233
The Pennsylvania State University The Graduate School College of Engineering PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING MEMBRANE PROCESSES A Dissertation in Chemical Engineering by Krisada Ruanjaikaen 2013 Krisada Ruanjaikaen Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2013

PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

  • Upload
    others

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

Page 1: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

The Pennsylvania State University

The Graduate School

College of Engineering

PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS

USING MEMBRANE PROCESSES

A Dissertation in

Chemical Engineering

by

Krisada Ruanjaikaen

2013 Krisada Ruanjaikaen

Submitted in Partial Fulfillment

of the Requirements

for the Degree of

Doctor of Philosophy

August 2013

Page 2: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

The dissertation of Krisada Ruanjaikaen was reviewed and approved* by the following:

Andrew L. Zydney

Head of the Department of Chemical Engineering

Walter L. Robb Chair and Professor of Chemical Engineering

Dissertation Advisor

Chair of Committee

Darrell Velegol

Distinguished Professor of Chemical Engineering

Michael Janik

John J. and Jean M. Brennan Clean Energy Early Career Professor and

Associate Professor of Chemical Engineering

Arnold A. Fontaine

Senior Scientist of Applied Research Laboratory

Professor of Bioengineering

*Signatures are on file in the Graduate School

Page 3: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

iii

ABSTRACT

There is growing interest in the use of pegylated therapeutic proteins due to the

improved therapeutic efficacy, with much greater serum circulation half-life leading to lower

dosage requirements and dosing frequency. One of the challenges in producing pegylated

proteins is the low product yield and difficult purification of the desired (typically mono-

pegylated) product. The overall objective of this thesis was to examine the use of

ultrafiltration for the production and purification of a desired mono-pegylated protein

product. The specific aims included: (1) evaluate the effects of electrostatic and solute-solute

intermolecular interactions on transmission of pegylated proteins through both neutral and

charged ultrafiltration membranes, (2) develop an ultrafiltration process to purify pegylated

proteins with different degree of pegylation, and (3) develop a combined reaction and

membrane-based separation process to enhance the yield of a desired pegylated product.

Experiments were performed using a model pegylated system produced by covalent

attachment of activated PEG to α-lactalbumin. Ultrafiltration experiments were performed

with UltracelTM cellulosic membranes, with a negatively charged version generated by

covalent attachment of sulfonic acid groups to the base cellulose.

The ultrafiltration data showed strong electrostatic exclusion of the pegylated proteins

from a charged membrane. A theoretical model was developed to describe the sieving

behavior of the pegylated proteins accounting for the increase in the effective protein size,

the elimination of the protonatable –NH2 group due to the pegylation reaction, and the

alteration of the electrostatic potential field around the protein due to the PEG layer. The

transmission of the pegylated proteins also increases with increasing PEG concentration due

to the increase in free energy of the pegylated protein in the bulk solution associated with the

Page 4: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

iv

strong intermolecular interactions. These intermolecular interactions also affect the bulk mass

transfer, which could be described theoretically using a modified concentration polarization

model. These models provide an appropriate framework to describe / optimize the

performance of ultrafiltration and diafiltration processes for the purification and formulation

of pegylated proteins.

The sieving data were used to develop a diafiltration process to purify the mono-

pegylated protein. Unreacted (native) protein and PEG were removed via a single

diafiltration step using a highly charged membrane with relatively large pore size (300 kDa);

the high retention of the mono-pegylated protein was due to a combination of steric and

electrostatic exclusion. The process provided greater than 90% yield with purification factors

of more than 20. A similar diafiltration process was developed for purification of a mono-

pegylated protein from the di- and tri-pegylated forms, with greater than 95% yield and more

than 20-fold purification.

A combined reaction-membrane based separation system was developed to produce a

mono-pegylated protein at high yield by continuously separating the product from the

reactor. The final product yield from this reaction-separation scheme was 69%, significantly

higher than the 50% yield obtained using a batch process. A simple mathematical model was

developed for this reaction-separation system, providing additional insights into the

underlying phenomena and a framework for the design and optimization of this type of

reaction-separation process. Overall, this thesis provided a clear demonstration of the

potential of using membrane-based systems for the purification and enhanced production of

desired protein-polymer conjugates.

Page 5: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

v

TABLE OF CONTENTS

LIST OF FIGURES ................................................................................................................. ix

LIST OF TABLES…………………………………………………………………………. ...xvi

ACKNOWLEDGEMENTS ..................................................................................................... xvii

Chapter 1 Introduction ............................................................................................................ 1

1.1 Therapeutic Proteins .................................................................................................. 1

1.2 Pegylated Therapeutic Proteins ................................................................................. 3

1.2.1 Benefits of Pegylation ........................................................................................ 5

1.2.1.1 Prolonged Circulation Time ..................................................................... 6

1.2.1.2 Reduced Immunogenicity and Side Effects .................................................... 9

1.2.1.3 Alternative Routes of Administration ............................................................. 10

1.3 Production of Pegylated Proteins: Pegylation Reaction ............................................ 10

1.3.1 Random Pegylation ............................................................................................ 11

1.3.2 Site-Specific Pegylation ..................................................................................... 12

1.4 Purification of Pegylated Proteins ............................................................................. 14

1.4.1 Ion Exchange Chromatography ......................................................................... 14

1.4.2 Size Exclusion Chromatography ........................................................................ 15

1.4.3 Hydrophobic Interaction Chromatography ........................................................ 16

1.5 Membrane processes .................................................................................................. 16

1.5.1 Membrane Processes in the Biopharmaceutical Industry .................................. 16

1.5.2 Ultrafiltration of Pegylated Proteins .................................................................. 18

1.6 Combined Reaction and Separation Processes .......................................................... 19

1.7 Thesis Summary ........................................................................................................ 21

1.7.1 Overall Objectives .................................................................................................. 21

1.7.1 Thesis Outline ......................................................................................................... 22

Chapter 2 Theoretical Background ......................................................................................... 24

2.1 Introduction ................................................................................................................ 24

2.2 Bulk Mass Transport .................................................................................................. 24

2.2.1 Stagnant Film Model.......................................................................................... 26

2.2.2 Bulk Mass Transfer Coefficient ......................................................................... 28

2.3 Membrane Transport of Solvent ................................................................................ 30

2.4 Membrane Transport of Solute .................................................................................. 31

2.4.1 Thermodynamic Partition Coefficient ............................................................... 34

2.4.1.1 Steric Interactions ..................................................................................... 34

2.4.1.2 Electrostatic Interactions .......................................................................... 35

2.4.1.3 Solute-solute Intermolecular Interactions ................................................. 37

2.4.2 Hydrodynamic Analyses .................................................................................... 38

2.5 Pore Size Distribution: Effect on Membrane Transport ............................................ 40

Page 6: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

vi

2.6 Effective Protein Radius ............................................................................................ 42

2.7 Protein Net Charge .................................................................................................... 43

Chapter 3 Materials and Methods ........................................................................................... 47

3.1 Introduction ................................................................................................................ 47

3.2 Experimental Materials .............................................................................................. 47

3.2.1 Polyethylene Glycol (PEG) ................................................................................ 47

3.2.2 Activated Polyethylene Glycol .......................................................................... 49

3.2.3 Proteins .............................................................................................................. 50

3.2.3.1 Pegylated Proteins .................................................................................... 52

3.2.3.2 Acetylated proteins ................................................................................... 53

3.2.4 Ultrafiltration Membranes.................................................................................. 54

3.2.5 Buffer Solutions ................................................................................................. 56

3.3 Experimental Methods ............................................................................................... 57

3.3.1 Ultrafiltration Apparatus .................................................................................... 57

3.3.2 Membrane Hydraulic Permeability .................................................................... 58

3.3.3 Sieving Experiments .......................................................................................... 59

3.3.4 Diafiltration ........................................................................................................ 59

3.3.5 Membrane Charge Characterization .................................................................. 60

3.4 Assays ........................................................................................................................ 63

3.4.1 Size Exclusion Chromatography (SEC) ............................................................. 63

3.4.2 Capillary Electrophoresis (CE) .......................................................................... 67

Chapter 4 Effect of Electrostatic Interactions on Transmission of Pegylated Proteins

through Charged Ultrafiltration Membranes .................................................................... 69

4.1 Introduction................................................................................................................ 69

4.2 Materials and Methods .............................................................................................. 71

4.2.1 Pegylated Protein Preparation ............................................................................ 71

4.2.2 Acetylated Protein Preparation .......................................................................... 72

4.2.3 Ultrafiltration Membranes ................................................................................. 72

4.2.4 Protein Characterizations ................................................................................... 73

4.2.5 Ultrafiltration Sieving Experiments ................................................................... 73

4.3 Results and Analysis .................................................................................................. 74

4.3.1 Ultrafiltration of Pegylated Proteins .................................................................. 74

4.3.2 Electrophoretic Mobility .................................................................................... 79

4.3.3 Partitioning Model ............................................................................................. 86

4.4 Conclusion ................................................................................................................. 91

Chapter 5 Separation of Pegylated Proteins from Reactants using a Single Charge-

modified Membrane ......................................................................................................... 94

5.1 Introduction................................................................................................................ 94

5.2 Materials and Methods .............................................................................................. 95

5.3 Results........................................................................................................................ 96

5.3.1 Sieving Experiments .......................................................................................... 96

5.3.2 Purification of Mono-pegylated Protein ............................................................ 103

Page 7: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

vii

5.4 Conclusions................................................................................................................ 107

Chapter 6 Removal of Multiply Pegylated Proteins using Charged Ultrafiltration

Membranes ....................................................................................................................... 108

6.1 Introduction................................................................................................................ 108

6.2 Materials and Methods .............................................................................................. 109

6.2.1 Preparation of Pegylated Proteins ...................................................................... 109

6.2.2 Ultrafiltration Membranes ................................................................................. 110

6.2.3 Ultrafiltration Experiments ................................................................................ 111

6.2.4 Diafiltration Experiments................................................................................... 111

6.3 Results and Analysis .................................................................................................. 112

6.3.1 Ultrafiltration Results......................................................................................... 112

6.3.2 Model Calculations ............................................................................................ 115

6.3.3 Diafiltration Experiments................................................................................... 120

6.4 Conclusion ................................................................................................................. 127

Chapter 7 Intermolecular Interactions during Ultrafiltration of Pegylated Proteins ............... 129

7.1 Introduction................................................................................................................ 129

7.2 Materials and Methods .............................................................................................. 130

7.2.2 Ultrafiltration Membranes…………...…………..........….............................. ...131

7.2.3 Ultrafiltration Experiments ................................................................................ 131

7.3. Results and Analysis ................................................................................................. 132

7.3.1 Sieving Behavior at Low Filtrate Flux ............................................................... 132

7.3.2 Concentration Polarization Effects .................................................................... 139

7.3.2.1 PEG-PEG Interactions .............................................................................. 139

7.3.2.2 PEG-PEG1 Interactions ............................................................................ 143

7.3.3 Diafiltration Process - PEG Removal ................................................................ 146

7.3.4 Batch Ultrafiltration ........................................................................................... 149

7.4 Conclusions................................................................................................................ 153

Chapter 8 Combined reaction and membrane-based separation process for enhanced

yield of protein conjugates ............................................................................................... 155

8.1 Introduction................................................................................................................ 155

8.2 Reaction--Separation System ..................................................................................... 156

8.3 Materials and Methods .............................................................................................. 159

8.4 Results and Discussions ............................................................................................. 161

8.4.1 Batch Pegylation ................................................................................................ 161

8.4.2 Combined Reaction-Separation ......................................................................... 163

8.4.3 Model Simulations ............................................................................................. 167

8.5 Single-Pass Ultrafiltration Process ........................................................................... 176

8.6 Conclusions................................................................................................................ 184

Chapter 9 Conclusions and Recommendations ....................................................................... 187

Page 8: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

viii

9.1 Conclusions................................................................................................................ 187

9.1.1Electrostatic Effects ............................................................................................ 188

9.1.2 Purification of Pegylated Proteins using Charged Membranes.......................... 189

9.1.3 Solute-Solute Intermolecular Interactions ......................................................... 191

9.1.4 Combined Reaction-Separation Systems ........................................................... 192

9.2 Recommendations ...................................................................................................... 195

REFERENCES ........................................................................................................................ 199

APPENDIX .............................................................................................................................. 211

Page 9: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

ix

LIST OF FIGURES

Figure 1.1: Comparison of in vitro and in vivo bioactivity of a pegylated erythropoietin as a

function of grafted PEG mass. Adapted from Harris et al. (2001)...................7

Figure 1.2: Concentration profiles of interferon-α2a in the body as a function of time for

the native (top panel) and pegylated form (bottom panel). Taken from

Kozlowski et al. 2001).........................................................................................8

Figure 2.1: Schematic representation of concentration polarization of a solute near the

membrane surface with the concentration polarization boundary layer

thickness, �.........................................................................................................28

Figure 2.2: Actual sieving coefficient as a function of membrane Peclet number.............33

Figure 2.3: Calculated net charge of α-lactalbumin as a function of the solution pH at

different ionic strength .................................................................................... 46

Figure 3.1`: Molecular structure of linear polyethylene glycol............................................48

Figure 3.2: X-ray structure of bovine α-lactalbumin including ion binding sites (for Ca2+

and Zn2+) adapted from Permyakov and Berliner (2000). Disulfide bridges are

shown in yellow.................................................................................................51

Figure 3.3: Pegylation reaction between a PEG bearing an N-hydroxylsuccinimide ester

(PEG-NHS) and a primary amine on a protein (e.g., a lysine group)................53

Figure 3.4: Acetylation reaction between an acetic anhydride and a primary amine on a

protein.................................................................................................................54

Figure 3.5: SEM image of an UltracelTM membrane cross-section provided by the

manufacturer.......................................................................................................55

Figure 3.6: Schematic of the negative charge modification of an UltracelTM membrane by

attachment of sulfonic acid groups. Adapted from Molek (2008)....................56

Figure 3.7: Schematic of ultrafiltration stirred cell apparatus..............................................57

Figure 3.8: Streaming potential apparatus for measuring membrane surface charge. Taken

from Burns and Zydney (2000) with permission...............................................61

Figure 3.9: Streaming potential as a function of applied transmembrane pressure for an

unmodified 300 kDa UltracelTM membrane and for a negatively charged version

that was charged for 24 hr..................................................................................62

Page 10: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

x

Figure 3.10: Size exclusion chromatograms for a pegylation mixture performed with a

Superdex 200, 10/300 column using UV detector (top panel) and RI detector

(bottom panel)....................................................................................................65

Figure 3.11: Calibration curve for α-lactalbumin using UV detection at 280 nm. The slope

corresponds to the specific UV response for α-lactalbumin of 3.36 x 104

mAU·s/(g/L)......................................................................................................66

Figure 3.12: Calibration curves for α-lactalbumin and 20 kDa PEG using RI detection. The

slopes correspond to the specific RI response of 3.23 x 104 mAU·s/(g/L) for α-

lactalbumin and 2.65 x 106 nRIU·s/(g/L) for PEG............................................67

Figure 4.1: Observed sieving coefficient of a 5 kDa PEG (right panel) and a pegylated α-

lactalbumin with one 5 kDa PEG chain (left panel) as a function of ionic

strength through both an unmodified and a 12-hr charged 100 kDa composite

regenerated cellulose membrane........................................................................76

Figure 4.2: Scaled sieving coefficient of a pegylated α-lactalbumin with one 20 kDa PEG

chain as a function of solution ionic strength during ultrafiltration through both

an unmodified and a 12-h charged 100 kDa composite regenerated cellulose

membrane...........................................................................................................77

Figure 4.3: Observed sieving coefficients of a 5 kDa pegylated α-lactalbumin and a mono-

acetylated α-lactalbumin as a function of solution ionic strength for

ultrafiltration through a 12-hr charged composite regenerated cellulose

membrane...........................................................................................................79

Figure 4.4: Electrophoretic mobility of the pegylated α-lactalbumin with different size PEG

chains as a function of the number of substituted lysine groups. Also shown for

comparison are data for the acetylated proteins. Experiments were performed

using 10 mM Tris–Glycine running buffer at pH 8.1. Error bars represent plus

or minus one standard deviation of the experimental data. Solid curve is model

calculation for acetylated proteins as described in the text...............................82

Figure 4.5: Drag ratio as a function of the effective radius for pegylated proteins containing

2, 5, 10, 20, or 30 kDa PEG chains. The solid and dashed curves are model

calculations as discussed in the text...................................................................86

Figure 4.6: Actual sieving coefficients of a 5 kDa pegylated α-lactalbumin and a mono-

acetylated α-lactalbumin as a function of solution ionic strength for

ultrafiltration through a 12-h charged 100 kDa composite regenerated cellulose

membrane. Solid curves are model calculations as described in the text..........90

Page 11: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xi

Figure 4.7: Actual sieving coefficients of the acetylated and pegylated α-lactalbumin (with

5 kDa PEG chains) through a negatively charged cellulose membrane as a

function of the number of substituted lysine groups at both 10 mM (left panel)

and 200 mM (right panel) ionic strength. Solid curves are model calculations as

described in text.................................................................................................91

Figure 5.1: Observed sieving coefficients for mono-pegylated α-lactalbumin, native α-

lactalbumin, and 20 kDa PEG as a function of ionic strength through an

unmodified 300 kDa UltracelTM membrane (blank symbols) and a 24-hr

negatively charged version of the membrane (filled symbols)......................97

Figure 5.2: Selectivity between the mono-pegylated α-lactalbumin and either the native α-

lactalbumin or the 20 kDa PEG through an unmodified 300 kDa UltracelTM

membrane (blank symbols) and a 24-hr negatively charged version of the

membrane (filled symbols)..............................................................................100

Figure 5.3: Observed sieving coefficients for mono-pegylated α-lactalbumin, native α-

lactalbumin, and the 20 kDa PEG as a function of solution pH through a 24-hr

negatively charged version of 300 kDa UltracelTM membrane.......................101

Figure 5.4: Selectivity for the removal of native α-lactalbumin and PEG from the mono-

pegylated α-lactalbumin as a function of solution pH through a 24-hr negatively

charged version of the 300 kDa UltracelTM membrane in 0.5 mM buffers.....102

Figure 5.5: Yield for the native α-lactalbumin, PEG, and mono-pegylated α-lactalbumin in

the retentate solution as a function of number of diavolumes for a diafiltration

performed with a 300 kDa UltracelTM membrane charged for 24 hr. Data were

obtained at pH 6.6, 0.5 mM ionic strength, and a filtrate flux of 8 µm/s. Solid

curves are model calculations described in the text.......................................104

Figure 5.6: Yield for the mono-pegylated α-lactalbumin as a function of purification factor.

Circle symbols are for removal of the native α-lactalbumin; squares are for

removal of the PEG. Solid and dashed curves are model calculations described

in the text..........................................................................................................106

Figure 6.1: Selectivity between the mono- and di-pegylated α-lactalbumin as a function of

solution ionic strength for ultrafiltration through a 300 kDa UltracelTM

membrane charged for 24 hr. Data were obtained at pH 5 using a filtrate flux of

approximately 8 µm/s. The solid curve is the model calculation as described in

the text..............................................................................................................115

Figure 6.2: Selectivity between the mono- and di-pegylated α-lactalbumin as a function of

solution pH for ultrafiltration through a 300 kDa UltracelTM membrane charged

for 24 hr. Data were obtained using acetate or BisTris buffers with

approximately 0.5 mM ionic strength at a filtrate flux of approximately 8 µm/s.

The solid curve is the model calculation as described in the text...................118

Page 12: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xii

Figure 6.3: Mass throughput (J∆S) as a function of solution pH for ultrafiltration through a

300 kDa UltracelTM membrane charged for 24 hr. Data were obtained using 0.5

mM buffer at a filtrate flux of approximately 8 µm/s.....................................120

Figure 6.4: Yield for the mono-, di, and tri-pegylated α-lactalbumin in the filtrate solution

as a function of number of diavolumes for a diafiltration performed with a 300

kDa UltracelTM membrane charged for 24 hr. Data were obtained at pH 5, 0.4

mM ionic strength, and a filtrate flux of 8 µm/s. Solid curves are model

calculations.......................................................................................................121

Figure 6.5: Yield for the mono-pegylated α-lactalbumin as a function of purification factor.

Filled circles are for removal of the di-pegylated protein; filled squares are for

removal of the tri-pegylated protein. Solid and dashed curves are model

calculations.......................................................................................................123

Figure 6.6: Size exclusion chromatograms showing the initial feed and the final retentate

(top panel) and the final filtrate (bottom panel) solutions after a 10-diavolume

diafiltration at pH 5 and 0.4 mM ionic strength..............................................126

Figure 7.1: Observed sieving coefficients of the 20 kDa PEG, α-lactalbumin, and the

mono-pegylated α-lactalbumin as a function of the difference in PEG

concentrations between the bulk and filtrate solutions. Data obtained at a filtrate

flux of 2.3 µm/s in a 200 mM ionic strength solution at pH 7 using an

unmodified UltracelTM 30 kDa membrane. Dashed lines are linear regression

fits. Solid curves are model calculations discussed in more detail

subsequently.....................................................................................................136

Figure 7.2: Observed sieving coefficients of a 1.5 kDa PEG, α-lactalbumin, and the mono-

pegylated α-lactalbumin (with a 20 kDa PEG) as a function of the PEG

concentration difference between bulk and filtrate solutions for a low molecular

weight (1.5 kDa) PEG. Data obtained at a filtrate flux of 2.3 µm/s using a pH

7, 200 mM ionic strength buffer with an unmodified 30 kDa UltracelTM

membrane. Solid lines are model calculations for α-lactalbumin, and the mono-

pegylated α-lactalbumin as described in text.................................................139

Figure 7.3: Observed sieving coefficient of a 20 kDa PEG as a function of filtrate flux at

both low (1.2 g/L) and high (14 g/L) PEG concentrations in a pH 7 and 10 mM

ionic strength buffer using an unmodified 30 kDa UltracelTM membrane. The

dashed curves are model calculations using the classical concentration

polarization model while the solid curves are those using the modified

concentration polarization model as described in the text.............................140

Figure 7.4: Observed sieving coefficients of the mono-pegylated α-lactalbumin as a

function of filtrate flux at both low (1.2 g/L) and high (14 g/L) concentrations

of the 20 kDa PEG in a pH 7 and 10 mM ionic strength buffer using an

Page 13: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xiii

unmodified 30 kDa UltracelTM membrane. The solid curves are the numerical

solution to the full model. The dashed curves are an approximate solution as

described in the text..........................................................................................144

Figure 7.5: Normalized concentrations of the mono-pegylated α-lactalbumin and the 20

kDa PEG as a function of the number of diavolumes for a diafiltration

performed with a negatively charged 300 kDa Ultracel membrane at pH 8, 2

mM ionic strength, and a filtrate flux of 8 um/s. Solid and dashed curves are

model calculations as described in the text.....................................................147

Figure 7.6: Filtrate product loss of mono-pegylated α-lactalbumin as a function of volume

concentration factor (VCF) at a filtrate flux of 10 µm/s. Data obtained in a pH

7 and 10 mM ionic strength buffer using an unmodified 10 kDa UltracelTM

membrane. Solid and dashed curves are model calculations as described in the

text....................................................................................................................150

Figure 7.7: Filtrate product loss of mono-pegylated α-lactalbumin as a function of volume

concentration factor (VCF) at a filtrate flux of 10 µm/s. Data obtained in a pH

7 and 10 mM ionic strength buffer using an unmodified 30 kDa UltracelTM

membrane. Solid and dashed curves are model calculations as described in the

text....................................................................................................................151

Figure 7.8: Calculated filtrate product loss of mono-pegylated α-lactalbumin as a function

of volume concentration factor (VCF). Model calculations were performed

using Jv/km = 3 with Sao = 10-4. Solid and dashed curves are model calculations

as described in the text.....................................................................................153

Figure 8.1: Schematic of the reaction and membrane-based separation system..............157

Figure 8.2: Schematic of Pellicon XLTM tangential flow filtration module (image provided

by Millipore Corp.)..........................................................................................160

Figure 8.3: Concentration of α-lactalbumin, 20 kDa PEG, and the differently pegylated α-

lactalbumins as a function of time for a batch reaction at pH 7. Curves are

model calculations as described in the text.....................................................162

Figure 8.4: Concentration of mono-pegylated α-lactalbumin as a function of time for the

reaction-separation system. Solid curves are model calculations for the

reaction-separation process as described in the text. Dashed curves are

corresponding model results for a batch process with different molar ratio of

PEG (N) relative to the mass of initial α-lactalbumin......................................166

Figure 8.5: Yield of mono-pegylated α-lactalbumin in the reactor, product tank, and in the

system as a whole (total yield) as a function of time. Curves are model

calculations as described in the text.................................................................167

Page 14: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xiv

Figure 8.6: Concentration of mono-pegylated, multiply-pegylated, and native α-

lactalbumin as a function of process time for the reaction-separation system.

