

Durapore is a trademark of Millipore Corp.

E-PAGE is a trademark of Invitrogen.

Mimetic and Purabead are trademarks of Prometic.

ZetaPlus is a trademark of Cuno Inc. a 3M company.

GE Healthcare Bio-Sciences AB, a General Electric Company

GE Healthcare UK Ltd Amersham Place, Little Chalfont, Buckinghamshire, HP7 9NA, UK

GE Healthcare Bio-Sciences Corp 800 Centennial Avenue, Piscataway, New Jersey 08855, USA

GE Healthcare Europe GmbH Munzinger Strasse 5, D-79111 Freiburg, Germany

GE Healthcare Bio-Sciences K.K. Sanken Building 3-25-1, Shinjuku-ku, Tokyo 169-0073, Japan

All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. GE Healthcare reserves the right, subject to any regulatory and contractual approval, if required, to make changes in specifications and features shown herein, or discontinue the product described at any time with notice or obligation. Contact your local GE Healthcare representative for the most current information © General Electric 2006 – All rights reserved.

The views expressed by the contributors and correspondents are their own and do not necessarily reflect the views of GE Healthcare Bio-Sciences.

Documents

Downstream Cap 05 abstracts - … Downstream – Cap ’05 abstracts Clariﬁcation of biomass harvest from a perfusion bioreactor Fudu Miao, Dan Borginis, David Jen, and Andrew Ramelmeier