Top panel is for the product tank operated at pH 4 while the bottom panel was

at pH 7..............................................................................................................169

Figure 8.7: Model calculations for the dimensionless mass of mono-pegylated, multiply-

pegylated, and native α-lactalbumin as a function of process time for the

reaction-separation system operated with the product tank at pH 4 (left panel)

and at pH 7 (right panel)..................................................................................170

Figure 8.8: Model calculations for the dimensionless mass mono-pegylated, multiply-

pegylated, and native α-lactalbumin for the combined reaction-separation

system as a function of membrane selectivity.................................................171

Figure 8.9: Model calculations for the dimensionless mass of α-lactalbumin, the mono-

pegylated protein, and the multiply-pegylated species in the combined reaction-

separation system as a function of the residence time in the reactor (left panel)

and product tank (right panel)..........................................................................173

Figure 8.10: Model calculations for the dimensionless species mass for the combined

reaction-separation system as a function of total process time for a constant

amount of PEG addition...................................................................................174

Figure 8.11: Model calculations for the dimensionless species mass for the combined

reaction-separation system as a function of total PEG feed molar ratio (moles of

added PEG to initial moles of α-lactalbumin)..................................................175

Figure 8.12: Schematic of the single-pass reaction and membrane-based separation

system...............................................................................................................177

Figure 8.13: Model calculations for the concentration of mono-pegylated, multiply-

pegylated and native α-lactalbumin as a function of process time for the single-

pass reaction-separation system performed with the base-case conditions.....179

Figure 8.14: Model calculations for the dimensionless mass of mono-pegylated, multiply-

pegylated, and native α-lactalbumin in the product tank (collected from the

retentate outflow) as function of process time for the single-pass reaction-

separation system.............................................................................................180

Figure 8.15: Model calculations for the dimensionless mass of α-lactalbumin, the mono-

pegylated protein, and the multiply-pegylated species produced by the single-

pass system as a function of the residence time in the reactor (left panel) and

UF module (right panel)...................................................................................181

Page 15: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xv

Figure 8.16: Model calculations for the yield of mono-pegylated protein as a function of

single-pass conversion for different values of the membrane selectivity. Dashed

curves are results assuming that there is no reaction in the UF module.........183

Figure 8.17: Product distribution for the production of pegylated protein obtained from the

batch reactor and the different reaction-separation schemes...........................184

Page 16: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xvi

LIST OF TABLES

Table 1.1: Commercially available pegylated therapeutics................................................4

Table 2.1: Expansion coefficients for hydrodynamic functions Kt and Ks..........................40

Table 2.2: Number (ni) and i

apK values of the charged amino acids in α-lactalbumin (Brew

et al., 1970; Nelson and Cox, 2008)......................................................................45

Table 3.1: Basic physical / chemical properties of 20 PEG, native, and 20 kDa pegylated α-

lactalbumins...........................................................................................................51

Table 3.2: Approximated effective pore size for UltracelTM membranes..............................54

Table 5.1: Best-fit values of the protein sieving coefficients for the diafiltration process...105

Table 6.1: Observed sieving coefficients for mono-, di-, and tri- pegylated α-lactalbumin for

membranes charged for different periods of time. Data were obtained in a 0.5 mM

acetate buffer at pH 5 using a filtrate flux of 8 µm/s..........................................113

Table 6.2: Best-fit values of the protein sieving coefficients for the diafiltration process...124

Table 7.1: Sieving coefficients of the unmodified α-lactalbumin, the 20 kDa PEG, and the

mono-pegylated α-lactalbumin alone and in mixtures with low (0.4 g/L) and high

(23 g/L) PEG concentrations. Data were obtained at a filtrate flux of Jv ≈ 2.3

µm/s in a pH 7, 200 mM ionic strength buffer using an unmodified 30 kDa

UltracelTM membrane..........................................................................................134

Table 8.1: Rate constants for the pegylation reaction of α-lactalbumin with 20 kDa PEG-

NHS in 1 mM bis-Tris (pH 6, 7, and 8) or acetate buffer (pH 4 and 5)...........163

Page 17: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xvii

ACKNOWLEDGEMENTS

This dissertation would not have been possible without the help and support from

many people that I have met along the way. I am most grateful to my dissertation advisor, Dr.

Andrew Zydney, for his support throughout my doctoral research. His depth of knowledge

and his invaluable guidance is very essential to the completion of this dissertation. I am

always amazed by his dedication, helpfulness, and courteous treatment of others, which

might have been the most important lessons of my graduate career that I have learned from

him. I also would like to thank my dissertation committee: Dr. Darrell Velegol, Dr. Michael

Janik, and Dr. Arnold Fontaine for their careful reading of the dissertation and their valuable

comments. I would like to express my sincere thanks to Dr. Manish Kumar for allowing me

to work in his lab during the last two semesters and for his advice on an industrial career

path.

My time in Zydney’s lab has been an invaluable experience. I have enjoyed the

company of both former and current members of our group including Mahsa Rohani, Meisam

Bakhshayeshi, Dharmesh Kanani, Dave Latulippe, Ehsan Borujeni, Achyuta Teella, Elaheh

Binabaji, Melissa Woods, and Mahsa Hadidi. I would like to specifically thank Meisam

Bakhshayeshi, Dave Latulippe, and Dharmesh Kanani, who helped teach me lab basics when

I first started in the lab. I also would like to express my sincere appreciation to Mahsa Rohani

for being a wonderful officemate and an amazing mentor. Thank you for taking your time

and being patient with my endless questions. I also learned a great deal from her during the

first few months in the lab helping with her project. The company of Ehsan Borujeni and

Page 18: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

xviii

Achyuta Teella has made long hours in the lab less tedious and oftentimes humorous. Melissa

Woods has always been helpful and cheerful. Thanks to Elaheh Binabaji and Mahsa Hadidi

for always keeping our lab clean and organized. Most of all, I have just enjoyed the

conversation and the time I spent in the lab with these wonderful group of people.

I was also fortunate to work with a number of talented and dedicated undergrad

students: Megumi Woltermann, Monica Perez, and Chris Rinschler. I wish them the best for

their future endeavor. I would like to recognize Jessica Molek, a former PhD graduate from

our group, who has done such an amazing work on ultrafiltration of pegylated proteins. I also

would like to thank her for the contributions to Chapter 4, especially the results and

discussions regarding the electrophoretic mobility. I also would like to thank the

administrative and technical staff from the Chemical Engineering department: Roger

Dunlap, MJ Smith, Stephen Black, Sue Ellen Bainbridge, Cathy Krause, Chris Jabco, Lisa

Haines, and Steven Smith for always being extremely helpful.

I would like to thank family: my dear mom and dad, Tuanjai and Manop

Ruanjaikaen, and my sister Jirattikan Ruanjaikaen for always believing in me and always

being there for me. I also would like to thank my cousin Manassanant Hansen for constantly

sending me Thai food and always making sure that I feel like home here.

Page 19: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

1

Chapter 1

Introduction

1.1 Therapeutic Proteins

The advent of recombinant DNA technology has enabled the production of

therapeutic proteins at large scale by cloning a foreign gene into a fast-growing host

organism. Protein therapeutics have since emerged as an important class of

biopharmaceuticals with the first recombinant human insulin (Humulin®) approved by the

U.S. Food and Drug Administration (FDA) in 1982 (Carter, 2011). The use of protein

therapeutics has a number of advantages over small-molecule drugs due to their highly

specific and complex biological response and the lower probability of adverse effects. The

earliest therapeutic proteins were recombinant version of naturally occurring proteins used to

replace a hormone or enzyme that was deficient or abnormal, e.g. insulin for the treatment of

diabetes or Factor VIII for the treatment of hemophilia. Recombinant proteins can also

provide novel functionality that may not be expressed naturally, for example asparaginase for

the treatment of leukemia (Leader et al., 2008).

Recombinant proteins are currently used for treatment of a wide range of deceases

including cancers, hepatitis, diabetes, arthritis, multiple sclerosis, hemophilia, etc. (Dimitrov,

2012). The past decade has also seen the rapid development and clinical application of

monoclonal antibodies (mAbs) for the treatment of cancers and immune disorders (Reichert,

2011). Protein-based vaccines are also currently used to generate protection against infectious

deceases (U.S. Food and Drug Administration, 2013), including influenza, hepatitis A and B,

Page 20: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

2

and most recently human papillomavirus (HPV) infections (Gardasil®). There are more than

200 products approved for clinical use and 2010 sales exceeded 100 billion dollars in the

USA and European Union (Dimitrov, 2012).

Second generation therapeutic proteins have recently been developed to address some

of the shortcomings of the natural products including low stability and solubility, short serum

half-lives, immunogenicity, and toxicity. These new products are typically produced by post-

translational modification of the natural protein (Carter, 2011), including Fc fusion proteins

with increased plasma half-life and targeted delivery (Czajkowsky et al., 2012) and

glycosylated proteins with enhanced biological function/activity (Walsh and Jefferis, 2006).

Another very attractive approach to generate enhanced therapeutics is conjugation of

a polymer chain(s) to the base protein. The major impact of protein-polymer conjugation is to

increase the biological half-life of the therapeutic by increasing the hydrodynamic volume

(size) of the molecule, significantly reducing clearance by the kidney (particularly for small

proteins that are able to pass through the glomerular membrane). Several naturally and

synthetically derived polymers have been studied for protein conjugation including

polyethylene glycol (PEG), N-(2-hydroxypropyl) methacrylamide copolymer (HPMA),

polysaccharides, and polyamino acids (Pasut and Veronese, 2007). The most clinically

successful approach has been the conjugation with PEG (typically referred to as pegylation);

the pegylated products not only provide greater half-life and enhanced therapeutic efficacy,

they also exhibit superior biocompatibility and low-toxicity (Carter, 2011; Pasut and

Veronese, 2007).

Page 21: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

3

1.2 Pegylated Therapeutic Proteins

Pegylation was first discussed by Abuchowski et al. (1997a, 1997b) as a method to

improve in vivo-half-life and reduce the immunogenicity of a protein. They demonstrated

that a pegylated bovine serum albumin (BSA) was maintained in vivo at a higher

concentration for a longer period of time compared to the unmodified BSA after injection

into rabbits (Abuchowski et al., 1977b). In addition, their results showed that the pegylated

proteins had reduced immunogenicity (Abuchowski, et al., 1977b) and slower degradation

rates (Abuchowski, et al., 1977a) compared to the native BSA.

The first pegylated therapeutic protein approved by FDA was Adagen®, a pegylated

bovine adenosine deaminase introduced by Enzon Pharmaceuticals in 1990 for the treatment

of severe combined immunodeficiency decease. To date, ten pegylated products have

received FDA approval; nine of which are pegylated proteins and one is a pegylated

oligonucleotides. Table 1.1 provides a list of these FDA approved therapeutics including

company information, primary indication, and number/molecular weight of the conjugated

PEG (Veronese and Pasut, 2005; Jevsevar et al., 2010; Li et al., 2013).

Page 22: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

4

Table 1.1 Commercially available pegylated therapeutics

Name Drug name Company PEG size (Da) Indication Year approved

Pegylated proteins:

Adagen® Pegadamase Enzon Multiple linear 5,000 SCID 1990

Oncaspar®

Pegaspargase Enzon Multiple linear 5,000 Leukemia 1994

PEG-INTRON® Peginterferon-α2b Schering-Plough Linear 12,000 Hepatitis C 2000

PEGASYS® Peginterferon-α2a Hoffman-La Roche Branched 40,000 Hepatitis C 2001

Neulasta® Pegfilgrastim Amgen Linear 2,000 Neutropenia 2002

Somavert® Pegvisomant Pharmacia & Upjohn 4-6 linear 5,000 Acromegaly 2003

Mircera ® mPEG-epoetin-β Hoffman-La Roche Linear 30,000 Anemia / renal failure 2007

Cimzia® Certolizumab pegol UCB Branched 40,000 Crohn’s disease 2008

Rheumatoid arthritis 2009

Puricase1®/ PEG-uricase Savient Multiple 10,000 Gout 2010

Krystexxa®

Pegylated oligonucleotides:

Mucagen® Pegaptanib Pfizer Branched 40,000 Age-related macular 2004

Degeneration

SCID: severe combined immunodeficiency disease

Page 23: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

5

In addition to the pegylated products on the market, several others are currently in

different stages of development, and several of these are expected to receive FDA approval in

the near future. For example, pegylated interferon-β1a from BiogenIdec received a Fast

Track designation from the FDA for a global phase III clinical trial for the treatment of

multiple sclerosis (Baker et al., 2006; Jevsevar et al., 2010); pegylated C-peptide

(CBX129081) from Cebix is in Phase II clinical trial for the treatment of type 1 diabetes

(Shah, 2013). There is also increasing interest in the use of pegylation for small-molecule

drugs, especially for anti-tumor agents to improve solubility and sustain in vivo release

(Kang et al, 2009; Li et al., 2013). Four of these products are currently in clinical trials

including pegylated irinotecan (phase II/III), pegylated docetaxel (phase I), pegylated SN38

(phase II) for the treatment of solid tumor, and orally administered PEG-naloxol for the

treatment of opioid-induced bowel dysfunction and constipation (phase III).

1.2.1 Benefits of Pegylation

The biocompatibility of PEG itself contributes to the success of pegylated

therapeutics. PEG has a long history of use as a non-toxic, non-immunogenic, and non-

biodegradable polymer, and it has been approved by FDA as “generally recognized as safe”.

PEG has been previously used in the food and cosmetic industries, as well as in the

pharmaceutical industry as an excipient for parenteral, topical, and ocular application (Knop

et al., 2010). PEG has also been used in blood and organ storage to reduce aggregation of red

blood cells (Mosbah et al., 2006). PEG copolymers are used as coatings for cardiovascular

devices, e.g. heart stents, in order to reduce thrombosis (Acharya et al., 2012).

Page 24: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

6

The improved therapeutic efficacy of pegylated proteins compared to the native

protein is due to changes in the protein half-life within the body, the biological activity, and /

or the immunogenicity of the pegylated molecule (Caliceti and Veronese, 2003).

1.2.1.1 Prolonged Circulation Time

One of the major benefits of pegylation is the increase in circulation half-life within

the body, resulting from a reduction of glomerular (renal) filtration rate. The extent of renal

filtration is primary controlled by the size of the molecule; proteins larger than 70 kDa (close

to the molecular weight of serum albumin) are largely retained by the glomerulus while

smaller proteins pass through the glomerular membrane and are secreted in the urine. Caliceti

and Veronese (2003) showed that a linear PEG molecule larger than 30 kDa was nearly

completely retained during renal filtration. This reflects the larger hydrodynamic radius of

the PEG due to its coiled / extended conformation compared to the compact globular

structure of proteins. In general, conjugation of one or two chains of high molecular weight

PEG is sufficient to produce a pegylated protein with enhanced half-life and high biological

activity (Caliceti and Veronese, 2003). For example, the attachment of a 10-20 kDa PEG

increased the circulation half-life of a recombinant interleukin-2 several fold (Katre et al.,

1987). Similarly, Gaertner and Offord (1996) reported that attachment of a single branched

40 kDa PEG to a small 10 kDa IL-8 increased the protein half-life more than 6-fold. Studies

provided by Clark et al. (1996) demonstrated that other mechanisms, such as reduced

proteolysis (hydrolysis of the peptide bonds by protease) in serum, also contribute to the

enhanced half-life of the proteins.

Page 25: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

7

The increase in circulation half-life due to pegylation can sometimes be offset by a

reduction in the biological activity of the molecule due to blockage of the active site(s) by the

attached PEG. The optimal degree of pegylation thus reflects a balance between increasing

clearance while maintaining sufficient biological activity. In general, a single PEG

attachment per protein is more likely to conserve biological activity, especially when the

activity depends on interactions between the protein and another molecule (Harris et al.,

2001). The balance between the reduced activity and increased half-life is shown in Figure

1.1 for data for a pegylated cytokine (erythropoietin) where pegylation provides a significant

increase in in vivo biological activity despite the reduction of the intrinsic (in vitro) activity

(Bailon and Berthold, 1998; Harris et al., 2001).

Figure 1.1 Comparison of in vitro and in vivo bioactivity of a pegylated erythropoietin as a

function of grafted PEG mass. Adapted from Harris et al. (2001).

Page 26: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

8

The increase in the in vivo half-life can directly benefit the patient by reducing the

amount and frequency of the required dosage. For example, Figure 1.2 shows data for the in

vivo concentration profile of unmodified interferon-α2a administered three times per week;

the plasma concentration dropped by more than an order of magnitude between doses. In

contrast, a single injection of PEGASYS® (pegylated interferon-α2a) gave a relatively

constant concentration profile over a full week, allowing for a far more convenient

administration schedule (Kozlowski et al., 2001).

Figure 1.2 Concentration profiles of interferon-α2a in the body as a function of time for the

native (top panel) and pegylated form (bottom panel). Taken from Kozlowski et

al. (2001).

Page 27: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

9

1.2.1.2 Reduced Immunogenicity and Side Effects

One of the main challenges of using proteins for therapeutic purposes especially those

of nonhuman origin is the risk of immunogenic response, which can induce adverse side

effects for the patients. Several studies showed that pegylation can lower the immunogenic

response by masking the surface of the protein and potential immunogenic sites, therefore

preventing recognition by the immune system (Harris et al., 2001). For example, Abuchowski

et al. (1977b) demonstrated that bovine serum albumin was rapidly cleared from rabbits by

immune-complex formation; however, the pegylated version showed minimal clearance by

blocking the development of antibodies. Chapman (2002) reported a similar behavior for IgG

versus pegylated IgG administered to monkeys.

A typical example of the benefits of pegylation is the use of uricase for the treatment

of gout. Unlike most mammals, humans lack the enzyme uricase, which is capable of

degrading high levels of uric acid in patients with hyperuricemia and gout. Sherman et al.

(2008) showed that a pegylated version of uricase had sufficiently low immunogenicity to

permit repeated dosing in clinical trials. Similar advantages have been reported for the

treatment of acute leukemia using pegylated asparaginase. The native enzyme causes

significant adverse reactions, including allergic reactions leading to anaphylactic shock

(Kawashima et al., 1991). The pegylated asparaginase showed much lower immune response,

allowing patients with hypersensitivity to the native enzyme to tolerate the pegylated version

(Keating et al., 1993; Harris et al., 2001).

Page 28: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

10

1.2.1.3 Alternative Routes of Administration

Although therapeutic proteins are usually administered intravenously, attachment of

the PEG opens up possibilities for alternative administration routes. Studies on the

pharmacokinetics profiles of pegylated superoxide dismutase (SOD) demonstrated that

reasonable bioavailability of the pegylated protein was maintained in both intramuscular and

subcutaneous administration, in sharp contrast to the very low bioavailability of the native

protein (Veronese et al., 1989).

Insulin analogs that can be administered orally are also of increasing interest for the

treatment of type 1 and 2 diabetes. Results from Phase II clinical trials demonstrated that an

orally administered pegylated insulin (lispro) provided comparable glycaemic control and

less hypoglycemia compared to a currently available insulin administered subcutaneously.

The pegylated insulin also appear to have greater flexibility with time-of-the-day dosing

(Zinman, 2013).

1.3 Production of Pegylated Proteins: Pegylation Reaction

Pegylated proteins are inherently more expensive to produce compared to the native

counterpart due to the additional costs associated with the pegylation reaction, the subsequent

separation, and the additional analyses required to demonstrate the success of the pegylation.

In general, pegylation is performed on highly purified proteins to reduce separation

challenges and improve process consistency/validation (Fee and van Alstine, 2006; Hoyle,

1991). Pegylation is performed by covalent reaction between the protein and an activated

Page 29: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

11

version of PEG. A number of chemistries have been developed for pegylation of different

sites on a protein; each offers its own advantages and disadvantages.

1.3.1 Random Pegylation

The most common chemistry for pegylation targets the ɛ-amino group on the lysine

residues of proteins. Generally, lysines account for approximately 10% of the amino acid

residues in proteins, which represents both opportunities and challenges for protein

conjugation. Their availability allows the pegylation reaction to proceed quickly under mild

condition but results in a heterogeneous mixture of conjugates with different degree of

pegylation. Although a number of activation chemistries have been employed, the most

common pegylation reagents are N-hydroxylsuccinimide (NHS) esters of PEG, which form

stable protein-PEG conjugates via amide bonds (Jevsevar et al., 2010). Depending on the

reaction conditions (e.g., reaction time, temperature, pH, protein, and activated PEG

concentration), mono-, di-, tri- and higher order pegylated conjugates are formed. The

resulting mono-pegylated conjugate can be a heterogeneous population of positional isomers

with the PEG chain attached to different available sites on the protein, some of which can

differ significantly in their biological properties.

Due to its simplicity, most of the commercial pegylated proteins have been produced

via random pegylation (Gaberc-Porekar et al., 2008; Jevsevar et al., 2010; Fee and Van

Alstine, 2006). The first two pegylated proteins, Adagen® and Oncaspar®, are actually

mixtures containing proteins with different degrees of pegylation. They exhibit distinctly

improved therapeutic properties over the native enzymes, in particular increased serum half-

life and decreased immunogenicity, respectively. Subsequently approved products, PEG-

Page 30: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

12

Intron® PEGASYS®, and Mircera® are mono-pegylated proteins produced via random

pegylation and then purified to remove the higher order pegylation products. Somavert® is a

human growth hormone pegylated with four to five chains of 5 kDa PEG.

1.3.2 Site-Specific Pegylation

Due to the challenges associated with product heterogeneity, several approaches have

been developed to increase the specificity of the pegylation reaction to obtain a molecularly

defined mono-pegylated product. One example is site-specific pegylation at the N-terminal of

the protein used in Neulasta® (Kinstler et al., 2002). The amino group of the N-terminal

amino acid has a lower pKa value compared to the ɛ-amino groups on lysine residues. Thus,

pegylation at a low pH (around pH 5) preferentially targets the unprotonated amino group of

the N-terminus. However, the reaction typically requires large excess amount of PEG due to

the low reactivity and this approach is only feasible if the N-terminus is not required for the

desired biological/therapeutic activity (Seely et al., 2005)

Another approach is to target the thiol group of free cysteine residues. The reaction

typically employs a PEG bearing a maleimide group due to its specificity to the thiol group

and the stability of the resulting linkage (Jevsevar et al., 2010). However, most proteins do

not have free cysteine residues available for conjugation since they are usually in the form of

disulfide bonds or their presence is required for biological activity. Several studies have

demonstrated that an unpaired cysteine can be genetically introduced using site-directed

mutagenesis; this approach was examined for both pegylated granulocyte macrophage

colony-stimulating factor (GM-CFS) (Doherty et al., 2005) and pegylated interferon-α2

Page 31: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

13

(Rosendahl et al., 2005). However, there are a number of technical challenges including

disulfide scrambling and protein mis-folding (Gaberc-Porekar et al., 2008).

More recently, Balan et al. (2007) reported a novel approach for specific pegylation

of a variety of proteins, including interferon-α2b and L-asparaginase, by targeting an

accessible, natural disulfide bond. The pegylation reaction proceeds by reduction of the

disulfide bond to release the two cysteine thiols followed by bis-alkylation using a three-

carbon bridge attached to the PEG chain. However, experiments with interferon-α2b, which

has two accessible disulfide bonds, gave a significant amount of di-pegylated species with

only 60-70% yield of mono-pegylated protein (Brocchini et al., 2008).

Another interesting approach that can be used for site-specific pegylation is the

incorporation of a non-natural amino acid in the protein. For example, phenylalanine bearing

an azido group was genetically introduced into proteins and then specifically reacted with

PEG bearing an alkyne group (Deiters et al., 2004; Nguyen et al., 2009). A mono-pegylated

human growth hormone produced using this technology (AmberTM Technology) is currently

in clinical trials (Jevsevar et al., 2010). Although this approach can afford a high specificity

and yield, the introduction of a non-natural amino acid residue is very time-consuming and

can potentially alter the protein’s biological activity and immunogenicity, limiting the use of

this technology (Thordarson et al., 2006).

Enzymatic approaches have also been developed for site-specific pegylation (Sato,

2002). For example, transglutaminase has been used to catalyze the incorporation of PEG

bearing an alkylamine group into a protein at natural or genetically introduced glutamine

residues. The process provided relatively high selectivity; however, the conjugation is

possible only when the target glutamine residue is present in a flexible or unfolded region of

the protein (Payne et al., 2011).

Page 32: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

14

1.4 Purification of Pegylated Proteins

There are two basic purification challenges in the production of pegylated proteins:

(i) the removal of the reactants (PEG and native protein) and small reaction by-product(s),

and (ii) the purification of the desired pegylated form from species with different degrees of

pegylation. This is typically accomplished using chromatographic approaches exploiting

differences in electrical charge, hydrodynamic radius, and hydrophobicity.

1.4.1 Ion Exchange Chromatography

Ion exchange chromatography has been used most frequently for the purification of

pegylated proteins. Cation chromatography can be used for separation of native protein, PEG,

and multiply pegylated proteins from the desired product (Kinstler et al., 2002; Fee and Van

Alstine, 2006; Edwards et al., 2003) exploiting the charge shielding provided by the PEG and

/ or the difference in net charge of the pegylated species associated with the conversion of the

positively-charged amino group into a neutral amide by the pegylation reaction. The more

heavily pegylated species typically elutes first in the presence of a salt gradient, with the

unmodified protein eluting last.

Unreacted PEG does not bind to the ion exchange resin and is eluted in the flow

through. However, the presence of PEG can reduce the resolution during chromatographic

separation. Therefore, the PEG is usually removed as soon as possible in the purification

process (Fee and Van Alstine, 2006). In addition, the presence of unreacted PEG can make

the pegylation solution very viscous, leading to high back pressure and column fouling.

Page 33: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

15

These can be avoided by dilution of the pegylation mixture before loading onto the column

(Jevsevar et al., 2010).

Although several studies have demonstrated the feasibility of using ion exchange

chromatography (Pabst et al., 2007; Piquet et al., 2002; Lee et al., 2008; Yun et al., 2005), the

attached PEG typically causes a dramatic reduction in dynamic binding capacity (on the

average of 10-fold) by shielding the protein surface charge, by providing a steric hindrance

for binding, and / or by reducing mass transfer rates (Fee and van Alstine, 2006). For

example, the dynamic binding capacity for pegylated BSA with a 30 kDa PEG to a

Fractoprep TMAE anion exchange resin was reduced by more than 100-fold compared to that

of the native protein (Pabst et al., 2007). Moosmann et al. (2010) reported similar trends for

mono-pegylated lysozyme using cation exchange chromatography.

1.4.2 Size Exclusion Chromatography

Size exclusion chromatography has been used extensively for small analytical scale

separation of pegylated productions. It can be used to remove low molecular weight

impurities including the by-products formed by hydrolysis of the activated group on the PEG

as well as the native protein. The separation efficiency depends on the molecular size

difference between the species; typically a ratio of hydrodynamic radii larger than 1.26

enables efficient separation (Fee and van Alstine, 2006). The separation resolution is

typically low for the higher order pegylated species and decreases with the degree of

pegylation. The main disadvantage of size exclusion chromatography is the large column

requirement, since only about 3-5% of the total column volume can be loaded per cycle to

obtain reasonable resolution (Fee and Damodaran, 2012). The large process volumes

Page 34: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

16

typically make SEC unsuitable for production of pegylated proteins at commercial scale

(Morar et al., 2006; Seely and Richey, 2001).

1.4.3 Hydrophobic Interaction Chromatography

Hydrophobic interaction chromatography can also be used for purification of

pegylated proteins, although most studies have been performed at analytical scale. For

example, Youn et al. (2004) purified pegylated growth hormone releasing factor using a C-8

hydrophobic chromatography column. The limited use of this method is due to poor

resolution between the differently pegylated species and the tendency of unreacted PEG to

bind to the column; the removal of PEG is somewhat unpredictable and depends on the

difference in size and hydrophobicity of the protein and PEG (Jevsevar et al., 2010). It is

possible to combine this approach with other chromatographic methods. For example, Zhang

et al. (2007) employed ion exchange chromatography to remove residual PEG and then used

hydrophobic chromatography to resolve mono-pegylated insulin from native and multiply

species.

1.5 Membrane processes

1.5.1 Membrane Processes in the Biopharmaceutical Industry

Membrane processes are very attractive for separation of biomolecules since they are

typically operated at mild conditions that cause little degradation or denaturation of the

biological product. A previous study comparing ultrafiltration and size exclusion

Page 35: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

17

chromatography for concentration and buffer exchange of therapeutic proteins has clearly

demonstrated the advantages of using ultrafiltration in terms of cost, throughput, and plant

space (Kurnik et al., 1995). Membrane processes have been used throughout downstream

processing in the biopharmaceutical industry, with the greatest interest in applications of

microfiltration, virus filtration, and ultrafiltration (van Reis and Zydney, 2007).

Microfiltration membranes with pore size between 0.05 and 10 µm are designed to retain

cells and cell debris while allowing proteins and smaller solutes to pass into the filtrate (van

Reis and Zydney, 2007). Microfiltration is commonly used for sterile filtration and bioburden

reduction to remove bacteria and particles from feedstock solutions. Microfiltration can also

be employed to harvest therapeutic proteins from the fermentation broth using either

tangential flow filtration or depth filtration, with the latter typically used in combination with

centrifugation (Russell et al., 2007).

Virus filters can provide a robust, size-based viral clearance mechanism that

complements other virus clearance steps (e.g. low pH inactivation, solvent / detergent

inactivation, UV inactivation). Virus filtration has become a fairly standard component of

most downstream purification processes (Phillips et al., 2007).

Ultrafiltration using membranes with pore size between 1 and 20 nm is widely used

for protein concentration and buffer exchange in large scale production of nearly all

recombinant proteins (van Reis and Zydney, 2007). Although ultrafiltration was originally

viewed as a purely size-based separation, it is now well established that solute transmission is

determined by both steric and electrostatic interactions between the protein and membrane

(van Reis and Zydney, 2007; Burns and Zydney, 2001; Mehta and Zydney, 2006). The

electrostatic interactions have been exploited to achieve high resolution protein separations,

with the charged membrane retaining the like-charged proteins/biomolecules while allowing

Page 36: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

18

relatively uncharged solutes to be recovered in the filtrate. Several studies have demonstrated

the potential of using charged ultrafiltration membranes, including purification of an antigen

binding fragment (Fab) from BSA (van Reis et al., 1999), purification of a monoclonal

antibody from Chinese hamster ovary (CHO) host cell proteins (Mehta et al., 2008), and

purification of an antibody fragment from E. Coli host cell proteins (Lebreton et al., 2008).

Optimization of these membrane separations typically involves selection of solution ionic

strength, pH, and membrane charge to control the extent of electrostatic interactions.

1.5.2 Ultrafiltration of Pegylated Proteins

There has been considerable interest in the use of membrane systems for the

purification and concentration of pegylated proteins (Mayolo-Deloisa et al., 2011). Since

pegylated proteins are frequently administered at a relatively high concentration (typically 10

g/L), ultrafiltration is well suited for the final processing step (Jevsevar et al., 2010).

Ultrafiltration has been used to concentrate a variety of pegylated proteins including α-

interferon (Arduini et al., 2004), human growth hormone (Clark et al., 1996), methioninase,

(Tan et al. 1998), and tumor necrosis factor receptor (Edwards et al., 2003).

Ultrafiltration/diafiltration (UF/DF) processes have also been used to remove small

impurities and achieve the desired final formulation for pegylated tumor necrosis factor

(Stoner et al., 2004) and pegylated gelonin, a ribosome inactivating protein (Arpicco et al.,

2002).

Molek and Zydney (2007) demonstrated the feasibility of using a two-stage

ultrafiltration/diafiltration system to remove unreacted protein, small reaction by-products,

and unreacted PEG from pegylated α-lactalbumin. The first stage employed an unmodified

Page 37: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

19

30 kDa ultrafiltration membrane to remove native α-lactalbumin and N-hydroxysuccinimide,

with the separation based on the differences in solute size. The resulting retentate was then

processed by a second diafiltration process using a negatively charged 100 kDa membrane at

low ionic strength to remove the neutral, unreacted 20 kDa PEG in the filtrate while the

pegylated protein was retained by a combination of steric and electrostatic interactions.

Molek and Zydney (2006) performed a fundamental study of the ultrafiltration

characteristics of pegylated proteins arising from the molecular flexibility of the grafted PEG.

They found that the sieving coefficients for a pegylated protein depended on both the total

mass of attached PEG and the number of PEG branches. This behavior was apparently due to

the deformation of the PEG chains associated with the elongation flow into the membrane

pores, resulting in an increase in the sieving coefficient. However, this analysis neglected the

effects of solute-solute intermolecular interactions, which could become significant at high

filtrate flux due to the accumulation of retained solutes near the upstream surface of the

membrane (concentration polarization).

1.6 Combined Reaction and Separation Processes

One of the challenges in producing a pegylated protein is generating a high yield of

the desired conjugate; for example, the maximum yield of a mono-pegylated protein reported

by several studies was only slightly greater than 50% (Gao et al., 2009; Piquet et al., 2002).

Attempts to drive the reaction forward, e.g., by the use of higher concentrations of the

activated PEG, led to the formation of multiply-pegylated products that had to be removed in

a subsequent purification step.

Page 38: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

20

Chavez and Orpiszewski (2004) used a sequential reaction – separation process to

increase the yield of a mono-pegylated lysozyme. This involved a batch pegylation process

with low conversion followed by recovery of the mono-pegylated product and recycle of the

unreacted (native) protein to a subsequent batch pegylation reaction. Although this process

did provide a slight increase in yield, there was significant product loss during the separation,

and the sequential reaction – separation process would be difficult to implement in a

commercial process. In addition, regulatory practices require traceability of the end product,

with the general philosophy that all of the product has been through an identical process (Fee

and van Alstine, 2006). Thus the use of a sequential batch recycle might create regulatory

concerns.

Fee (2003) designed a novel method for simultaneous pegylation and separation

called Size Exclusion Reaction Chromatography (SERC). The pegylation reaction was

performed in a size exclusion chromatography column, with the pegylated product selectively

removed from the moving reaction zone due to the increase in migration velocity associated

with the increased size of the protein conjugate. However, it was difficult to control the

extent of pegylation, and the low capacity of size exclusion chromatography columns would

make this approach difficult to implement at industrial scale.

More recently, Milunović et al. (2012) produced a pegylated tumor necrosis factor

(TNF-α) using immobilized metal affinity chromatography (IMAC), in which the TNF-α was

bound to the IMAC resin via histidine residues. This restricted access of the large PEG to

potential pegylation sites, reducing the formation of undesired conjugates. However, the

final yield of mono-pegylated TNF- α was still only 43%.

Page 39: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

21

1.7 Thesis Summary

1.7.1 Overall Objectives

Although recent studies have examined the potential of using ultrafiltration for

purification of pegylated proteins, there are still a number of critical issues that have yet to be

examined. For example, none of the previous studies have examined the ability of

ultrafiltration to selectively separate the differently pegylated species. There is also no

fundamental understanding of the effect of the attached PEG on the nature or magnitude of

the electrostatic interactions between the protein and membrane or on the intermolecular

interactions in the highly concentrated solutions often encountered in ultrafiltration

processes.

The overall objective of this thesis was to examine the use of ultrafiltration for the

production and purification of a desired mono-pegylated protein product. The specific aims

include: (1) evaluate the effects of electrostatic and solute-solute intermolecular interactions

on protein transmission through both neutral and charged ultrafiltration membranes over a

range of conditions, including development of an appropriate theoretical framework to

calculate the magnitude of these phenomena, (2) develop an ultrafiltration process to purify

pegylated proteins with different degree of pegylation, (3) develop a combined reaction and

membrane-based separation process to enhance the yield of a desired pegylated product.

Page 40: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

22

1.7.1 Thesis Outline

Chapter 2 provides the general theoretical background used to analyze the

ultrafiltration results. This includes theoretical models for solvent and solute transport,

including contributions from both bulk and membrane transport. A brief discussion of the

theoretical analysis of the net charge and hydrodynamic radius of pegylated proteins is also

provided.

Chapter 3 presents the basic experimental systems, materials, and methods used

throughout the thesis.

Chapter 4 examines the effects of electrostatic interactions on the transport of

pegylated proteins through negatively charged ultrafiltration membranes, with a specific

focus on the effects of the attachment of the PEG chain(s) on the steric and electrostatic

interactions.

Chapter 5 shows results for the separation of both unreacted native protein and PEG

from pegylated α-lactalbumin using a single negatively charged membrane.

Chapter 6 examines the purification of mono-pegylated α-lactalbumin from multiply

pegylated species using a negatively charged membrane.

Chapter 7 presents results for the effects of the PEG concentration on protein

transmission through both unmodified and charged ultrafiltration membranes including the

development of an appropriate theoretical framework to optimize the performance of

membrane processes for purification pegylated products.

Chapter 8 discusses the development of a combined reaction and membrane-based

separation process for enhanced yield of a desired protein-polymer conjugate. The results

demonstrate the feasibility of this combined reaction-separation system and provide a

Page 41: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

23

framework for the design of novel processes for the production of desired protein conjugates

with enhanced yield.

Chapter 9 presents a summary of final conclusions and implications of the results as

well as recommendations for appropriate future research in this area.

Page 42: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

24

Chapter 2

Theoretical Background

2.1 Introduction

This chapter provides a brief review of the theoretical models that have been

developed to describe protein transport in ultrafiltration. Briefly, the overall rate of protein

transport through a semipermeable ultrafiltration membrane is governed by i) the rate of

transport from the bulk feed solution to the membrane surface (Section 2.2), and ii) the rate

of transport through the membrane pore (discussed in Sections 2.3 and 2.4). Much of the

discussion presented in this Chapter is based on the reviews of protein transport presented by

Zeman and Zydney (1996), Molek (2008), Rohani (2011), and Rao (2006). A brief discussion

on the calculation of the net electrical charge and hydrodynamic radius for proteins and their

pegylated forms is provided at the end of the Chapter.

2.2 Bulk Mass Transport

Ultrafiltration is a pressure-driven process where both solvent and solutes are

convectively transported towards and then through a semipermeable membrane. When the

membrane is partially or completely retentive to a given solute, there is an accumulation of

the retained solute near the upstream surface of the membrane. This phenomenon is generally

referred to as concentration polarization and is shown schematically in Figure 2.1.

Concentration polarization causes the solute concentration to vary from a value of Cb in the

bulk feed to a much higher value of Cw adjacent to the membrane surface; this variation

Page 43: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

25

occurs over the distance defined as the concentration polarization boundary layer thickness,

�. At steady-state, the net rate of solute transport towards the membrane surface is balanced

by the convective flow through the membrane and the diffusive flux back into the bulk

solution; the mathematical description of the concentration profile is discussed subsequently.

Figure 2.1 Schematic representation of concentration polarization of a solute near the

membrane surface with the concentration polarization boundary layer thickness �.

The accumulation of solute at the membrane surface can reduce the filtration rate

(filtrate flux) by three mechanisms (Zeman and Zydney, 1996). First, the presence of a high

concentration of retained solute at the upstream surface of the membrane causes an osmotic

pressure difference between the two sides of the membrane, resulting in a reduction of the

effective transmembrane pressure driving force for filtration. This effect is more significant

Page 44: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

26

1)

for small solutes, which tend to have large osmotic pressures, for example, retained salts

during reverse-osmosis. The osmotic effects can also be significant for proteins at very high

concentrations due to the effect of protein-protein interactions on the thermodynamic

behavior; for example, the osmotic pressure for bovine serum albumin in a 150 mM NaCl

solution at pH 7 increases from 2.9 to 220 mmHg as the protein concentration increases from

10 to 200 g/L (Vilker et al., 1981). Second, the accumulated solute can form a dense cake

(gel layer) which provides an additional hydraulic resistance to flow (Bowen and Jenner,

1995). Third, the high solute concentration at the surface of the membrane can lead to

irreversible membrane fouling both on and within the membrane pores, decreasing the

membrane hydraulic permeability.

2.2.1 Stagnant Film Model

The most commonly used model to describe concentration polarization effects in

membrane systems is the stagnant film model. The model provides an approximate analysis

of the concentration profile within a stagnant layer upstream of the membrane surface,

neglecting the complexities associated with the detailed fluid flow within the membrane

device and the coupling between mass and momentum transport. The model assumes that

solute-solute intermolecular interactions are negligible, and that the solute diffusivity and

fluid viscosity are both independent of the solute concentration. At steady state, the solute

flux through the membrane and into the filtrate solution (-JvCf) is equal to the net solute flux

towards the upstream surface of the membrane:

dz

dCDCJCJ vfv −−=− (2.1)

Page 45: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

27

2)

3)

where Jv is the filtrate flux, Cf is the concentration of solute in the filtrate, C is the local

solute concentration at a position z above the membrane surface, and D is the solute

diffusivity. The first term on the right hand side represents the convective solute transport to

the membrane surface while the second term represents the diffusion away from the

membrane surface where the solute is more concentrated (Fick’s Law). Equation (2.1) can be

integrated over the concentration boundary layer thickness (�) with C = Cw at z = 0 and C =

Cb at z = � to give:

−=

fb

fw

vCC

CCDJ ln

δ (2.2)

The ratio of the solute diffusion coefficient to the boundary layer thickness is typically set

equal to the solute mass transfer coefficient, km = D/δ , with the evaluation of km discussed

subsequently. More details about the derivation and validity of the stagnant film model is

provided by Zydney (1997).

Concentration polarization also increases the rate of solute transport through the

membrane by increasing the local solute concentration at the membrane surface. The solute

transport is typically described in terms of the observed sieving coefficient (So), defined as

the ratio of the solute concentration in the filtrate solution to the concentration in the bulk

feed (So =Cf/Cb). The observed sieving coefficient can be related to the actual sieving

coefficient (Sa), defined as the ratio of the solute concentration in the filtrate to the

concentration adjacent to the membrane (Sa= Cf/Cw), by rearranging Equation (2.2):

So =Sa exp(Jv / km )

(1− Sa )+ Sa exp(Jv / km ) (2.3)

Equation (2.3) has been used to successfully analyze experimental data for a variety of

macromolecules including proteins, DNA, dextrans, etc. However, at very high degrees of

Page 46: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

28

polarization, the observed sieving coefficient is often found to be a weaker function of the

filtrate flux than given by Equation (2.3); this is typically attributed to the effects of solute-

solute intermolecular interactions on bulk transport (Zydney, 1992). In particular, the solute-

solute interactions in the bulk tend to enhance the diffusive flux away from the membrane

surface, reducing Cw relative to the value given by the classical polarization model. A

modified concentration polarization model including the effects of solute-solute

intermolecular interactions was developed by Zydney (1992). This model can also be used to

analyze data for a PEG – protein system at high PEG concentration as discussed in Chapter 7.

2.2.2 Bulk Mass Transfer Coefficient

Although in principle the mass transfer coefficient can be evaluated by solving the

governing mass transfer equations for the particular device geometry of interest, it is often

difficult to develop complete solutions due to the complexities of the fluid flow and the

coupling between mass and momentum transfer. Therefore, semi-empirical correlations are

typically developed based on a combination of experimental data and simplified theoretical

analyses (Zeman and Zydney, 1996). The mass transfer coefficient in the stirred cell

geometry used in this thesis was evaluated from the empirical correlation provided by (Smith

et al., 1968):

Sh = Ac(Re)d(Sc)0.33 (2.4)

where Sh is the Sherwood number (Sh = kmr/D), Re is the Reynolds number (Re= ρωr2/η ),

and Sc is the Schmidt number (Sc =η /ρD). D is the solute diffusion coefficient in an

infinitely dilute solution, r is the radius of the stirred cell, ω is the stirring speed, ρ is the

solution density, and η is the solution viscosity.

Page 47: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

29

The solute diffusion coefficient in an infinitely dilute solution can be evaluated using

the Stokes-Einstein equation:

s

B

r

kD

πη6= (2.5)

where kB is the Boltzman’s constant (1.38 x 10-23 J/K), T is the absolute temperature, and rs is

the hydrodynamic radius of the solute.

The viscosity for a dilute protein solution is approximately equal to the viscosity of

water. However, the viscosity of an aqueous PEG solution is a stronger function of solute

concentration. The analysis of km for a solution with high PEG concentration (provided in

Chapter 7) used the following correlation for the solution viscosity in terms of the mass

fraction of the PEG (w) as given by Mei et al., (1995):

( ) 2/ln bwawo +=ηη

(2.6)

where oη is the viscosity of water and a = 35.74 and b = −88.362 for a 20 kDa PEG.

The coefficient Ac in Equation (2.4) is a weak function of the stirred cell geometry

and was taken as Ac = 0.23 with d = 0.59 based on previous results (Opong and Zydney,

1991). The diffusion coefficients of the PEG and the pegylated proteins were calculated using

the Stokes-Einstein equation with the hydrodynamic radii determined from the correlations

provided by Fee and van Alstine (2004) as discussed in section 2.6.

2.3 Membrane Transport of Solvent

The rate of solvent transport through a membrane is described in terms of the

membrane permeability (Lp):

Page 48: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

30

=P

LJ pv

η (2.7)

where η is the solution viscosity, Jv is the filtrate flux, and ∆P is the transmembrane pressure.

For a membrane with uniformly distributed cylindrical pores, the filtrate flux can be

evaluated using the Hagen-Poiseuille equation as:

m

p

v

PrJ

ηδ

ε

8

2∆= (2.8)

where rp is the pore radius, ε is the porosity of the membrane, and mδ is the membrane

thickness. Equation (2.8) is valid when end effects are negligible, i.e. mδ >> rp,, which is true

for all commercially available ultrafiltration membranes.

The rate of solvent transport through the membrane pore also depends on the

membrane surface charge and solution ionic strength due to electrokinetic effects. The

solvent flux through a charged pore is reduced compared to that through a neutral pore due to

the interactions between the fluid flow and the ions adjacent to the pore boundary. The

presence of a net charge on the pore wall causes an accumulation of counterions in the region

adjacent to the pore wall (the region typically referred to as the electrical double layer). The

convective flux through the membrane due to the applied transmembrane pressure will create

a net convective flux of counter-ions through the pore. As a result, an induced electrical

(streaming) potential is developed to generate a back ion transport that exactly balances the

convective ion flux, resulting in a steady state where there is no electrical current flow

through the pore. This phenomenon is often referred to as counter-electroosmosis. A detailed

review of the subject as well as the theoretical model describing the reduced filtrate flux are

provided elsewhere (Zeman and Zydney, 1996; Pujar, 1996).

Page 49: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

31

2.4 Membrane Transport of Solute

For a membrane with isotropic pores (pore properties independent of axial position

through the pore), the radially averaged local solute flux through the membrane sN has

contributions from both convective and diffusive transport (Deen, 1987):

dz

CdDKCVKN

s

dscs −= (2.9)

where sC is the radially averaged solute concentration inside the pore, V is the radially

averaged velocity of the fluid (solvent flux), D is the solute diffusivity in free solution, and z

is the axial position within the pore. The coefficients Kc and Kd are the hindrance factors

associated with convection and diffusion, respectively. These coefficients describe the

additional drag on the solute molecule due to the presence of the pore wall. The evaluation of

these parameters is discussed in section 2.4.2.

The radially averaged solute concentration in the pore sC can be related to the

solute concentrations immediately outside the pore (Cf and Cw) using the equilibrium

partition coefficient (φ ):

f

mzs

w

ozs

C

C

C

Cδφ == == (2.10)

Equation (2.9) can be integrated over the membrane thickness ( mδ ) to give:

1exp

exp

=

m

d

c

fm

d

cw

cs

K

K

D

V

CK

K

D

VC

VKN

δφφ

δφφ

φ (2.11)

Page 50: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

32

The solute flux through the membrane is equal to the solute flux into the filtrate

solution, fs CVN = ; substitution of this relationship into Equation (2.11) yields an

expression for the actual sieving coefficient (Sa = Cf /Cw) in terms of the asymptotic sieving

coefficient ( ∞S ) and the membrane Peclet number (Pem):

1)exp(

)exp(

−+=

m

ma

PeS

PeSS (2.12)

where

=

= ∞

D

V

K

S

D

V

K

KPe

m

d

m

d

cm

δφ

δ (2.13)

cKS φ=∞ (2.14)

d

effK

D

Dφ= (2.15)

The membrane Peclet number (Pem) describes the relative contributions of the convective and

diffusive fluxes within the membrane pore. At very high filtration velocities, solute transport

is governed by convection and the solute flux across the membrane reduces to

ws CVSN ∞= . At a very low Pem, solute transport is determined by diffusion with

)( fwds CCDKN −= φ .

Figure 2.2 shows a typical plot for the actual sieving coefficient as a function of the

membrane Peclet number (Pem) calculated using Equation (2.12) for different values of ∞S .

The actual sieving coefficient (Sa) decreases from a value of one at a very low Pem to a

constant value equal to the asymptotic sieving coefficient ( ∞S ) at very high Pem, in good

agreement with experimental observations (Zeman and Zydney, 1996). In order to determine

the actual rate of solute transport through the membrane (i.e. the actual sieving coefficient), it

Page 51: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

33

is necessary to evaluate the equilibrium partitioning coefficient (ɸ) and the hydrodynamic

parameters, Kc and Kd, as a function of the solute and membrane properties as discussed in

the following sections.

Figure 2.2 Actual sieving coefficient as a function of membrane Peclet number.

2.4.1 Thermodynamic Partition Coefficient

The equilibrium partition coefficient (ɸ) for a spherical solute in a cylindrical pore

can be expressed in terms of the energy of interaction between the solute and the pore

boundary as (Anderson and Quinn, 1974; Zydney and Pujar, 1998):

Page 52: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

34

−=

1

0

exp2 ββψ

φ dTk B

total (2.16)

where β is the dimensionless radial position (β= r/rp) and the total interaction potential is the

sum of the contributions from steric, electrostatic, and intermolecular (solute-solute) forces:

rermoleculaticelectrostasterictotal intψψψψ ++= (2.17)

2.4.1.1 Steric Interactions

The equilibrium partition coefficient for a spherical solute in a cylindrical pore in the

presence of purely hard-sphere (steric) interactions can be evaluated directly from Equation

(2.16) using ∞→stericψ when the solute overlaps the pore wall and 0=stericψ in the pore

interior giving:

2

1

0

)1(2 λββφλ

−== ∫−

d (2.18)

where λ is the ratio of the solute to pore radii (λ= rs/rp).

Equation (2.18) is only valid for a membrane with uniform cylindrical pores of a

given pore radius. Giddings et al. (1968) analyzed the partitioning behavior of a wide range

of spherical and non-spherical solute in model membranes constructed from an array of

intersecting planes. The partition coefficient was governed primarily by the quantity:

sR 2/** =λ (2.19)

where R* is the mean projected solute radius (e.g. R* = rs for a spherical solute) and s is the

specific area of the pore (the pore volume divided by the pore surface area), with the partition

coefficient given as:

Page 53: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

35

*)2exp( λφ −= (2.20)

2.4.1.2 Electrostatic Interactions

The first rigorous analytical expressions for evaluating the electrostatic potential for a

charged spherical solute in a charged cylindrical pore were developed by (Smith and Deen,

1980). The solutions were obtained by solving the linearized Poisson-Boltzmann equation

(valid at electrical potentials less than about 25 mV) using matched asymptotic expansions in

cylindrical and spherical coordinates. The results for the dimensionless electrostatic energy of

interaction for constant surface charge density can be expressed as:

denpppsspss

B

E AAAATk

/)( 22 σσσσψ

++= (2.21)

where σs and σp are the dimensionless surface charge densities of the solute (protein) and

pore respectively:

RT

qFr

ro

sp

s εεσ = (2.22)

RT

qFr

ro

pp

p εεσ = (2.23)

where oε is the permittivity of free space, εr is the dielectric constant of the solution, F is the

Faraday’s constant, R is the ideal gas constant, T is the absolute temperature, rp is the pore

radius, and qs and qp are the dimensional surface charge densities of the solute and pore,

respectively. As, Asp, Ap and Aden are all positive coefficients, which are functions of the

solution ionic strength, solute size, and pore size:

Page 54: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

36

[ ][ ] θ

θτθτ

τλπτλ τλ

dI

KeAs ∫

++

+=

0

2/122

1

2/122

1

4

)(

)(

1

4 (2.24)

[ ]

)(

)1()1(2

1

2

2

τττλτλπ τπτπ

I

eeAp

−−+=

(2.25)

)(

4

1

22

τλπ

IAsp = (2.26)

[ ] [ ][ ] θ

θτθτ

τλτλτλπτ τλτλτλ dI

KeeeAden ∫

∞−−−

++

−−+−+=0

2/122

1

2/122

1

)(

)()1()1()1( (2.27)

where I1 and K1 are modified Bessel functions, prκτ = is the dimensionless pore radius, and

κ is the inverse Debye length:

2/1

1

22

= ∑

=

N

i

ii

or

CzRT

F

εεκ (2.28)

where zi and Ci are the valence and concentration of each ion in the solution.

The three terms in Equation (2.21) represent the potential energy of interaction

associated with the distortion of the electrical double layer around the solute, direct charge-

charge interactions between the solute and the pore, and the distortion of the electrical double

layer adjacent to the pore wall, respectively. The change of interaction energy associated with

the deformation of the electrical double layer around the solute and the pore wall is always

positive, leading to a reduction in the value of the partition coefficient. The direct charge-

charge interaction term is positive when σs and σp have like charges and negative when they

are of opposite charge.

Equations (2.24) to (2.27) are valid for a solute located at the pore axis (centerline

approximation). Further analyses have extended this basic theoretical framework to account

for a solute at arbitrary radial position (Smith and Deen, 1983), for interactions at constant

Page 55: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

37

surface potential (Smith and Deen, 1980), and for the effects of charge regulation which

account for the change in surface charge / potential of the protein and the pore associated

with the distortion of the local electric field when the protein enters the pore (Pujar and

Zydney, 1997).

2.4.1.3 Solute-Solute Intermolecular Interactions

The effects of solute-solute interactions on sieving are not generally observed during

protein ultrafiltration unless the feed concentration approaches 5% by volume or

approximately 50 g/L for a typical protein solution (Zeman and Zydney, 1996). However,

these effects can be significant at lower solute concentrations for long polymer chains and in

multi-component mixtures with non-ideal solution behavior. For example, the presence of

BSA can significantly decrease the sieving coefficient for lysozyme due to attractive

electrostatic interactions between the oppositely charged proteins in the bulk solution

(Ingham et al., 1980). Similar behavior was observed for BSA in the presence of IgG (Baker

and Strathmann, 1970).

The effects of solute-solute interactions on the partition coefficient can be evaluated

theoretically using Equation (2.16) by accounting for the change of the solute chemical

potential between the external solution adjacent to the membrane and the solution space

inside the membrane pores, i.e. by treating the solution and pore space as two distinct phases:

)exp(Tk

G

B

∆−=φ (2.29)

where ∆G is the change in the solute chemical potential. Equation (2.29) indicates that the

partition coefficient associated with solute-solute interactions could be greater than or less

Page 56: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

38

than one depending upon whether the free energy for the solute in the pore is less than or

greater than that in the external solution. More details on the application of this theoretical

framework to evaluate the sieving coefficients in PEG – protein mixtures are provided in

Chapter 7.

2.4.2 Hydrodynamic Analyses

The hindrance factors for convection (Kc) and diffusion (Kd) arise from

hydrodynamic interactions between the solute and the pore boundary. Expressions for Kc and

Kd were originally developed using the centerline approximation, assuming that the spherical

solute is located at the axis (centerline) of a cylindrical pore yielding (Bungay and Brenner,

1973; Deen 1987):

]2[ φ−= GKc (2.30)

1−= KK d (2.31)

where ϕ is the equilibrium partition coefficient discussed previously. G is the lag coefficient,

equal to the velocity of the particle relative to the unperturbed velocity evaluated at the

particle center, and K is the enhanced drag coefficient, equal to the drag in the pore

normalized by that in an unbounded fluid. Bungay and Brenner (1973) developed analytical

expressions for G and K using matched asymptotic expressions with the results expressed as:

t

s

K

KG

2= (2.32)

tK

Kπ61 =− (2.33)

where Kt and Ks are given as:

Page 57: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

39

∑∑=

=

− +

−+−=

7

3

32

1

2/52 )1(1)1(24

9

n

n

n

n

n

nt aaK λλλπ (2.34)

∑∑=

=

− +

−+−=

7

3

32

1

2/52 )1(1)1(24

9

n

n

n

n

n

ns bbK λλλπ (2.35)

with the expansion coefficients (an and bn) provided in Table 2.1.

It is important to note that Equations (2.30) to (2.35) were developed by neglecting

electrostatic interactions between the pore and the solute. More detailed analyses for Kc and

Kd, including the effects of electrostatic interactions on these coefficients, are provided by

Dechadilok and Deen (2009a; 2009b). The results suggest that the electrostatic effects are of

secondary importance.

Page 58: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

40

Table 2.1 Expansion coefficients for hydrodynamic functions Kt and Ks

Subscript (n) an bn

1 -73/60 7/60

2 77,293/50,400 -2,227/50,400

3 -22.5083 4.0180

4 -5.6117 -3.9788

5 -0.3363 -1.9215

6 -1.216 4.392

7 1.647 5.006

2.5 Pore Size Distribution: Effect on Membrane Transport

The analyses presented in the previous sections are limited to membranes with a

uniform (single) pore radius. In order to apply these expressions to actual ultrafiltration

membranes, it is often necessary to include the effects of the pore size distribution. Most

previous studies have employed a log-normal pore size distribution (Mochizuki and Zydney,

1993; Saksena and Zydney, 1995), which is conveniently expressed as:

+

−=2

2ln

2

1exp

2)(

b

r

r

bbr

nrn o

π (2.36)

with the parameter b given by

+=2

1lnr

(2.37)

where r is the pore radius, r is the mean pore radius, no is the number of pores at the

maximum of the distribution function, and σ is the standard deviation of the distribution

Page 59: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

41

The average solute flux (sN ) and the average solvent flux (V ) through a membrane

with a pore size distribution are evaluated by integration of sN and V over the pore size

distribution (Mochizuki and Zydney, 1993):

∫∞

=

0

2

0

2

)(

)(

drrrn

drrrnN

N

s

s

π

π (2.38)

∫∞

=

0

2

0

2

)(

)(

drrrn

drrrnV

V

π

π (2.39)

where sN and V are the solute and solvent flux in a pore of radius rp.

The average asymptotic sieving coefficient though a membrane ( ∞S ) is defined

experimentally as wCVSN ∞= , which can be evaluated using Equation (2.38) and (2.39) as:

∫∞

∞ =

0

4

0

4

)(

)(

drrrn

drrrnS

S

π

π (2.40)

Similarly, the average effective diffusion coefficient ( dKφ ) is defined experimentally as

( )fw

m

ds CC

DKN −=

δφ

giving:

∫∞

=

0

4

0

4

)(

)(

drrrn

drrrnK

K

d

d

π

πφφ (2.41)

Page 60: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

42

The r4 dependent in the numerator in Equations (2.40) and (2.41) arises from the r2

dependence of the circular cross sectional area of the cylindrical pores in combination with r2

dependence of the filtration velocity.

The average value of the actual sieving coefficient ( aS ) cannot be evaluated

explicitly. However, it can be determined iteratively for a given solute radius using Equation

(2.11) with the values of ∞S and dKφ determined by numerical integration over the pore size

distribution using Equation (2.40) and (2.41). More details on the theoretical analysis of pore

size distribution effects on membrane transport are provided by Mochizuki and Zydney

(1993).

2.6 Effective Protein Radius

The effective radii (b) of the pegylated proteins examined in this thesis were

determined from the correlations provided by Fee and Van Alstine (2004). These

correlations were developed based on the retention volume in size exclusion chromatography

(SEC), which was used to evaluate the viscosity radius of an equivalent sphere for a variety

of proteins (α-lactalbumin, β-lactoglobulin, and bovine serum albumin) pegylated with linear

PEGs of different molecular weight (2, 5, and 20 kDa). Their results are expressed as:

b =A

6+

2

3ARPEG

2 +RPEG

3 (2.42)

with

A = 108a3 + 8RPEG

3 + 12 81a6 + 12a3RPEG

3( )1

2

13

(2.43)

Page 61: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

43

where RPEG and a are the radii of the isolated PEG and protein, respectively. The radius of a

free PEG molecule was calculated as:

RPEG = 0.01912 × MW 0.559 (2.44)

where RPEG is in nm and the molecular weight (MW) is in Da (Fee and Van Alstine, 2004).

The radius of the unmodified α-lactalbumin is a = 2.0 nm.

2.7 Protein Net Charge

The net electrical charge on a protein is determined by the dissociation of the

ionizable amino acid residues, which is directly related to the intrinsic dissociation constant

( i

apK ) of that ionizable group. For example, the dissociation / protonation of an α-carboxylic

acid, RCOOH, is given as:

][

]][[

RCOOH

HRCOOK i

a

+−

= (2.45)

where the square brackets refer to the molar concentration for that species. Equation (2.45) is

typically written in the form of the classic Henderson-Hasselbach equation:

−+=

ii

ii

arn

rpKpH log (2.46)

where ni is the total number of each ionizable amino acid and ri is the number of

unprotonated residues. The pH in Equation (2.46) refers to the H+ concentration at the protein

surface, which is different from that in the bulk solution due to electrostatic interactions

between the protein and the hydrogen ion. The local pH is evaluated assuming a Boltzmann

distribution:

Page 62: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

44

−= ++

Tk

eHH

B

sb

ψexp][][ (2.47)

where [Hb+] is the bulk hydrogen ion concentration, e is the electron charge, and ψs is the

electrical potential at the protein surface. ψs can be directly related to the net protein charge

(Z) assuming that the protein is a hard sphere with uniform surface charge:

)1(4 ssro

srr

eZ

κεπεψ

+= (2.48)

where κ is the inverse Debye length calculated using Equation (2.28). The protein net charge

is determined from the difference between the maximum number of positively charged

residues on the protein ( +maxZ ), equivalent to the net protein charge at very low pH (where all

residues are protonated), and the number of unprotonated groups:

∑=

+ −=n

i

irZZ1

max (2.49)

Equations (2.46) to (2.49) can be solved iteratively to evaluate the net protein charge as a

function of bulk pH and ionic strength. Note that the calculated values of the net charge are

determined assuming that each type of amino acid residue has the same pKa, which ignores

the detailed interactions and local ionic distribution around a given amino acid on the protein

surface. More sophisticated models for calculating the net protein charge are available, but

they require very detailed information about the protein structure (Sharma et al., 2003). A

more detailed discussion of the theoretical evaluation of the protein charge using this

theoretical framework is provided by Menon and Zydney (2000).

Table 2.2 shows the number of each ionizable amino acid present in α-lactalbumin

(Brew et al., 1970) and its corresponding pKa (Nelson and Cox, 2008). Typical results for the

calculated values of the net charge of α-lactalbumin in 1, 10, and 100 mM ionic strength

Page 63: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

45

solutions are shown in Figure 2.2 as a function of solution pH. The absolute value of the

charge increases with increasing ionic strength due to the shielding of the electrostatic

interactions between the charged protein and the H+ ions in the higher salt concentration

solution.

Table 2.2 Number (ni) and i

apK values of the charged amino acids in α-lactalbumin (Brew

et al., 1970; Nelson and Cox, 2008)

Type Number i

apK

C terminal 1 2.36

N terminal 1 9.67

Asp (D) 9 3.65

Glu (E) 8 4.25

His (H) 3 6.00

Lys (K) 12 10.53

Tyr (Y) 4 10.07

Arg (R) 0 12.48

Page 64: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

46

Figure 2.3 Calculated net charge of α-lactalbumin as a function of the solution pH at

different ionic strength.

Page 65: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

47

Chapter 3

Materials and Methods

3.1 Introduction

This chapter provides information about the materials, methods, and apparatus

employed for experiments commonly used throughout this dissertation. Details about

additional specific experiments / procedures are provided in individual chapters as

appropriate.

3.2 Experimental Materials

3.2.1 Polyethylene Glycol (PEG)

Polyethylene glycol (PEG) is an amphiphilic molecule composed of repeating units

of ethylene oxide with terminal hydroxyl group(s) as shown in Figure 3.1. PEGs are usually

synthesized by ring opening polymerization of ethylene oxide (Pasut and Veronese, 2007);

they are commercial available with different lengths and branching. The polymers with a

molecular weight up to 100,000 kDa are typically called PEGs while those with higher

molecular weight are classified as polyethylene oxide (PEO). Linear PEG molecules can be

divided into two subgroups: PEG with free hydroxyl groups at both ends and PEGs with one

or two methoxylated end group(s) where the –OH group is replaced by –OCH3 (Hamidi et al.,

2006). The hydroxyl group(s) can be chemically modified by reaction with specific

functional groups, e.g., for producing an activated PEG.

Page 66: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

48

Figure 3.1 Molecular structure of linear polyethylene glycol.

PEG is generally considered to be a non-toxic molecule and has been used for many

years as a formulation excipient in the pharmaceutical and cosmetic industries (Working et

al., 1997). PEG with molecular weight > 400 Da is approved by FDA for intravenous

application; however, PEG chains shorter than 400 Da can be metabolized in vivo by alcohol

dehydrogenase, which generates toxic metabolites in humans (Knop et al., 2010). PEG

possesses unique properties such as high solubility in water, low immunogenicity, and

minimal toxicity. PEGs are commercially available with low polydispersity, with the ratio of

Mw/Mn (where Mw and Mn are the weight and number average molecular weights,

respectively) ranging from 1.01 for PEG smaller than 5 kDa to 1.1 for 50 kDa PEG (Pasut

and Veronese, 2007). In general, a polydispersity less than 1.1 is used for pharmaceutical

applications to ensure a high degree of reproducibility with respect to immunogenicity and in

vivo residence time (Knop et al., 2010).

One disadvantage of PEG is that it is non-biodegradable. PEG with molecular weight

below 20 kDa can be secreted in the urine via renal clearance; however, higher molecular

weight PEG is retained by the glomerulus and accumulation / clearance in the liver becomes

predominant (Knop et al., 2010; Jevsevar et al., 2010). The fate of high molecular weight

Page 67: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

49

PEG at the cellular level is not fully understood; there are no systematic studies examining

the role of PEG at the site of accumulation. PEGs in this study were obtained from Sigma

Aldrich (St Louis, MO) and Creative PEGwork (Winston Salem, NC) with nominal

molecular weights of 1.5 and 20 kDa, respectively.

3.2.2 Activated Polyethylene Glycol

Several chemistries have been developed to covalently attach PEG to different amino

acids including both primary and N-terminal lysines (amino groups), cysteines (thiol groups),

glutamines (amide groups), and serine and threonine (hydroxyl groups) (Veronese and Pasut,

2005). Each method requires different functionalization of the activated group on the PEG

molecule. The most common covalent attachment site of PEG to a protein is via the primary

amine of a lysine group due to the ease of attachment and the mild reaction conditions

(Gaberc-Porekar et al., 2008). The reaction is usually performed with PEG containing an N-

hydroxylsuccinimide (NHS) ester as a leaving group (Jevsevar et al., 2010; Fee, 2003).

In this study, pegylated proteins were produced using N-hydroxysuccinimide

activated PEG with nominal molecular weight of 20 kDa and polydispersity <1.08 as

provided by the manufacturer (Catalog number ME-200HS; NOF Corporation, Tokyo,

Japan). Smaller activated PEGs were also used to study the effects of electrostatic

interactions during ultrafiltration, with details provided in Chapter 5.

Page 68: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

50

3.2.3 Proteins

α-lactalbumin with a molecular weight of 14.2 kDa was used as a model protein due

to its availability, high solubility over a wide range of pH and ionic strength, and well defined

properties. In addition, the protein has a very similar size to that of several proteins used in

commercially available pegylated systems, e.g., interferon (MW = 19.3 kDa) and human

granulocyte stimulating factor (18.8 kDa). α-lactalbumin is a small, acidic protein found in

milk of many mammalian species. It is a regulatory protein of lactose synthase, which

catalyzes lactose synthesis in the lactating mammary gland. The protein also strongly binds to

Ca2+, which is required for protein folding and native disulfide bond formation (Chrysina et

al. 2000; Permyakov and Berliner, 2000).

In this study, bovine α-lactalbumin was obtained from Sigma Chemicals (Catalog

Number L5385, MW = 14.2 kDa). Figure 3.2 shows the X-ray structure for α-lactalbumin

(Permyakov and Berliner, 2000). Basic properties of the protein are shown in Table 3.1. The

number of lysine groups was obtained from the published amino acid sequence of the protein

(Viaene et al., 1991); the hydrodynamic radius was provided by Smith (1970). The diffusion

coefficient in an infinitely dilute solution (D) was calculated using the Stokes-Einstein

equation based on the hydrodynamic radius.

Protein solutions were prepared by weighing an appropriate amount of dry protein

using a digital balance Model AG104 obtained from METTLER TOLEDO (Columbus, OH)

and dissolving it into an appropriate buffer. The protein solutions were then filtered through

an Acrodisc® syringe filter with Supor® membrane with 0.2 µm pore size (Pall Corporation,

Ann Arbor, MI) to remove any large aggregates. When not in use, protein solutions were

stored at 4 oC.

Page 69: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

51

Table 3.1 Basic physical / chemical properties of 20 PEG, native, and 20 kDa pegylated α-

lactalbumins.

Properties PEG α-lac PEG1 PEG2 PEG3

Molecular weight (kDa) 21.4 14.2 35.5 56.8 78.1

Lysine groups 0 12 11 10 9

Hydrodynamic radius (nm) 51 2.0 52 74 93

D (x10-10 m/s2) 0.43 1.1 0.42 0.29 0.23

Figure 3.2 X-ray structure of bovine α-lactalbumin including ion binding sites (for Ca2+ and

Zn2+) adapted from Permyakov and Berliner (2000). Disulfide bridges are shown

in yellow.

Page 70: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

52

3.2.3.1 Pegylated Proteins

The pegylated α-lactalbumins were prepared by reaction of the native protein with

the 20 kDa N-hydroxylsuccinimide (NHS) ester activated PEG as shown in Figure 3.3. The

α-lactalbumin was dissolved in a 10 mM Bis-Tris buffer at pH 7 (unless otherwise stated).

The activated PEG was then added and the solution was slowly stirred at room temperature

(21-24oC) for a minimum of 8 hr to allow the reaction to go essentially to completion. The

resulting product solution, which contained the pegylated α-lactalbumin, the un-reacted

protein, the hydrolyzed PEG reagent, and N-hydroxysuccinimide (produced from hydrolysis

of the activated PEG), was then diluted approximately four-fold with Bis-Tris buffer. The

solution was pre-filtered through a 0.2 µm pore size Acrodisc syringe filter (Pall Corporation,

Ann Arbor, MI) to remove any particulate matter and larger aggregates prior to use. The

solution ionic strength was adjusted to the desired value by addition of 1 M KCl or NaCl; in

some cases the solution was diafiltered through a 10 kDa Ultracel membrane using an

appropriate diafiltration buffer. Pegylated proteins were stored at 4oC when not in use.

The properties of the α-lactalbumin pegylated with different numbers of 20 kDa

PEGs are shown in Table 3.1 where PEG1, PEG2, and PEG3 represent the mono-, di, and tri-

pegylated α-lactalbumin, respectively. The number of available lysine groups was reduced by

the pegylation reaction, with the isoelectric point of the pegylated proteins calculated from

the available amino acid sequence based on the pKa values of the amino acids as discussed

previously in Chapter 2. The hydrodynamic radii for the pegylated proteins were calculated

using the correlations provided by Fee and van Alstine (2004) as discussed in Chapter 2.

Page 71: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

53

Figure 3.3 Pegylation reaction between a PEG bearing an N-hydroxylsuccinimide ester

(PEG-NHS) and a primary amine on a protein (e.g., a lysine group).

3.2.3.2 Acetylated proteins

An acetylated protein is a modified protein in which the free amine group on one or

more lysine amino acids was reacted with acetic anhydride. Acetylated α-lactalbumin was

synthesized using the general procedure described by Gao and Whitesides (1997). α-

lactalbumin was added to deionized water to a concentration of 2 g/L. The solution was

chilled in an ice bath to 5oC and the pH was adjusted to 12 by addition of 0.1 M NaOH. Four

molar equivalents of acetic anhydride (as a 9.75 g/L solution in dioxane) were then added to

the protein solution. The reaction mixture was stirred continuously for 15 min while slowly

adding 0.1 M NaOH to maintain pH ≈ 12. The reaction was then quenched by adding 0.5 M

HCl to rapidly reduce the pH to approximately 6. A constant volume diafiltration was

performed through a 10 kDa UltracelTM membrane (Millipore Corp., Bedford, MA) using 10

mM Bis-Tris buffer for a minimum of five diavolumes to remove acetic acid, unreacted

acetic anhydride, and other small impurities. The resulting solution was then filtered through

a 0.2 µm Pall Supor Acrodisc syringe filter prior to use.

Page 72: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

54

Figure 3.4 Acetylation reaction between an acetic anhydride and a primary amine on a

protein.

3.2.4 Ultrafiltration Membranes

Ultrafiltration experiments were performed using UltracelTM composite regenerated

cellulose membranes obtained from EMD Millipore (Bedford, MA) with nominal molecular

weight cut-off of 10, 30, 100, and 300 kDa. UltracelTM membranes are asymmetric (Figure

3.5) with a thin skin layer (approximately 1 µm thick) that provides the desired separation

selectivity and a highly permeable substrate that provides the mechanical strength (Zeman

and Zydney, 1996). The effective pore size for the UltracelTM membranes was estimated from

the hydraulic permeability using Equation (2.8), with results provided in Table 3.2.

Table 3.2 Approximated effective pore size for UltracelTM membranes

MW cut-off (kDa) 10 30 100 300

Pore radius (nm) 2.4 3.2 6.4 8.5

Page 73: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

55

Figure 3.5 SEM image of an UltracelTM membrane cross-section provided by the

manufacturer.

Membrane disks were cut off from a large flat sheet using a special cutting tool,

soaked in 90% isopropanol for 40 min to remove any storage agents and fully wet the pore

structure, and flushed with 100 L/m2 deionized water. Negatively-charged versions of the

UltracelTM membranes were produced by chemical modification of the base cellulose by

attachment of sulfonic acid groups to the free hydroxyls using the base activated chemistry

provided by van Reis (2006). Membranes were first soaked in 0.1 M NaOH for 12 hr and

then immersed in a 2 M solution of 3-Bromopropanesulfonic acid sodium salt (Catalogue

#B2912, Sigma Chemical) in 0.1 N NaOH for a specified period of time. The membrane was

then thoroughly washed with 0.1 M NaOH followed by DI water, 0.2 M acetic acid, DI water

again, and finally the buffer solution that was to be used in the ultrafiltration experiment.

Page 74: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

56

Figure 3.6 Schematic of the negative charge modification of an UltracelTM membrane by

attachment of sulfonic acid groups. Adapted from Molek (2008)

3.2.5 Buffer Solutions

Appropriate buffers were used to maintain the desired solution pH. Buffer solutions

were prepared by dissolving pre-weighed amounts of the required salts in deionized water

obtained from a NANOpure Diamond water purification system (Barnstead Thermolyne

Corporation, Dubuque, IA). All salts were certified ACS grade or higher and obtained from

Sigma Aldrich (St. Louis. MO) unless otherwise stated. The solution pH was measured using

a Thermo Orion pH meter Model 402 (Beverly, MA) and adjusted within 0.05 pH unit of the

desired value by addition of 0.1 M HCl or NaOH as required. 4 M HCl or NaOH were

occasionally used when adjusting pH of solutions with very high buffer concentrations.

Page 75: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

57

Buffer solutions were then filtered through a 47 mm Supor® membrane with 0.2 µm pore size

(Pall Corporation, Ann Arbor, MI) with a vacuum pump to remove any particulates before

use. The solution ionic strength (I) was evaluated as:

∑=i

iiCzI 2

2

1 (3.1)

where zi and Ci are the net charge and total concentration for each dissolved ion.

3.3 Experimental Methods

3.3.1 Ultrafiltration Apparatus

Figure 3.7 Schematic of ultrafiltration stirred cell apparatus.

Ultrafiltration experiments were performed with Amicon Stirred cells Model 8010,

8050, and 8200 with effective membrane areas of 4.1, 13.4, and 28.7 cm2, respectively. An

ultrafiltration membrane of interest was placed at the bottom of the stirred cell on a porous

Page 76: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

58

layer of Tyvek®, which was used as a structural support to prevent deformation of the

membrane at high pressure. The stirred cell was equipped with a stirred bar, which was

suspended from the top of the cell so that it was located directly above the membrane. The

stirring speed was set to 600 rpm using a VWR 205 Autostirrer magnetic stirred plate, with

the stirring speed calibrated using a Strobotac 1531-AB phototachometer (General Radio

Company, Concord, MA). The stirred cell was air pressurized to control the filtrate flux,

which was measured by timed collection. The applied pressure was measured using an

Ashcroft 0-30 or 0-60 psig digital pressure gauge (Ashcroft, Stratford, CT). Alternatively, the

filtrate line from the stirred cell was connected to a variable-speed peristaltic pump

(Dynamax, Rainin Instrument, CA) for more accurate control of the filtrate flux. A schematic

of the ultrafiltration apparatus is shown in Figure 3.7.

A tangential flow filtration (TFF) module was also used in this thesis for the study of

a combine reaction-separation process. The details of this apparatus and the associated

operating procedures are provided in Chapter 8.

3.3.2 Membrane Hydraulic Permeability

The membrane hydraulic permeability (Lp) was evaluated by measuring the filtrate

flux (Jv) as a function of transmembrane pressure (∆P) using a 10 mM Bis-Tris buffer

containing 500 mM NaCl at pH 7. The high salt concentration was employed in order to

minimize the contribution of counter electro-osmosis. Filtrate flux data were collected for at

least 4 transmembrane pressures. The permeability was calculated from the slope of the data

using:

Page 77: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

59

∆=

ηP

LJ pv (3.2)

where η is the solution viscosity. The viscosity for a dilute protein solution was approximated

using the viscosity of water while the viscosity of an aqueous PEG solution was evaluated as

discussed in Chapter 2. The membrane permeability was evaluated immediately before and

after the protein ultrafiltration experiments to provide a measure of the extent of fouling.

3.3.3 Sieving Experiments

After evaluating the membrane hydraulic permeability, the stirred cell was rinsed

with deionized water and the membrane was flushed with at least 100 L/m2 of deionized

water. The stirred cell was then filled with the desired protein solution. Filtration was

performed at a constant filtrate flux, adjusted either by air pressurization or use of a

peristaltic pump on the filtrate line. The actual filtrate flux was evaluated by timed collection.

A minimum of 1.5 mL of filtrate was collected before each sieving measurement to remove

the dead volume beneath the membrane and eliminate any transients associated with the

change in pressure. Small samples of the filtrate and retentate solutions were collected for

subsequent analysis.

3.3.4 Diafiltration

A diafiltration process was used for buffer exchange and actual separations using the

apparatus shown in Figure 3.7, except that a buffer reservoir was attached to the feed of the

stirred cell. The solution reservoir was air-pressurized, with the filtrate flux adjusted to the

Page 78: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

60

desired value using a pressure regulator. Alternatively, the filtrate line was connected to a

peristaltic pump with the diafiltration buffer continuously drawn into the stirred cell. The

filtrate flux was evaluated at multiple time points over the course of the diafiltration, with

filtrate samples collected periodically for subsequent analysis. At the end of the diafiltration,

the stirred cell was opened and a retentate sample was obtained to verify closure of the mass

balance.

3.3.5 Membrane Charge Characterization

The membrane surface charge density was evaluated from streaming potential

measurements following the procedure described by (Burns and Zydney, 2000) using the

apparatus shown in Figure 3.8. The Ag/AgCl electrodes were prepared by reducing

appropriate lengths of pure silver wire (1 mM diameter). The silver wires were first lightly

sanded and placed in a beaker of concentrated nitric acid for 10 s. The wire was then washed

with DI water and placed in a beaker containing 1 M KCl. A DC power source was then

connected to the silver electrode and a steel wire in a separate beaker containing 1 M KCl. A

commercial Kimwipe was used as a salt bridge. A current of 20 mA was maintained for 20

minutes to reduce the silver to AgCl. Electrodes were stored in 0.5 M KCl between

experiments.

The membrane of interest was placed and sealed with O-rings in between two

Plexiglas chambers. The apparatus was slowly filled with a desired buffer taking care to

remove any entrapped air bubbles from both chambers. Ag/AgCl electrodes were screwed

tightly into the ends of the chambers. The electrodes were wired to a KEITHLEY 200

multimeter (Keithley Instruments, Inc., Cleveland, OH). A feed reservoir containing the same

Page 79: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

61

buffer solution was attached to the inlet chamber and air pressurized. The buffer flow from

the exit port was directed to drain. The system was allowed to stabilize for 1 hr before the

first measurement. The transmembrane voltage (Ez) was evaluated as a function of the

applied pressure, with the system allowed to equilibrate for 15 min between measurements.

Figure 3.8 Streaming potential apparatus for measuring membrane surface charge. Taken

from Burns and Zydney (2000) with permission.

Typical experimental data obtained using a 1 mM Bis-Tris buffer with 10 mM NaCl

at pH 7 are shown in Figure 3.9 for an unmodified 300 kDa UltracelTM membrane and a

negatively-charged version that was charged for 24 hr. The apparent zeta potential (ζ) was

evaluated from the slope of the measured streaming potential as a function of the applied

pressure using the Helmholtz-Smoluchowski equation (Hunter, 1981):

Pd

dEz

r ∆

=

εεµγ

ζ0

(3.3)

Page 80: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

62

where Ez is the measured voltage, �is the solution conductivity, �� is the permittivity of free

space, and rε is the dielectric constant of the solution. The results in Figure 3.9 gave ζ = -3.0

± 0.2 for the unmodified membrane and ζ = -11.7 ± 0.2 mV for the negatively charged

membrane. The small negative charge on the unmodified membrane is likely due to the

preferential adsorption of negative ions from the bulk electrolyte.

Figure 3.9 Streaming potential as a function of applied transmembrane pressure for an

unmodified 300 kDa UltracelTM membrane and for a negatively charged version

that was charged for 24 hr.

The surface charge density of the membrane pores was evaluated as (Burns and

Zydney, 2000):

= −

RT

FFCq p

2sinh4 1

0

ζκ (3.4)

Page 81: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

63

where C0 is the bulk ion concentration, F is Faraday’s constant, 1−κ is the Debye length, R is

the universal gas constant, and T is the temperature. The calculated surface charge density for

the membrane charged for 24 hr was qp = -2.7 mC/m2.

3.4 Assays

3.4.1 Size Exclusion Chromatography (SEC)

Size exclusion chromatography was employed to evaluated the concentrations of the

pegylated α-lactalbumin, the unreacted protein, and the PEG in the filtrate or bulk samples.

Data were obtained using a Superdex 200 G/L (GE Healthcare, Uppsala, Sweden) column

with a running buffer of 150 mM NaCl with 50 mM phosphate buffer at pH 7.0 using a flow

rate of 0.3 mL/min. Sample detection was performed using an Agilent 1100 series refractive

index detector and an Agilent 1200 series UV-Vis detector at 280 nm, with the two detectors

operated in series. The chromatography system was operated using Chemstation software

Rev B.02.01-SR2 (260) (Agilent Technologies, Santa Clara, CA).

The SEC chromatogram for a typical pegylation mixture is shown in Figure 3.10

using the UV detector (top panel) and RI detector (bottom panel). Peak areas were

determined by numerical integration. Overlapping peaks were simply split at the location of

the minimum. The PEG was invisible in the UV, allowing accurate determination of the

protein concentrations to ±0.002 g/L using the calibration curve provided in Figure 3.11. The

concentrations of PEG (CPEG) could be accurately measured to within 0.02 g/L (with baseline

resolution of the peaks) using the RI detector, assuming that the total RI response is given by

the weighted sum (Kunitani et al., 1991):

Page 82: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

64

PEG

PEG

RIprotein

protein

RIRI C

dc

dnC

dc

dnn

+

= (3.5)

where (dnRI / dc)protein and (dnRI / dc)PEG are the specific RI responses for the pure α-

lactalbumin (3.23x106 nRIU·s/(g/L)) and pure PEG (2.65x106 nRIU·s/(g/L)) determined

using the calibration curves shown in Figure 3.12.

Page 83: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

65

Figure 3.10 Size exclusion chromatograms for a pegylation mixture performed with a

Superdex 200, 10/300 column using UV detector (top panel) and RI detector

(bottom panel).

Page 84: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

66

Figure 3.11 Calibration curve for α-lactalbumin using UV detection at 280 nm. The slope

corresponds to the specific UV response for α-lactalbumin of 3.36 x 104

mAU·s/(g/L).

Page 85: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

67

Figure 3.12 Calibration curves for α-lactalbumin and 20 kDa PEG using RI detection. The

slopes correspond to the specific RI response of 3.23 x 104 mAU·s/(g/L) for α-

lactalbumin and 2.65 x 106 nRIU·s/(g/L) for PEG.

3.4.2 Capillary Electrophoresis (CE)

The electrophoretic mobilities of the pegylated and acetylated α-lactalbumin were

determined using an Agilent G1600A high performance capillary electrophoresis system

equipped with a dual-polarity variable high voltage DC supply (0-30 kV) and a diode array

UV / visible absorbance detector (214 nm wavelength for detection). Experiments were

performed with negatively charged fused-silica capillaries (inner diameter of 50 µm) with a

total length of 80.5 cm and an effective length (from injection to the detection window) of

70.2 cm. The capillary and solution reservoirs were filled with 10 mM Tris / Glycine at pH

8.1. Protein samples (approximately 40 - 140 nL) containing 5 mM of mesityl oxide as a

Page 86: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

68

neutral marker were injected for 4 to 25 s at a pressure of 4 kPa. Data were obtained at a

constant applied voltage of 25 kV, with the field direction chosen so that the bulk

electroosmotic flow was toward the detector (and cathode). The current was kept less than

45 µA to minimize Joule heating.

The electrophoretic mobility was calculated from the migration times of the protein

and neutral marker as:

−=

oz

d

ettE

L 11µ (3.6)

where Ld is the effective capillary length, Ez is the applied electric field, and t and to are the

migration times for the protein and neutral marker to reach the detector.

Page 87: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

69

Chapter 4

Effect of Electrostatic Interactions on Transmission of Pegylated

Proteins through Charged Ultrafiltration Membranes

Note: The results in this Chapter were adapted from Molek, J.R., Ruanjaikaen K., Zydney

A.L., 2010. Effect of electrostatic interactions on transmission of PEGylated protein

through charged ultrafiltration membranes. Journal of Membrane Science 353, 60-69.

4.1 Introduction

The importance of electrostatic interactions in protein ultrafiltration has been well-

established over the past decade. Pujar and Zydney (1994) showed that a reduction in

solution ionic strength caused a 100-fold decrease in albumin transmission through a

negatively-charged ultrafiltration membrane. Similar effects have been seen with a number

of other proteins over a range of solution pH (Burns and Zydney, 1999) and membrane

surface charge density (Mehta and Zydney, 2006).

Several studies, e.g., Mehta and Zydney (2006) and Burns and Zydney (2001), have

shown that the magnitude of these electrostatic interactions can be well described using the

theoretical analysis developed by Smith and Deen (1980) for the partitioning of a charged

sphere in an infinitely long charged cylindrical pore:

( )

−−=

TkKS

B

Eca

ψλ exp1

2 (4.1)

where Sa is the actual sieving coefficient, defined as the ratio of the protein concentration in

the filtrate solution to that in the solution immediately upstream of the membrane. Equation

Page 88: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

70

(4.1) is only valid at high filtration velocities where solute diffusion across the membrane is

negligible relative to the convective flux. The development of the expression for the actual

sieving coefficient is discussed in more detail in Chapter 2. The term (1-λ)2 describes the

steric (hard-sphere) exclusion of the sphere from the region within one solute radius of the

pore wall (with λ equal to the ratio of the solute radius to the pore radius). Kc is the hindrance

factor associated with convection and

TkB

Eψ is the dimensionless electrostatic energy of

interaction:

denpppsspss

B

E AAAATk

/)( 22 σσσσψ

++= (4.2)

where As, Asp, and Ap and Aden are functions of the solution ionic strength, solute size, and

pore size, with their expressions provided in Chapter 2. σs and σp are the dimensionless

surface charge densities of the solute (protein) and pore. The three terms in Equation (4.2)

represent the energy of interaction associated with the distortion of the electrical double layer

around the solute, direct charge-charge interactions between the solute and the pore, and the

distortion of the electrical double layer adjacent to the pore wall, respectively.

Equations (4.1) and (4.2) were developed for hard sphere solutes in which the charge

is uniformly distributed over the external surface of the sphere. There are currently no

experimental or theoretical results for the possible effect of the attached polyethylene glycol

in a pegylated protein on the nature of these electrostatic interactions. The objective of this

study was to obtain quantitative data for the effect of solution ionic strength and membrane

surface charge on the transmission of pegylated proteins during ultrafiltration and to develop

a more fundamental understanding of how the PEG layer alters the intermolecular

electrostatic repulsion between the pegylated protein and the membrane pore. Data were

Page 89: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

71

obtained with un-modified and negatively charged membranes with pegylated α-lactalbumin

over a range of ionic strength, with the charge characteristics of the proteins also studied

using capillary electrophoresis.

4.2 Materials and Methods

4.2.1 Pegylated Protein Preparation

Experiments were performed with pegylated α-lactalbumin having one or more 2, 5,

10, or 20 kDa PEG branches. The pegylated α-lactalbumin was prepared by reaction of the

native protein (obtained from Sigma Chemicals, Catalog Number L5385, MW = 14.2 kDa)

with an N-hydroxylsuccinimide (NHS) ester activated PEG obtained from Nektar

Therapeutics (Huntsville, AL) and NOF Corporation (Tokyo, Japan). The α-lactalbumin was

dissolved in a 10 mM Bis-Tris buffer at pH 7. The activated PEG was then added, and the

solution was slowly stirred at room temperature for a minimum of 8 hr to allow the reaction

to go to completion. The resulting product solution, which contained the pegylated α-

lactalbumins, the un-reacted protein, the hydrolyzed PEG reagent, and N-

hydroxysuccinimide (produced from hydrolysis of the activated PEG), was then diluted

approximately four-fold with Bis-Tris buffer. If needed, a buffer exchange was performed

with a desired buffer through a 10 kDa UltracelTM membrane. The solution was pre-filtered

through a 0.2 µm pore size Acrodisc syringe filter (Pall Corporation, Ann Arbor, MI) to

remove any particulate matter and larger aggregates prior to use. The solution ionic strength

Page 90: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

72

was adjusted to the desired value by addition of 1 M KCl or NaCl containing the buffer

species of interest. Pegylated proteins were stored at 4oC when not in use.

4.2.2 Acetylated Protein Preparation

To obtain additional insights into the nature of the electrostatic interactions,

experiments were also performed with acetylated α-lactalbumin in which the free amine

group on one or more lysine amino acids was reacted with acetic anhydride (instead of an

activated PEG). Acetylated α-lactalbumin was synthesized using the procedure described by

Gao and Whitesides (1997)with the details discussed in Chapter 3. The resulting solution

contained a mixture of acetylated α-lactalbumin with different degree of acetylation (i.e. with

different number of capped lysines) was then filtered through a 0.2 µm Pall Supor Acrodisc

syringe filter prior to use.

4.2.3 Ultrafiltration Membranes

UltracelTM composite regenerated cellulose membranes were obtained from EMD

Millipore (Bedford, MA) with a 100 kDa nominal molecular weight cut-off. Negatively

charged versions of the UltracelTM membranes were produced by chemical modification of

the base cellulose by attachment of sulfonic acid groups using the base activated chemistry as

discussed in Chapter 3. The membrane surface charge density was evaluated from streaming

potential measurements as described in Chapter 3.

Page 91: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

73

4.2.4 Protein Characterizations

The electrophoretic mobilities of the pegylated and acetylated α-lactalbumin were

determined using an Agilent G1600A high performance capillary electrophoresis system

following the methods provided in Chapter 3.

4.2.5 Ultrafiltration Sieving Experiments

All ultrafiltration experiments were performed in a 25 mm diameter stirred

ultrafiltration cell (Amicon Model 8010, Millipore Corp., Bedford, MA) following the

general procedure provided in Chapter 3. Membranes were flushed with at least 40 L/m2 of

Bis-Tris buffer prior to exposure to protein and then soaked overnight in a solution

containing the protein solution of interest. The device was air-pressurized, with the filtrate

flux controlled by a pressure regulator (Scott Specialty gases, Plumsteadville, PA). The

pressure was measured by an Ashcroft Model 0518 (0-30 psi) or Model 8920 (0-15 psi)

pressure gauge. A minimum of 4 mL of filtrate was passed through the membrane to ensure

stable operation; this also served to flush the dead space beneath the membrane in the stirred

cell. For each experimental condition, a small filtrate sample was collected followed directly

by a small sample of the bulk solution from the stirred cell. The stirred cell was then refilled

with the pegylated protein solution and a repeat measurement obtained. The process was

repeated using solutions of different ionic strength or pH to cover the range of solution

conditions.

Size exclusion chromatography with a Superdex 200 G/L (GE Healthcare, Uppsala,

Sweden) column was used to evaluate the concentrations of the pegylated α-lactalbumins, the

Page 92: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

74

unreacted protein, and the PEG in the filtrate and bulk samples. Details regarding the column

operation and evaluation of the solute concentration are provided in Chapter 3. The

concentrations of the different acetylated α-lactalbumin species were determined by capillary

electrophoresis as described subsequently.

4.3 Results and Analysis

4.3.1 Ultrafiltration of Pegylated Proteins

Figure 4.1 shows typical data for the observed sieving coefficients of a 5 kDa PEG

(right panel) and a pegylated α-lactalbumin with a single 5 kDa PEG chain (left panel) as a

function of solution ionic strength through both an unmodified and a negatively-charged

version of the 100 kDa UltracelTM membrane. The charged membrane was produced by

reaction of the base UltracelTM membrane with 3-Bromopropanesulfonic acid sodium salt for

12 hr, giving a membrane with an apparent zeta potential of -9.2 ±0.3 mV. The data were

obtained using a mixture containing the PEG, unreacted α-lactalbumin, and the pegylated

protein, with the concentrations of each component evaluated by size exclusion

chromatography. The observed sieving coefficient was calculated from the ratio of the

filtrate to feed (bulk) concentrations, with results shown for three repeat experiments using

the same membrane. All experiments were performed at a filtrate flux of approximately 7.8

x 10-6 m/s (28 L m-2 hr-1) using Bis-Tris buffer solutions at pH 7.

The sieving coefficients of the 5 kDa PEG through the unmodified membrane (right

panel) were essentially independent of solution ionic strength, varying between 0.86 and

0.90, consistent with the absence of any significant electrostatic interactions for the neutral

Page 93: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

75

polyethylene glycol when using the unmodified cellulose membrane. The data for the 5 kDa

PEG with the charged membrane show a small reduction in sieving coefficient with

decreasing ionic strength. This is likely due to the increase in free energy associated with the

distortion of the electrical double layer within the pores of the charged membrane due to the

presence of the PEG (described qualitatively by the third term on the right-hand side of

Equation 2). The sieving coefficients of the 5 kDa PEG through the unmodified and charged

membranes at high ionic strength (200 mM) were nearly identical, suggesting that the

attachment of the small charged ligand had minimal effect on the pore size of the 100 kDa

composite regenerated cellulose membrane. This is consistent with the very similar values of

the hydraulic permeability for the unmodified and charged membranes (difference in Lp of

less than 10%).

The sieving coefficients of the pegylated α-lactalbumin (left panel) were a much

stronger function of solution ionic strength, varying from 0.4 to 0.72 for the unmodified

membrane and from less than 0.02 to more than 0.60 for the charged membrane. The large

increase in sieving coefficient with increasing ionic strength is consistent with the shielding

of the electrostatic interactions between the charged protein and the membrane by the added

electrolyte; this effect is much more pronounced for the negatively charged membrane due to

the direct charge-charge interactions between the membrane and the pegylated protein as

described qualitatively by the second term on the right-hand side of Equation (4.2). Note that

one cannot use Equation (4.2) to directly evaluate the magnitude of the electrostatic

interactions since the charge on the pegylated α-lactalbumin is located on the surface of the

protein and is thus "buried" beneath the polyethylene glycol layer that is attached to the α-

lactalbumin. This is discussed in more detail subsequently.

Page 94: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

76

Figure 4.1 Observed sieving coefficient of a 5 kDa PEG (right panel) and a pegylated α-

lactalbumin with one 5 kDa PEG chain (left panel) as a function of ionic strength

through both an unmodified and a 12-hr charged 100 kDa composite regenerated

cellulose membrane.

Corresponding data for α-lactalbumin attached to a single 20 kDa PEG chain are

shown in Figure 4.2. The data are plotted as the scaled sieving coefficient, defined as the

sieving coefficient at a given ionic strength divided by that in the high (200 mM) ionic

strength solution to highlight the effects of electrostatic interactions. The sieving coefficient

data for the charged membrane show a much stronger dependence on the solution ionic

strength, decreasing by a factor of 15 as the ionic strength is reduced from 200 to 2 mM

compared to only a 3-fold reduction in the sieving coefficient through the unmodified

membrane.

Page 95: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

77

Figure 4.2 Scaled sieving coefficient of a pegylated α-lactalbumin with one 20 kDa PEG

chain as a function of solution ionic strength during ultrafiltration through both an

unmodified and a 12-h charged 100 kDa composite regenerated cellulose

membrane.

The addition of one or more PEG chains to a protein can have at least 3 separate

effects on the sieving behavior: (1) the PEG increases the effective protein size, reducing the

accessibility of the space within the membrane pores, (2) the attachment of the PEG to the

lysine amino group eliminates the presence of a protonatable –NH2 group, thereby increasing

the net negative charge on the protein, and (3) the presence of the PEG alters the electrostatic

potential field around the protein, modifying the details of the electrostatic interactions

between the charged protein and the pore.

Page 96: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

78

In order to explore these effects in more detail, sieving data were obtained with a 5

kDa-pegylated and a mono-acetylated α-lactalbumin through a negatively charged

membrane, with the results shown in Figure 4.3. In each case, the observed sieving

coefficients are plotted for 3 repeat measurements at solution ionic strength of 10, 25, 50, and

200 mM. The mono-pegylated and mono-acetylated proteins have the same number of

blocked lysine amino groups and thus the same number of available ionizable groups on the

surface of the protein. The sieving coefficients of the mono-pegylated α-lactalbumin are

uniformly smaller than those of the mono-acetylated protein. At high ionic strength (200

mM), this difference (So = 0.94 ± 0.08 for the acetylated protein versus So = 0.61 ± 0.01 for

the pegylated) is due entirely to the increase in the effective size of the pegylated protein.

The sieving coefficients of both the pegylated and acetylated proteins decrease with

decreasing ionic strength, with the dependence on ionic strength being somewhat greater for

the pegylated α-lactalbumin. The observed sieving coefficient for the pegylated α-

lactalbumin decreased by more than a factor of 15 as the ionic strength was reduced from 200

to 10 mM, while the observed sieving coefficient of the acetylated protein decreased by only

a factor of 7 over the same range. Thus, the presence of the polyethylene glycol layer

appears to increase the magnitude of the electrostatic interactions, which is exactly opposite

the behavior reported previously for ion exchange chromatography of pegylated proteins

(Pabst et al., 2007).

Page 97: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

79

Figure 4.3 Observed sieving coefficients of a 5 kDa pegylated α-lactalbumin and a mono-

acetylated α-lactalbumin as a function of solution ionic strength for ultrafiltration

through a 12-hr charged composite regenerated cellulose membrane.

4.3.2 Electrophoretic Mobility

In order to understand the underlying electrostatic interactions in more detail,

experimental studies were performed to evaluate the electrophoretic mobility of both the

pegylated and acetylated α-lactalbumin. Figure 4.4 shows data for pegylated α-lactalbumin

formed by attachment of PEG chains with molecular weight of 2, 5, 10, and 20 kDa along

with corresponding results for acetylated α-lactalbumin with different numbers of acetylated

lysine groups. In each case, the data are plotted as a function of the number of modifications

to the native protein; thus, the species with n = 1 represents α-lactalbumin with either a single

PEG chain or a single acetylated lysine. The electrophoretic mobility of the acetylated

Page 98: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

80

species increases with increasing number of acetyl groups due to the increase in net negative

charge associated with the conversion of the protonated amino group to a neutral amide:

( )af H

ro

e κη

ζεεµ

3

2= (4.3)

where εo is the permittivity of a vacuum, εr is the dielectric constant of the solution, η is the

solution viscosity, and ζ is the zeta (or surface) potential of the protein, which is proportional

to the net protein charge. The quantity fH(κa) is typically referred to as Henry's function and

can be evaluated as an expansion in κa as (Menon and Zydney, 2000):

f κa( )=1+1

16κa( )2

−5

48κa( )3

−1

96κa( )4

+1

96κa( )5

+1

8κa( )4

exp κa[ ] 1−κa( )2

12

exp −t[ ]t

dtκa

∫ (4.4)

where a is the protein radius and κ is the inverse Debye length:

κ =1

εrεokTzi

2Cio

i=1

N

1/ 2

(4.5)

where k is Boltzmann's constant, T is the absolute temperature, and zi and Cio are the valence

and bulk concentration of all mobile ions.

The predicted values of the electrophoretic mobility given by Equations (4.3) – (4.5),

with the zeta potential evaluated in terms of the net protein charge as:

( )aa

Ze

or κεεπζ

+=

14 (4.6)

are shown as the solid curve in Figure 4.4 using a = 1.98 nm and e = 1.609 x 10-23 C/electron,

where Z is the net number of negative charge groups on the protein. The model is in good

agreement with the data for the acetylated proteins. The slight discrepancy for the highest

degree of modification may be due to the effects of charge regulation. At high degrees of

modification, the increase in negative charge associated with the acetylation will increase the

Page 99: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

81

local H+ concentration near the protein surface, which in turn leads to a slight protonation of

other basic or acidic amino acid residues. This phenomenon is discussed in more detail by

Menon and Zydney (2000).

In contrast to the data for the acetylated proteins, the electrophoretic mobility of the

pegylated α-lactalbumin decreased with increasing number of attached PEG groups even

though the pegylation reaction eliminates multiple amine groups (analogous to the

acetylation reaction). For example, the mobility of the α-lactalbumin with a single 5 kDa

PEG is approximately 34 % smaller than that of the native protein while the mobility of the

pegylated α-lactalbumin with a single 30 kDa PEG is more than 75 % smaller than that of the

un-modified α-lactalbumin. The electrophoretic mobility of the pegylated α-lactalbumin

with one 20 kDa PEG chain is slightly smaller than that of the pegylated protein with two 10

kDa PEG chains which is turn smaller than that for the pegylated protein with four 5 kDa

PEG chains even though these species all have basically the same molecular weight and the

same effective size as determined by size exclusion chromatography (Fee, 2007). These

differences are a direct result of the change in molecular charge; the pegylated proteins

formed with more PEG chains have a greater negative charge due to the conversion of

multiple amine groups to the corresponding amide.

Page 100: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

82

Figure 4.4 Electrophoretic mobility of the pegylated α-lactalbumin with different size PEG

chains as a function of the number of substituted lysine groups. Also shown for

comparison are data for the acetylated proteins. Experiments were performed

using 10 mM Tris–Glycine running buffer at pH 8.1. Error bars represent plus or

minus one standard deviation of the experimental data. Solid curve is model

calculation for acetylated proteins as described in the text.

In order to account for the change in both size and net charge of the pegylated

proteins, the experimental data in Figure 4.4 have been analyzed in terms of the drag ratio,

KD, which is equal to the electrophoretic mobility of the pegylated α-lactalbumin divided by

the mobility of the corresponding acetylated species with the same number of chemical

substitutions. Thus, the drag ratio for the pegylated α-lactalbumin with two 5 kDa PEG

groups was evaluated by normalizing its mobility using the corresponding mobility of the

doubly-acetylated protein. This method effectively accounts for the effects of pegylation and

acetylation on the protein charge since both species have the same number of blocked lysine

Page 101: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

83

amino groups. Sharma and Carbeck (2005) used a similar approach to calculate the effective

size of pegylated proteins using capillary electrophoresis.

Experimental data for the drag ratio for the different pegylated species are shown in

Figure 4.5 as a function of the effective radius of the pegylated protein (b), which was

calculated using the correlation presented by Fee and van Alstine (2004) as given by

Equations (2.42) to (2.44) in Chapter 2. The data for α-lactalbumin pegylated with different

numbers of 2, 5, 10, 20, and 30 kDa PEG chains all collapse to a single curve when plotted in

terms of the drag ratio. For example, the drag ratio for the pegylated α-lactalbumin with a

single 10 kDa PEG chain is 0.44, which is within 10% of the value for the pegylated α-

lactalbumin having two 5 kDa PEG chains (which is within the standard deviation of the

measurements). This is in sharp contrast to the 40% difference in the electrophoretic mobility

of these species arising from the difference in electrical charge caused by the modification of

the lysine groups.

Although the detailed structure of the pegylated protein is not known, a reasonable

approach to evaluate the electrophoretic mobility would be to use the analytical expression

presented by Ohshima (2002) for the mobility of a hard sphere covered by an ion-penetrable

uncharged polymer layer (analogous to the PEG layer in the pegylated protein). The results

are expressed in terms of the parameter:

2/1

=

ηα

f (4.7)

where f is the frictional coefficient within the polymer layer. In the limit of α→∞,

corresponding to the situation where the slip plane moves to the outer radius of the pegylated

protein, the electrophoretic mobility becomes:

Page 102: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

84

µe =εoεrζ

ηa

b

exp −κ b − a( )[ ]f H κb( ) (4.8)

The term (a/b) exp [−κ(b−a)] describes the decay in the electrostatic potential as one moves

through the ion-penetrable polymer. The drag ratio, KD, evaluated from the ratio of the

electrophoretic mobility of the pegylated protein (given by Equation 4.8) to that of the

acetylated protein (given by Equation 4.3), is shown by the dashed curve in Figure 4.5

assuming that the zeta potential of the pegylated and acetylated proteins are the same. The

model is in good agreement with the data for very small values of b, but it significantly

under-predicts the data for the more heavily pegylated proteins. For example, the α-

lactalbumin with a single 30 kDa PEG chain is predicted to have a drag ratio of 0.06 which is

3.5 times smaller than the experimental value of KD = 0.21.

One possible explanation for the large discrepancy between the model and data is that

the PEG layer provides relatively little hydrodynamic resistance, corresponding to a

relatively small value of α. However, previous studies of the hydrodynamic radius of

pegylated interferon using dynamic light scattering (Kusterle et al, 2008) indicate that the

pegylated protein behaves nearly as a hard sphere with radius b. An alternative explanation is

that the PEG layer has a much lower ion concentration, corresponding to a larger Debye

length, than that in the bulk electrolyte. The thermodynamics of PEG–salt systems have been

studied quite extensively (Willauer et al., 2002). These systems tend to phase separate due to

the strong “negative” interactions between the salts and the polyethylene glycol. The salt

concentration in the PEG phase can be as much as seven times smaller than the salt

concentration in the non-PEG phase (Willauer et al., 2002). Although it is difficult to directly

extrapolate these data for PEG-salt systems to the behavior of pegylated proteins, the results

clearly indicate that there may be a significant exclusion of salts from the polyethylene glycol

Page 103: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

85

layer in the pegylated protein. The simplest approach to include this effect in the evaluation

of the drag ratio is to modify the expression for the electrophoretic mobility of the pegylated

protein given by Equation (4.8) so that the Debye length that describes the decay in

electrostatic potential within the PEG layer is different from that in the bulk solution giving:

KD =a

b

f H κb( )f H κa( )

exp −κPEG b − a( )[ ] (4.9)

Note that the Debye length used in both Henry’s functions is equal to that determined from

the bulk electrolyte concentration since this function describes the relaxation of the double

layer in the bulk solution external to the PEG layer. The calculated values of the drag ratio

given by Equation (4.9) with κPEG / κ = 0.38, which is based on a 7-fold reduction in ionic

strength within the PEG layer, are shown as the solid curve in Figure 4.5. The model

calculations are in very good agreement with the experimental data over the entire range of

effective radius. More detailed results for KD developed using the complete solution for the

electrostatic potential surrounding a composite sphere are in good agreement with both

Equation (4.9) and the experimental data (Molek, 2008), providing strong support for the idea

of ion exclusion from the PEG layer of the pegylated protein.

Page 104: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

86

Figure 4.5 Drag ratio as a function of the effective radius for pegylated proteins containing

2, 5, 10, 20, or 30 kDa PEG chains. The solid and dashed curves are model

calculations as discussed in the text.

4.3.3 Partitioning Model

The results for the electrophoretic mobility suggest a simple approach for calculating

the sieving coefficient of the pegylated protein using Equations (4.1) and (4.2) by evaluating

the dimensionless surface charge density of the solute (protein) at the outer edge of the

pegylated protein to account for the ion exclusion from the PEG layer. The electrostatic

potential at the outer surface of the pegylated protein can be approximated in terms of the

zeta potential of the α–lactalbumin core based on the decay in potential through the PEG

layer as was done for the electrophoretic mobility:

Page 105: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

87

ψr= b =a

b

exp −κPEG b − a( )[ ]ζ (4.10)

The surface charge density at the outer edge of the pegylated protein is then evaluated

directly from Eq. (4.6) as:

( )[ ] sPEGPEG qaba

b

b

a

b

eZ−−

++

== κκκ

πσ exp

1

1

4

2

2 (4.11)

where qs is the charge density on the surface of the α–lactalbumin core. The ratio of the

surface charge density of a pegylated protein to that of the α–lactalbumin core ranges from

σPEG/σs, = 0.42 to 0.30 for the ionic strength between 10 to 200 mM using Eq. (4.11) with a =

1.98 nm, b = 3.16 nm, and κPEG/κ = 0.38. Note that this reduction in surface charge density

does not necessarily imply a reduction in the electrostatic interaction since the coefficients in

Equation (4.2) are strong functions of the protein radius; a large pegylated protein can have a

stronger electrostatic exclusion than the native protein even though it has a lower surface

charge density since the charged groups on the protein surface are in much closer proximity

to the pore wall.

In order to compare the experimental data with the model calculations, it is first

necessary to evaluate the actual sieving coefficient (Sa) from the measured values of the

observed sieving coefficient (So) by accounting for the effects of concentration polarization

as described in Section 2.2 of Chapter 2. Theoretical calculations were performed using

Equations (4.1) and (4.2) with the hindrance factor for convection (Kc) evaluated as a

function of λ = rs/rp using the hydrodynamic models described in Chapter 2 (Equations

(2.30), (2.32), (2.34), and (2.35). The surface charge density of the pore was evaluated as

described in section 3.3.5 of Chapter 3 based on the measured streaming potential. The

surface charge density of the mono-acetylated α-lactalbumin was evaluated from the known

Page 106: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

88

amino acid sequence based on the pKa values of the different ionizable groups (as described

in section 2.7 of Chapter 2). The best-fit value of the pore radius was determined by fitting

the data for the acetylated protein to the partitioning model giving rp = 4.3 nm. This value is

in good qualitative agreement with independent estimates of the pore size determined from

the measured hydraulic permeability and from dextran sieving data (Molek, 2008). This same

pore size was then used to calculate the sieving coefficients of the pegylated α-lactalbumin,

with the effective size of the pegylated protein determined from correlations provided by Fee

and van Alstine (2004) (discussed in section 2.6 in Chapter 2) and the protein surface charge

density given by Equation (4.11). The model calculations are in good agreement with the

experimental data for both the acetylated and mono-pegylated α-lactalbumin over the entire

range of ionic strength (Figure 4.6). The good agreement between the model and data for the

acetylated protein is consistent with prior results in the literature (Burns and Zydney, 1999;

Mehta & Zydney, 2006; Burns and Zydney, 2001). The good agreement with the data for the

pegylated protein suggests that the simple approach of shifting the protein charge to the outer

edge of the pegylated protein, accounting for the reduction in electrostatic potential

associated with the reduced electrolyte concentration within the PEG layer, provides an

appropriate framework for describing the electrostatic interactions of pegylated proteins

during ultrafiltration. More detailed calculations accounting for charge regulation effects

(Pujar and Zydney, 1997) might provide more accurate predictions for the actual sieving

coefficients, although it was difficult to justify including either of these phenomena given the

relatively simple approximation used to evaluate the surface charge density of the pegylated

proteins. The small discrepancy observed between the theory and data could also reflect some

conformational change in the α-lactalbumin associated with the pegylation and/or acetylation,

Page 107: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

89

although previous studies of pegylated and acetylated proteins have generally shown minimal

alteration in the three-dimensional protein structure due to these surface modifications.

Figure 4.7 shows experimental data for the actual sieving coefficients for the 5 kDa

pegylated and acetylated α-lactalbumin as an explicit function of the number of substitutions.

Thus, the data point plotted at n = 1 corresponds to either the mono-pegylated or mono-

acetylated species. At high ionic strength (200 mM, right panel), the sieving coefficients of

the acetylated protein are essentially independent of the number of modifications, with values

above Sa = 0.7, since the small acetyl group has no significant effect on the protein size. In

contrast, the actual sieving coefficient of the pegylated species decreases from Sa = 0.85 for

the mono-pegylated protein to Sa = 0.017 for the protein with three 5 kDa PEG groups due to

the increase in effective size associated with the PEG chains. The data at 10 mM ionic

strength (left panel) are strongly influenced by electrostatic interactions due to the low degree

of shielding provided at the low salt concentration. In this case, the actual sieving coefficients

for the acetylated and pegylated species both decrease with increasing degree of

modification. The reduction in actual sieving coefficient of the acetylated species is due to

the increase in electrostatic exclusion arising from the increased negative charge associated

with the elimination of one or more protonated amine groups. This effect also influences the

behavior of the pegylated proteins, although in this case the very large reduction in sieving

coefficient with increasing number of substituted amine groups reflects the combined effects

of: (1) the increase in negative charge, (2) the increase in effective size, and (3) the

displacement of the effective protein charge to the outer edge of the PEG layer associated

with the exclusion of bulk ions from the PEG layer.

The solid curves in Figure 4.7 are model calculations based on the simple partitioning

model using the same model parameters as in Figure 4.6. The calculated values of the sieving

Page 108: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

90

coefficient for the acetylated α-lactalbumin were developed directly from Equations (4.1) and

(4.2) with the protein charge evaluated by subtracting one for each blocked lysine group. The

calculations for the pegylated protein account for all three of the phenomena discussed in the

prior paragraph. The increase in negative charge was evaluated using the calculated charge

for the corresponding acetylated species. The increase in effective size was evaluated using

the correlation provided by Fee and van Alstine (2004). The displacement of the effective

protein charge to the outer edge of the PEG layer was evaluated using Equations (4.10) and

(4.11). The model is in good agreement with the experimental data for both the acetylated

and pegylated proteins.

Figure 4.6 Actual sieving coefficients of a 5 kDa pegylated α-lactalbumin and a mono-

acetylated α-lactalbumin as a function of solution ionic strength for ultrafiltration

through a 12-h charged 100 kDa composite regenerated cellulose membrane.

Solid curves are model calculations as described in the text.

Page 109: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

91

Figure 4.7 Actual sieving coefficients of the acetylated and pegylated α-lactalbumin (with 5

kDa PEG chains) through a negatively charged cellulose membrane as a function

of the number of substituted lysine groups at both 10 mM (left panel) and 200

mM (right panel) ionic strength. Solid curves are model calculations as described

in text.

4.4 Conclusion

Although several previous studies have demonstrated the importance of electrostatic

interactions in protein ultrafiltration, there have been no prior studies of these interactions for

pegylated proteins formed by covalent attachment of a neutral PEG chain to a charged

protein. In contrast to studies of ion exchange chromatography, in which the attached PEG

significantly reduces the extent of binding to the charged chromatography resin (Pabst et al.,

2007), the data obtained in this study demonstrate that the pegylated proteins can show even

stronger electrostatic repulsion than the native (unmodified) protein. Experiments with

pegylated α-lactalbumin through a charged composite regenerated cellulose membrane show

more than an order of magnitude reduction in the sieving coefficient as the ionic strength is

Page 110: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

92

reduced from 200 to 10 mM compared to only a 7-fold reduction in sieving coefficient of an

acetylated version of the protein under the same conditions.

The addition of the PEG chains has three separate effects on protein transmission

during ultrafiltration: (1) the PEG increases the effective protein size, reducing the

accessibility of the space within the membrane pores, (2) the attachment of the PEG to the

lysine amino group eliminates the presence of a protonatable–NH2 group, increasing the net

negative charge on the protein, and (3) the presence of the PEG alters the electrostatic

potential field around the protein. Experimental data for the electrophoretic mobility were in

good agreement with a simple model in which the plane of shear is displaced to the outer

edge of the PEG layer with the electrostatic potential at the outer surface of the pegylated

protein evaluated by accounting for the ion exclusion from the PEG. An analogous model

was developed for the protein sieving coefficient, with the experimental data in good

agreement with the resulting model calculations. All model parameters were evaluated from

independent measurements: the membrane pore size was determined from sieving data

obtained with the acetylated proteins, the membrane surface charge density was determined

from streaming potential measurements, and the charge on the protein core was determined

from the known amino acid sequence and pKa values of the ionizable groups accounting for

the conversion of one or more amine groups to the corresponding amide. The model

accurately describes the key experimental observations: the reduction in sieving coefficient at

high ionic strength is due to the increase in effective size of the pegylated protein while the

reduction in sieving coefficient at low ionic strength is due to both the increase in effective

size and the strong electrostatic interactions arising from the displacement of the effective

protein charge to the outer surface of the large pegylated species. This theoretical description

Page 111: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

93

provides an appropriate framework for analyzing the retention characteristics of pegylated

proteins during ultrafiltration.

Page 112: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

94

Chapter 5

Separation of Pegylated Proteins from Reactants using a Single Charge-

modified Membrane

5.1 Introduction

One of the challenges in the production of pegylated proteins is the purification of the

desired conjugate, typically the mono-pegylated protein, from the unreacted PEG and native

protein and also from any multiply pegylated conjugates. Although the use of site-specific

pegylation can produce a molecularly defined mono-pegylated product with few (if any)

multiply pegylated species (Jevsevar et al., 2010), the coupling reactions are typically with

excess PEG and even then the reactions do not go to completion. For example, Kinstler et al.,

(2002) performed a site-specific pegylation of recombinant human granulocyte colony-

stimulating factor (G-CSF) via its N-terminal using 5 molar excess of a 6 kDa PEG,

providing 92% yield of the mono-pegylated protein with 8% of unreacted G-CFS and a great

deal of unreacted PEG.

Previous work by Molek and Zydney (2007) demonstrated the feasibility of using a

2-stage diafiltration process for removal of the unreacted protein and PEG from the singly

pegylated product. The native protein was removed in the filtrate in the first stage using a

neutral 30 kDa membrane, exploiting the size difference between the small native protein and

the pegylated conjugate. The second stage employed a negatively charged 100 kDa

membrane to remove the neutral PEG while the pegylated proteins were retained due to the

strong electrostatic interactions (repulsion) between the membrane and protein. However,

Page 113: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

95

this process required two membrane ultrafiltration steps, increasing the cost and time for the

separation. The final purification factor was more than 20-fold with only 75% product yield,

with much of the product loss occurring during transfer of the feed between the two stages.

This chapter presents an alternative process for removal of the unreacted protein and

PEG using a single ultrafiltration step employing a relatively large pore size (300 kDa

molecular weight cut-off) negatively-charged membrane. The results showed a greater yield

of the singly pegylated protein with good purification, clearly demonstrating the feasibility of

this single-step ultrafiltration process.

5.2 Materials and Methods

Pegylated α-lactalbumin was prepared by reaction with N-hydroxysuccinimide

activated PEG as described in Chapter 3. A negatively charged cellulose ultrafiltration

membrane was prepared by surface modification of a 300 kDa UltracelTM membrane (EMD

Millipore, Bedford, MA) by attachment of sulfonic acid groups using the base-activated

chemistry described in Chapter 3.

Protein sieving experiments were performed in a 25 mm Amicon stirred cell at a

stirring speed of 600 rpm. Actual protein separations were performed by diafiltration. The

concentrations of the pegylated proteins, the native α-lactalbumin, and PEG were determined

by size exclusion chromatography as described in Chapter 3.

Page 114: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

96

5.3 Results

5.3.1 Sieving Experiments

Initial sieving experiments were performed to identify appropriate conditions for

separating PEG and native α-lactalbumin from the mono-pegylated protein. Figure 5.1

shows typical data for the observed sieving coefficient (So), defined as the ratio of the solute

concentration in the filtrate solution to that in the feed, through both an unmodified and a 24-

hr negatively charged version of the 300 kDa UltracelTM membrane as an explicit function of

the solution ionic strength. The data were obtained using a feed mixture containing the PEG

(approximately 0.51 g/L), unreacted α-lactalbumin (0.34 g/L), and the mono-pegylated

protein (1.40 g/L), with the concentration of each component evaluated by size exclusion

chromatography. Results for the di- and tri-pegylated proteins, which were also present in the

feed solution, are not shown. In each case, the membrane was used to filter a solution of the

feed mixture in a 0.5 mM Bis-Tris buffer at pH 6.6. The filtrate flux was maintained at

approximately 8 µm /s to minimize concentration polarization effects in the stirred cell due to

the relatively low bulk mass transfer coefficients in this device (Kwon et al., 2008). The

experiments were performed from low to high ionic strength, which was adjusted by addition

of 1 M NaCl solution containing 0.5 mM Bis-Tris buffer.

Page 115: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

97

Figure 5.1 Observed sieving coefficients for mono-pegylated α-lactalbumin, native α-

lactalbumin, and 20 kDa PEG as a function of ionic strength through an

unmodified 300 kDa UltracelTM membrane (blank symbols) and a 24-hr

negatively charged version of the membrane (filled symbols).

The sieving coefficients for all three species through the unmodified membrane were

a relatively weak function of the solution ionic strength. The transmission of the 20 kDa PEG

through the unmodified membrane was essentially independent of solution ionic strength,

varying between 0.69 and 0.72, consistent with the absence of significant electrostatic

interactions between the neutral PEG and the unmodified cellulose membrane. The sieving

coefficient for the native α-lactalbumin decreased slightly from 0.98 at the highest ionic

strength to 0.75 at the lowest ionic strength. The increase in protein retention at low ionic

Page 116: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

98

strength is due to the small but not negligible electrostatic interactions between the negatively

charged protein and the unmodified membrane, which possessed a slightly negative surface

charge as discussed in Chapter 3. The sieving coefficient for the mono-pegylated α-

lactalbumin decreased from 0.57 to 0.13 as the ionic strength was reduced from 200 to 0.5

mM.

In contrast, the sieving data for both the mono-pegylated and native α-lactalbumins

though the negatively charged membrane is a strong function of the solution ionic strength.

The sieving coefficient for the α-lactalbumin decreased approximately 6-fold from 0.99 to

0.17 as the solution ionic strength was reduced from 200 mM to 0.5 mM while the sieving

coefficient for the mono-pegylated α-lactalbumin decreased by more than 100-fold over the

same range of conditions. This large reduction in the sieving coefficient is a direct result of

the repulsive interactions between the highly negatively charged membrane and the

negatively charged proteins. The more pronounced ionic strength dependence for the mono-

pegylated protein is due to several factors: (1) the mono-pegylated protein has a greater

negative charge than the native α-lactalbumin due to the removal of one positive lysine amine

group due to the pegylation reaction, (2) the pegylated protein is larger than the α-

lactalbumin, and (3) the presence of the PEG alters the electrostatic potential surrounding the

protein due to salt partitioning in the PEG layer. This latter effect was discussed in Chapter 4.

The sieving coefficient for the PEG was essentially independent of the solution ionic

strength, varying between 0.67 and 0.72, implying negligible distortion of the electrical

double layer adjacent to the pore wall at low ionic strength due to the relatively large pore

size (small λ).

The experimental data in Figure 5.1 were used to calculate the selectivity, with the

results shown in Figure 5.2. The selectivity is defined as the ratio of the sieving coefficient of

Page 117: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

99

the impurity (either the PEG or the native protein) relative to the desired product (the mono-

pegylated protein). As expected, the selectivity for the separation of PEG from the mono-

pegylated protein increased as the ionic strength was reduced due to the increased retention

of the product, with the maximum selectivity of 150 attained in the 0.5 mM ionic strength

solution. Note that it might be possible to achieve a higher selectivity at an even lower ionic

strength; however, it was difficult to maintain the desired pH with a lower buffer

concentration. Although the sieving coefficients for both the native and mono-pegylated α-

lactalbumin decreased with decreasing ionic strength, the selectivity between the native α-

lactalbumin and the mono-pegylated protein increased at low ionic strength due to the more

pronounced retention of the mono-pegylated protein giving a maximum selectivity of 36 in

the lowest ionic strength solution.

Page 118: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

100

Figure 5.2 Selectivity between the mono-pegylated α-lactalbumin and either the native α-

lactalbumin or the 20 kDa PEG through an unmodified 300 kDa UltracelTM

membrane (blank symbols) and a 24-hr negatively charged version of the

membrane (filled symbols).

The effect of solution pH on the sieving behavior is examined in Figure 5.3. The data

were obtained using a 0.5 mM acetate buffer at pH 5, a 0.5 mM Bis-Tris at pH 6.0-7.0, and a

0.5 mM Trizma base at pH 7.5. The protein feed was placed in the desired buffer by

diafiltration through a 10 kDa UltracelTM membrane as described in Chapter 3. In each case

the pH was adjusted to the desired value by addition of 0.1 M HCl or NaOH as needed. The

very low ionic strength was chosen to maximize the electrostatic interactions between the

Page 119: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

101

membrane and proteins. The filtrate flux was maintained at approximately 8 µm/s. The

sieving coefficients for both α-lactalbumin and the mono-pegylated α-lactalbumin decreased

significantly as the pH was increased from 5 to 7.5. This is due to the protonation of

ionizable amino acids on the proteins, resulting in an increase in the net negative charge and

a corresponding increase in the electrostatic repulsion. The sieving coefficient for the PEG

remained approximately constant over this pH range since the PEG is electrically neutral.

Figure 5.3 Observed sieving coefficients for mono-pegylated α-lactalbumin, native α-

lactalbumin, and the 20 kDa PEG as a function of solution pH through a 24-hr

negatively charged version of 300 kDa UltracelTM membrane.

The sieving coefficients from Figure 5.3 were used to calculate the selectivity with

results shown in Figure 5.4. The selectivity for the removal of PEG increased substantially

Page 120: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

102

with increasing pH due to the large reduction in the transmission of the mono-pegylated

protein. The maximum selectivity of 270 was obtained at pH 7.5. In contrast, the selectivity

for the separation of the mono-pegylated protein from the α-lactalbumin attained a maximum

value of 47 at pH 6.4 before decreasing significantly at higher pH. This more complex

behavior arises from the changes in electrostatic interactions of the pegylated and native

protein with solution pH.

Figure 5.4 Selectivity for the removal of native α-lactalbumin and PEG from the mono-

pegylated α-lactalbumin as a function of solution pH through a 24-hr negatively

charged version of the 300 kDa UltracelTM membrane in 0.5 mM buffers.

Page 121: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

103

5.3.2 Purification of Mono-pegylated Protein

A diafiltration process was designed to separate the unreacted α-lactalbumin and

PEG from the mono-pegylated protein using a 300 kDa UltracelTM membrane that was

negatively charged for 24 hr. The data were obtained using a feed mixture containing the

PEG (0.95 g/L), unreacted α-lactalbumin (0.33 g/L), and the mono-pegylated protein (1.30

g/L). Some di- and tri-pegylated proteins were also present in the feed solution; the results

are not shown. The diafiltration was performed at a filtrate flux of 8 µm/s using a 0.5 mM

ionic strength Bis-Tris buffer at pH 6.6. Results are shown in Figure 5.5 for the yield (Y) in

the retentate solution:

i

ff

VC

CVY −= 1 (5)

where fV is the cumulative volume of collected filtrate, fC is the average concentration of a

given species in the filtrate, V is the feed volume, and iC is the initial concentration in the

feed (retentate) solution. The results are plotted as a function of the number of diavolumes

(N), which is simply equal to the total collected filtrate volume divided by the constant

retentate volume. The yield of PEG and α-lactalbumin decreased rapidly, reaching 5.0% and

11% after 8 diavolumes, while more than 90% of the mono-pegylated protein was recovered

in the retentate.

Page 122: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

104

Figure 5.5 Yield for the native α-lactalbumin, PEG, and mono-pegylated α-lactalbumin in

the retentate solution as a function of number of diavolumes for a diafiltration

performed with a 300 kDa UltracelTM membrane charged for 24 hr. Data were

obtained at pH 6.6, 0.5 mM ionic strength, and a filtrate flux of 8 µm/s. Solid

curves are model calculations described in the text.

The solid curves in Figure 5.5 are model calculations developed from solution of the

overall mass balance for a constant volume diafiltration (van Reis and Saksena, 1997):

)exp( oNSY −= (6)

where the protein sieving coefficient (So) is assumed to remain constant throughout the

diafiltration. The best fit values of So for the PEG, native α-lactalbumin, and mono-pegylated

α-lactalbumin were determined by minimizing the sum of the squared residuals between the

model and data, with results summarized in Table 5.1. The model calculations are in

Page 123: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

105

excellent agreement with the experimental data, properly capturing the variation in the

solution concentrations during the diafiltration process.

Table 5.1 Best-fit values of the protein sieving coefficients for the diafiltration process

Solute Observed Sieving Coefficient, So

α-lactalbumin 0.36

PEG 0.29

Mono-pegylated 0.0096

The separation performance of the diafiltration process was examined in more detail

by constructing a plot of the tradeoff between the yield of the mono-pegylated α-lactalbumin

product and purification factor (P). The results are shown in Figure 5.6 using PEG and the

native α-lactalbumin as the impurities. The purification factor is defined as the ratio of the

yield of the mono-pegylated α-lactalbumin in the retentate to that of a given impurity. The

diafiltration begins in the upper left-hand corner of Figure 5.5 with 100% of the mono-

pegylated protein in the retentate and a purification factor of 1 (since all of the impurity is

also in the retentate). The purification factor increases throughout the diafiltration process as

the PEG and native protein are removed in the filtrate, with the yield of the mono-pegylated

protein decreasing slightly because of the slow leakage of the product through the membrane.

The final diafiltration provided a purification factor of 22 for PEG removal and 29 for α-

lactalbumin removal with 90% yield of the desired mono-pegylated protein.

Page 124: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

106

Figure 5.6 Yield for the mono-pegylated α-lactalbumin as a function of purification factor.

Circle symbols are for removal of the native α-lactalbumin; squares are for

removal of the PEG. Solid and dashed curves are model calculations described

in the text.

The solid curves in Figure 5.6 are model calculations for a diafiltration process with

the product collected in the retentate solution (van Reis and Saksena, 1997):

Ψ−= 1YP (6)

where ψ is the selectivity between the mono-pegylated α-lactalbumin and the impurity,

either the PEG or native α-lactalbumin. The solid (blue) curve in Figure 5.6 was developed

using ψα−lac = 37 while the dashed (green) curve was for ψPEG = 31 based on the best-fit

values of the sieving coefficients given in Table 5.2. The model calculations are in good

agreement with the experimental results. The small discrepancies between the data and model

Page 125: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

107

could be due to small fluctuations in the retentate volume during the diafiltration.

Extrapolation of the model to larger numbers of diavolumes shows that a purification factor

greater than 100 could be obtained using N= 13.2 with a product yield of 88% for removal of

the α-lactalbumin and N = 16.2 with a product yield of 86% for removal of PEG.

5.4 Conclusions

The results obtained in this chapter provided the first demonstration that it is

possible to remove unreacted (native) protein and PEG from a desired pegylated product

using a single ultrafiltration step by operating at very low ionic strength and with an

electrically charged membrane. The use of the highly charged UltracelTM membrane provided

high retention of the mono-pegylated protein due to the strong electrostatic interactions while

the relatively large pore size (300 kDa MW cut-off) allowed the smaller reactants to pass

easily into the filtrate. The process provided greater than 90% yield with purification factors

of more than 20 for both impurities. This performance is actually better than that reported by

Molek and Zydney (2007) using a two-stage ultrafiltration process employing a neutral and

charged membrane in the two stages (20-fold purification but only 75% yield of the mono-

pegylated protein). The ability to remove the unreacted protein and PEG in a single

membrane step provides opportunities for coupling the separation with the pegylation

reaction to increase the overall yield of the desired mono-pegylated product. This is discussed

in more detail in Chapter 8.

Page 126: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

108

Chapter 6

Removal of Multiply Pegylated Proteins using Charged Ultrafiltration

Membranes

Note: The material presented in this Chapter was taken from: Ruanjaikaen, K., Zydney,

A.L., 2011. Purification of singly-Pegylated alpha-lactalbumin using charged

ultrafiltration membranes. Biotechnology and Bioengineering 108, 822 – 829.

6.1 Introduction

The higher order (multiply) pegylated species generated during the random

pegylation process are typically considered as process impurities and must be removed

during downstream processing. For example, Grace et al. (2001) described the production of

pegylated interferon α-2b in which the concentration of the di-pegylated protein was reduced

to approximately 3% in the final product by cation exchange chromatography (with

undetectable amounts of more heavily pegylated species). Several studies have demonstrated

the feasibility of using ion exchange chromatography (Pabst et al., 2007; Piquet et al., 2002;

Lee et al., 2008; Kinstler et al., 1996; Yun et al., 2005) to separate the differently pegylated

species; however, the attached PEG typically causes a dramatic reduction in dynamic binding

capacity by shielding the protein surface charge, by providing a steric hindrance for binding,

and / or by reducing mass transfer (Fee and van Alstine, 2006). For example, the dynamic

binding capacity for pegylated BSA with a 30 kDa PEG to a Fractoprep TMAE anion

exchange resin was reduced by more than 100-fold compared to that of the native protein

Page 127: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

109

(Pabst et al., 2007). Moosmann et al. (2010) reported similar trends for mono-pegylated

lysozyme using cation exchange chromatography.

Molek and Zydney (2007) demonstrated that a two-stage diafiltration process can

remove the unreacted protein and PEG from the product; similar separation can be obtained

with improved yield using a single charged membrane as discussed in Chapter 3. However,

these membrane systems provided no separation between the differently pegylated species.

On the other hand, the data presented in Chapter 4 demonstrated that multiply pegylated

proteins could be very strongly retained by electrically charged ultrafiltration membranes due

to a combination of electrostatic and steric interactions, suggesting that it might be possible

to exploit these interactions to separate differently pegylated species by ultrafiltration.

The objective of the work described in this Chapter was to examine the feasibility of

using ultrafiltration for the purification of a singly pegylated α-lactalbumin pegylated from

the multiply pegylated conjugates using a diafiltration process with buffer conditions chosen

to maximize the electrostatic exclusion of the multiply pegylated species from the charged

membrane.

6.2 Materials and Methods

6.2.1 Preparation of Pegylated Proteins

Pegylated α-lactalbumins were prepared by reaction with N-hydroxysuccinimide

activated PEG similar to that explained in Chapter 3. The activated PEG was added at a

molar ratio of 2.5:1 to a solution of α-lactalbumin in 10 mM Bis-Tris buffer at pH 7. The

unreacted α-lactalbumin, PEG, and N-hydroxysuccinimide were removed from the mixture

Page 128: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

110

using the two-stage diafiltration process developed by Molek and Zydney (2007). Briefly, the

mixture was diluted four-fold with 10 mM Bis-Tris buffer. The first diafiltration used an

unmodified 30 kDa ultrafiltration membrane to remove α-lactalbumin and N-

hydroxysuccinimide. The final retentate was then processed by a second diafiltration using a

negatively charged 100 kDa UltracelTM membrane at low ionic strength to remove the neutral

PEG. All diafiltrations were performed in Amicon stirred ultrafiltration cells (Millipore

Corp., Bedford, MA) using the procedure described in Chapter 3.

The pH of the resulting solution was adjusted by buffer exchange using 0.5 mM Bis-

Tris (for pH 6.0 - 7.0) or acetate (for pH 4.5 - 5) as the diafiltration buffer. The solution ionic

strength was then adjusted to the desired value by addition of 1 M NaCl. The final solution,

which contained the mono-pegylated α-lactalbumin at a concentration of approximately 0.7

g/L, was filtered through a 0.2 µm pore size Acrodisc syringe filter (Pall corporation, Ann

Arbor, MI) to remove any insoluble aggregates before use.

6.2.2 Ultrafiltration Membranes

Ultrafiltration and diafiltration were performed using unmodified and negatively

charged versions of the UltracelTM composite regenerated cellulose membrane with nominal

molecular weight cut-off of 300 kDa (Millipore Corp., Bedford, MA). Negatively-charged

membranes were made by covalent attachment of sulfonic acid groups using the procedure

described in Chapter 3. The hydraulic permeability of each membrane was evaluated before

and after each experiment to obtain a measure of the extent of fouling. The membrane charge

was determined using streaming potential measurements following the procedure discussed

in Chapter 3.

Page 129: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

111

6.2.3 Ultrafiltration Experiments

Ultrafiltration (sieving) experiments were performed to identify appropriate

conditions for the separation of the mono-pegylated protein from the multiply pegylated

species. Data were obtained using an Amicon 8010 stirred cell (EMD Millipore, Bedford,

MA) following the procedure provided in Chapter 3. The filtrate flux was controlled by

applied air pressure. A minimum of 1.5 mL of filtrate was collected before each measurement

to remove the dead volume underneath the membrane and to eliminate any transients

associated with the change in pressure.

Small samples of the filtrate and retentate solutions were collected for subsequent

analysis by size exclusion chromatography performed using a Superdex 200, 10/300 column.

Protein concentrations were determined from their UV signals at 280 nm. Details about the

HPLC operation and evaluation of the protein concentrations are provided in Chapter 3.

6.2.4 Diafiltration Experiments

The actual separation of the mono-pegylated protein from the multiply pegylated

species was performed using diafiltration. The membrane was first equilibrated with a small

amount of the feed mixture containing the mono- and multiply pegylated proteins to

minimize potential transients associated with protein adsorption. The stirred cell was then

filled with the feed mixture and connected to a solution reservoir containing protein-free

buffer. The solution reservoir was air-pressurized, with the filtrate flux adjusted to the desired

value using a pressure regulator. The filtrate flux was evaluated at multiple time points over

the course of the diafiltration, with filtrate samples collected periodically for subsequent

Page 130: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

112

analysis by size exclusion chromatography. At the end of the diafiltration, the stirred cell was

opened and a retentate sample was obtained to verify closure of the mass balance.

6.3 Results and Analysis

6.3.1 Ultrafiltration Results

Initial sieving experiments were performed to identify appropriate conditions for

separating the mono-pegylated protein from the multiply pegylated species. Table 1 shows

typical data for the observed sieving coefficient (So), defined as the ratio of the protein

concentration in the filtrate solution to that in the retentate, for membranes charged for

different periods of time. The data for zero charging time were obtained using an unmodified

300 kDa UltracelTM membrane, with the other results obtained using negatively charged

versions produced by reacting the UltracelTM membrane with 3-bromopropanesulfonic acid

for the stated time. In each case, the membrane was used to filter a solution of the pegylated

α-lactalbumin (containing the mono-, di-, and tri- pegylated protein) in a 0.5 mM acetate

buffer at pH 5.0. The filtrate flux was maintained at approximately 8 µm/s (corresponding to

29 L/m2/h) to minimize concentration polarization effects in the stirred cell due to the

relatively low bulk mass transfer coefficients in this device (Kwon et al., 2008).

Page 131: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

113

Table 6.1 Observed sieving coefficients for mono-, di-, and tri- pegylated α-lactalbumin for

membranes charged for different periods of time. Data were obtained in a 0.5 mM

acetate buffer at pH 5 using a filtrate flux of 8 µm/s.

Charging Time (hr)

Observed Sieving Coefficient, So

Mono-pegylated Di-pegylated Tri-pegylated

0

12

24

0.74 ± 0.01

0.52 ± 0.02

0.66 ± 0.02

0.49 ± 0.01

0.08 ± 0.01

0.005 ± 0.001

0.25 ± 0.02

0.009 ± 0.003

0.001 ± 0.000

The data in Table 6.1 represent the mean values of the observed sieving coefficients

determined from at least two repeat measurements using the same membrane. The observed

sieving coefficients for the tri-pegylated protein are uniformly smaller than those for the di-

and mono-pegylated proteins due to the greater size of the more heavily pegylated species.

The observed sieving coefficients for the di- and tri-pegylated proteins decrease with

increasing charging time, consistent with the electrostatic exclusion of these negatively

charged proteins from the negatively charged pores. In contrast, the observed sieving

coefficients for the mono-pegylated protein remained approximately constant, with So

varying between 0.50 and 0.75. This behavior is discussed in more detail subsequently. The

net result is that the unmodified UltracelTM (zero charging time) provided little separation

between the differently pegylated species while the membrane charged for 24 hr provided

significant transmission of the mono-pegylated protein with greater than 99% retention of

both the di- and tri-pegylated α-lactalbumin.

Page 132: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

114

The effect of solution ionic strength on the ultrafiltration behavior through the

membrane that was charged for 24 hr is examined in Figure 6.1. The data are plotted in terms

of the selectivity (ψ), which is the key parameter describing the effectiveness of a membrane

separation process (van Reis and Saksena, 1997):

ψ =SPEG1

SPEG 2

(6.1)

where SPEG1 is the observed sieving coefficient of the mono-pegylated protein (in this case

the desired product) and SPEG2 is the observed sieving coefficient of the di-pegylated protein

(the key impurity). The selectivity increases significantly with decreasing solution ionic

strength, going from a value of ψ = 2.2 at high ionic strength to more than 130 in the 0.4 mM

ionic strength solution. The solid curve is a model calculation which is discussed in more

detail in the following section.

Page 133: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

115

Figure 6.1 Selectivity between the mono- and di-pegylated α-lactalbumin as a function of

solution ionic strength for ultrafiltration through a 300 kDa UltracelTM membrane

charged for 24 hr. Data were obtained at pH 5 using a filtrate flux of

approximately 8 µm/s. The solid curve is the model calculation as described in

the text.

6.3.2 Model Calculations

In order to understand the effects of solution ionic strength and membrane charge on

the sieving behavior of the pegylated proteins in more detail, the experimental data were

analyzed using the simple theoretical model for the transmission of pegylated proteins

through narrow pore ultrafiltration membranes, which was discussed in detail in Chapter 4.

The model evaluates the partition coefficient of a pegylated protein between the bulk solution

and a cylindrical pore accounting for: (1) the increase in effective protein size associated with

Page 134: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

116

the attached PEG, (2) the increase in net negative charge associated with the elimination of

one protonatable –NH2 group associated with the reaction at the lysine amine, and (3) the

alteration in the electrostatic potential field around the protein due to the presence of the PEG

layer. The pegylated protein is modeled as a charged sphere with the electrostatic potential at

the outer surface of the sphere evaluated from the charge on the protein core accounting for

the ion exclusion from the PEG layer. The resulting expression for the surface charge density

of the pegylated protein is provided as Equation (4.11).

The solid curve in Figure 6.1 is the model calculation with the protein charge density

determined using Equation (4.11). The model accounts for the effects of convection and

diffusion as discussed in Chapter 2. The membrane surface charge density was evaluated

from the measured streaming potential as described in Chapter 3, yielding qp = -2.7 mC/m2.

The calculations were performed assuming a log-normal pore size distribution as discussed in

Chapter 2 with the mean pore size determined from the hydraulic permeability of the

membrane using Equation (2.8), assuming a coefficient of variation of 2.0/ =rσ based on

previous studies of the pore size distribution of ultrafiltration membranes (Molek, 2008;

Mehta & Zydney, 2005; Zeman and Zydney, 1996). The model calculations are in good

qualitative agreement with the experimental data, properly capturing the significant reduction

in selectivity with increasing solution ionic strength. At high ionic strength, the selectivity

between the mono- and di-pegylated proteins is due entirely to the difference in size, which

leads to very small values of ψ (similar to the results for the unmodified membrane in Table

1). However, at low ionic strength, the di-pegylated protein is strongly excluded from the

membrane pores due to its large size and the presence of a significant effective surface

charge density; the calculated value of qPEG2 was -0.18 mC/m2 as given by Equation (4.11).

The mono-pegylated protein has a much smaller net negative charge (qPEG1 = -0.089 mC/m2)

Page 135: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

117

due to the presence of an additional positively charged amine group (since there is one less

amine coupled to a PEG). The net result is a highly selective separation between the mono-

and di-pegylated species. This selectivity is completely lost when using an unmodified

(essentially neutral) membrane (Table 6.1) due to the absence of any significant electrostatic

exclusion under these conditions.

The effect of solution pH on the selectivity between the mono- and di-pegylated

proteins is examined in Figure 6.2. Protein solutions were buffered with 0.5 mM Bis-Tris at

pH 6.5 and 7.0 and with 0.5 mM acetate at pH 4.5 and 5.0. In each case, the pH was adjusted

to the desired valued by addition of 0.1 M HCl or NaOH as needed. No additional salt was

added to the solutions; the very low ionic strength was chosen to enhance the electrostatic

interactions. The selectivity initially increases with decreasing pH attaining a maximum value

of ψ = 130 at pH 5, which is near the isoelectric point of the mono-pegylated protein. The di-

pegylated protein still has a net negative charge at this pH due to the removal of an additional

positive amine group associated with the attachment of the second PEG chain. The reduction

in selectivity at pH 4.5 reflects the similar electrostatic energy of interaction for mono- and

di-pegylated proteins, both of which have a small positive charge under these conditions.

Page 136: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

118

Figure 6.2 Selectivity between the mono- and di-pegylated α-lactalbumin as a function of

solution pH for ultrafiltration through a 300 kDa UltracelTM membrane charged

for 24 hr. Data were obtained using acetate or BisTris buffers with

approximately 0.5 mM ionic strength at a filtrate flux of approximately 8 µm/s.

The solid curve is the model calculation as described in the text.

The solid curve in Figure 6.2 is the model calculation using an average solution ionic

strength of 0.45 mM. The model calculations properly capture the increase of the selectivity

towards pH 5, although there are significant discrepancies between the model and data at

both low and high pH. One possible explanation for this behavior is that the model

calculations were performed assuming that the membrane charge remained constant at qp = -

2.7 mC/m2, neglecting the possible variation in membrane charge with solution pH (although

this effect should be small given the pKa of the sulfonic acid groups that provide the

Page 137: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

119

membrane charge). Another contribution to the model uncertainty is the use of the linearized

form of the Poisson-Boltzmann equation to evaluate the partition coefficient (and thus the

sieving coefficient) could lead to significant errors at the very high surface potentials that

exist under these low salt conditions. This effect would be greatest at pH both above and

below the isoelectric point where the protein is move heavily charged, which is where the

discrepancy between model and data is greatest. Alternatively, the solution pH and ionic

strength were both simply estimated from the known composition of the diafiltration buffer

used for the buffer exchange. This neglects the effects of excipient partitioning due to

Donnan effects during diafiltration of highly charged proteins at low ionic strength (Stoner et

al., 2004).

In addition to the selectivity, the separation performance is also determined by the

mass throughput, J∆S, where J is the filtrate flux and ∆S is the difference in the observed

sieving coefficients between the product (in this case the mono-pegylated protein) and the

impurity (the di-pegylated protein) (van Reis and Saksena, 1997). The J∆S values

corresponding to the experiments in Figure 6.2 are shown in Figure 6.3. J∆S attains its

maximum value of 19 L/m2/h at pH 5 due to the high transmission of the product associated

with the absence of any significant electrostatic exclusion of the mono-pegylated protein

under these conditions. J∆S decreases at pH both above and below pH 5 due to the strong

electrostatic interactions for both the mono- and di-pegylated species under these conditions.

Page 138: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

120

Figure 6.3 Mass throughput (J∆S) as a function of solution pH for ultrafiltration through a

300 kDa UltracelTM membrane charged for 24 hr. Data were obtained using 0.5

mM buffer at a filtrate flux of approximately 8 µm/s.

6.3.3 Diafiltration Experiments

Based on the experimental data and model calculations, a diafiltration process was

developed to separate the multiply pegylated proteins from the mono-pegylated α-

lactalbumin using a 300 kDa UltracelTM membrane that was charged for 24 hr. The

diafiltration was performed at pH 5 with the proteins dissolved in a 0.4 mM ionic strength

acetate buffer. Experimental results are shown in Figure 6.4 for the yield of each protein in

the collected filtrate solution:

Page 139: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

121

i

ff

VC

CVY = (3.2)

where fV is the cumulative volume of collected filtrate, fC is the average concentration of a

given species in the filtrate, V is the constant retentate volume, and iC is the initial

concentration in the feed solution. The results are plotted as a function of the number of

diavolumes (N), which is simply equal to the total collected filtrate volume divided by the

constant retentate volume. The yield of the mono-pegylated α-lactalbumin in the filtrate is

greater than 99% for N > 7, while less than 10% of the di-pegylated protein (and less than 3%

of the tri-pegylated species) pass into the filtrate even after 10 diavolumes.

The solid curves in Figure 6.4 are model calculations developed from solution of the

overall mass balance (van Reis and Saksena, 1997):

Y =1−exp(−NSo) (3.3)

where the protein sieving coefficient (So) is assumed to remain constant throughout the

diafiltration. This is consistent with the absence of any fouling during the diafiltration; the

difference in permeability values before and after the diafiltration was less than 5%. The best

fit values of So for the mono-, di-, and tri-pegylated α-lactalbumin were determined by

minimizing the sum of the squared residuals between the model and data with values

summarized in Table 2. The mono-pegylated α-lactalbumin shows minimal retention (So =

0.70 ± 0.01), while the di- and tri-pegylated forms both had greater than 99% retention,

consistent with the data shown previously. The model calculations are in excellent agreement

with the experimental data, properly capturing the accumulation of the mono-pegylated

protein in the filtrate solution as it is washed through the membrane and out of the stirred

cell.

Page 140: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

122

Figure 6.4 Yield for the mono-, di, and tri-pegylated α-lactalbumin in the filtrate solution as

a function of number of diavolumes for a diafiltration performed with a 300 kDa

UltracelTM membrane charged for 24 hr. Data were obtained at pH 5, 0.4 mM

ionic strength, and a filtrate flux of 8 µm/s. Solid curves are model calculations

The separation performance of the diafiltration process can be quantified in terms of

the tradeoff between the yield and purification factor (P) for the mono-pegylated α-

lactalbumin product, with results shown in Figure 6.5 using both the di- (filled circles) and

tri-pegylated protein (filled squares) as the impurity of interest. The purification factor is

defined as the ratio of the product yield in the filtrate to that of a given impurity. The solid

curves are model calculations for a diafiltration process with the product collected in the

filtrate solution (van Reis and Saksena, 1997):

Page 141: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

123

ψ/1)1(1 Y

YP

−−= (3.4)

where ψ is the selectivity between the mono-pegylated α-lactalbumin and the impurity. The

model calculations are in excellent agreement with the experimental results using the

selectivities evaluated using Equation (6.1) and the best-fit values of the sieving coefficients

given in Table 6.2.

Figure 6.5 Yield for the mono-pegylated α-lactalbumin as a function of purification factor.

Filled circles are for removal of the di-pegylated protein; filled squares are for

removal of the tri-pegylated protein. Solid and dashed curves are model

calculations.

Page 142: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

124

Table 6.2 Best-fit values of the protein sieving coefficients for the diafiltration process

Solute Observed Sieving Coefficient, So

Mono-pegylated 0.70 ± 0.01

Di-pegylated 0.0084 ± 0.0002

Tri-pegylated 0.0023 ± 0.0001

The diafiltration begins with zero product yield since the mono-pegylated α-

lactalbumin is fully contained in the retentate at the start of the process. The initial

(maximum) purification factor is simply equal to the selectivity, with Pmax = 83 ± 2 for

removal of the di-pegylated protein and Pmax = 300 ± 20 for removal of the tri-pegylated

species. The purification factor decreases with increasing product yield as the impurities

begin to leak through the membrane over the course of the diafiltration. The purification

factor with respect to the di-pegylated protein is greater than 20 with more than 97% product

yield and remains at P > 12 even after more than 99% yield of the mono-pegylated α-

lactalbumin. The purification factor with respect to the tri-pegylated protein is more than 50-

fold at 99% product yield, demonstrating that the diafiltration process is able to reduce the

concentration of the higher order pegylated species (beyond n = 2) to negligible levels.

Figure 6.6 shows the size exclusion chromatogram of the initial feed, the final

retentate, and the final filtrate solutions after the 10 diavolume process. The feed was

composed of 36% mono-pegylated α-lactalbumin with approximately 48% of the di-

pegylated species and 16% of the tri-pegylated protein. The protein concentrations were

evaluated directly from the area under the curves with the peaks simply cut at the location of

Page 143: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

125

the minimum. The peak associated with the mono-pegylated protein is undetectable in the

final retentate, while the tri-pegylated protein is nearly totally recovered (>99%). The final

filtrate was predominantly composed of the mono-pegylated α-lactalbumin which was

recovered with 97% yield; the peak for the tri-pegylated species was essentially undetectable

while that for the di-pegylated protein was only a small fraction of the feed concentration.

Note that the protein concentration in the final filtrate solution was diluted by the diafiltration

buffer; in this case the final concentration of the mono-pegylated α-lactalbumin in the filtrate

was 0.1 g/L after 6 diavolumes compared to 0.7 g/L in the initial feed. This dilution could be

eliminated using a cascade filtration system (van Reis and Zydney, 2007), or the filtrate

product could be re-concentrated using a second ultrafiltration step as part of the final

formulation.

Page 144: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

126

Figure 6.6 Size exclusion chromatograms showing the initial feed and the final retentate

(top panel) and the final filtrate (bottom panel) solutions after a 10-diavolume

diafiltration at pH 5 and 0.4 mM ionic strength.

Page 145: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

127

6.4 Conclusions

Although the early pegylated protein products were sold as mixtures of both singly

and multiply pegylated species, the pharmacokinetics and activity of these mixtures could be

highly variable due to differences in the properties of the differently pegylated proteins. The

purification of the desired mono-pegylated protein conjugate from the high order (multiply)

pegylated species is typically done using chromatography, although the low binding

capacities and poor resolution make this a challenging separation. The results presented in

this Chapter provide the first demonstration that it is possible to use ultrafiltration for the

separation of a mono-pegylated protein from the di- and tri- pegylated forms. A diafiltration

process using a negatively charged version of the 300 kDa UltracelTM membrane provided

greater than 95% yield of the mono-pegylated α-lactalbumin with more than 20-fold

purification by exploiting both the larger size and the greater electrostatic interactions of the

di-pegylated species.

The separation of the pegylated proteins was performed using a very low ionic

strength buffer (0.4 mM) to enhance the electrostatic exclusion of the di-pegylated species;

higher ionic strength solutions with greater buffering capacity could probably be employed

with membranes having a greater surface charge density or slightly smaller pore size to

obtain the desired retention of the di-pegylated protein. The process was operated at a filtrate

flux of 29 L/m2/h to minimize concentration polarization effects in the stirred cell due to the

relatively low mass transfer coefficient in this device; higher filtrate flux could be employed

in tangential flow filtration modules which have better mass transfer characteristics. These

tangential flow filtration devices are available in linearly scalable formats that should enable

operation at full manufacturing scale using existing membrane modules and systems (van

Page 146: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

128

Reis and Zydney, 2007). It is also important to note that the mono-pegylated α-lactalbumin

was obtained in the filtrate solution as a dilute product solution due to dilution by the

diafiltration buffer (final concentration about 6 times less than the feed). It would be possible

to eliminate this dilution effect by using a cascade ultrafiltration system (van Reis and

Zydney, 2007) or the product could be re-concentrated using a second ultrafiltration step as

part of the final product formulation.

Page 147: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

129

Chapter 7

Intermolecular Interactions during Ultrafiltration of Pegylated Proteins

Note: The material presented in this Chapter was adapted from: Ruanjaikaen, K.,

Zydney, A.L. Intermolecular interactions during ultrafiltration of pegylated

proteins. Biotechnology Progress (in press).

7.1 Introduction

There has been considerable interest in the use of membrane systems for the

purification and concentration of pegylated proteins (Mayolo-Deloisa et al., 2011).

Ultrafiltration has been used to concentrate a variety of pegylated proteins including α-

interferon (Arduini et al., 2004), human growth hormone (Clark et al. 1996), methioninase,

(Tan et al. 1998), and tumor necrosis factor receptor (Edwards et al., 2003). Diafiltration

processes have been used to remove small impurities and achieve the desired final

formulation (Stoner et al., 2004). Arpicco et al. (2002) used ultrafiltration to remove low

molecular weight (2 and 5 kDa) PEG from pegylated-gelonin, a ribosome inactivating

protein. More recently, Molek and Zydney (2007) demonstrated the feasibility of using a

two-stage ultrafiltration / diafiltration system for the removal of unreacted protein and PEG

from a pegylated α-lactalbumin.

The large majority of the published data on the ultrafiltration of pegylated proteins

were obtained at relatively low concentrations. Chavez and Orpiszewski (2005) reported a

reduction in lysozyme transmission in the presence of high concentrations of the pegylated

protein, which the authors attributed to the formation of a gel layer (i.e., fouling) on the

Page 148: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

130

surface of the membrane. However, it is also well known that there can be strong

intermolecular interactions between polyethylene glycol and proteins (Atha and Ingham,

1981), a phenomenon that has been exploited in the development of aqueous two phase

systems for protein separations (Albertsson, 1986).

The objective of the work described in this Chapter was to quantitatively evaluate the

effects of intermolecular interactions on the transmission of PEG, unreacted protein, and

pegylated proteins during ultrafiltration. Data were obtained using a model protein, α-

lactalbumin, with a molecular weight of 14.2 kDa, which was pegylated with a 20 kDa PEG.

7.2 Materials and Methods

7.2.1 Pegylated Proteins

Pegylated α-lactalbumins were prepared according to the procedures provided in

Chapter 3, using an initial α-lactalbumin concentration of 5 g/L with the activated PEG added

in approximately 1.5:1 molar ratio. The mono-pegylated α-lactalbumin was purified as

follows. First, the unreacted α-lactalbumin and N-hydroxysuccinimide were removed from

the reaction mixture by diafiltration through an unmodified 30 kDa ultrafiltration membrane

(Molek and Zydney 2007). The collected retentate was then diafiltered through a negatively

charged 300 kDa UltracelTM membrane at pH 8 with an ionic strength of 2 mM to remove the

neutral PEG. The more heavily pegylated species were then removed by diafiltration through

a negatively charged 300 kDa UltracelTM membrane at pH 5 and 0.5 mM acetate. The mono-

pegylated α-lactalbumin (collected in the filtrate) was then concentrated by ultrafiltration

through a 10 kDa UltracelTM membrane. The diafiltration processes were performed in 10

Page 149: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

131

mL or 50 mL Amicon stirred cells (EMD Millipore, Bedford, MA) using the procedures

described in Chapter 3.

The concentration of free PEG in the mixtures was adjusted by addition of either a

methoxy-PEG with 20 kDa nominal molecular weight and polydispersity of 1.05 (Creative

PEGWorks, catalog number PJK-202; Winston Salem, NC) or a PEG with 1.5 kDa nominal

molecular weight (Sigma-Aldrich, catalog number P5402; Saint Louis, MO). All solutions

were filtered through a 0.2 µm pore size Acrodisc syringe filter (Pall Corporation, Ann

Arbor, MI) to remove any insoluble materials before use.

7.2.2 Ultrafiltration Membranes

UltracelTM composite regenerated cellulose ultrafiltration membranes were provided

by EMD Millipore (Bedford, MA) with molecular weight cut-offs of 10, 30, and 300 kDa.

Unmodified and negatively charged membranes were prepared according to the procedures

provided in Chapter 3. The membrane hydraulic permeability (Lp) was evaluated before and

after each ultrafiltration experiment as a measure of membrane fouling using the procedures

discussed in Chapter 3.

7.2.3 Ultrafiltration Experiments

Sieving experiments were performed in a 25 mm diameter stirred ultrafiltration cell

(Amicon Model 8010, EMD Millipore, Bedford, MA). A membrane was placed in the

bottom of the stirred cell on top of a supporting Tyvek® disk. The filtrate line from the

stirred cell was connected to a peristaltic pump (Dynamax, Rainin Instrument Co., CA). A

minimum of 1.5 mL of filtrate was collected before each sieving measurement to remove the

Page 150: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

132

dead volume beneath the membrane and eliminate any transients associated with the change

in pressure. Small samples of the filtrate and retentate solutions were collected for

subsequent analysis by size exclusion chromatography.

The same system was used for the diafiltration experiments, with the stirred cell

connected to a solution reservoir containing protein-free diafiltration buffer. Filtrate samples

were collected periodically for subsequent analysis by size exclusion chromatography. A

retentate sample was obtained directly from the stirred cell at the end of the diafiltration

experiment to verify closure of the mass balance.

Batch ultrafiltration was performed using a similar system but with a 45 mm diameter

stirred ultrafiltration cell (Amicon Model 8050, EMD Millipore, Bedford, MA). The stirred

cell was initially filled with 50 mL of feed solution. The filtrate flux was adjusted by a

peristaltic pump connected to the filtrate line of the stirred cell. The concentration of the

retained solute in the stirred cell increases with time as the filtrate is removed. Solute

concentrations in the initial feed and filtrate samples were determined by size exclusion

chromatography as discussed in Chapter 3.

7.3. Results and Analysis

7.3.1 Sieving Behavior at Low Filtrate Flux

Typical experimental data for the transmission of α-lactalbumin, the 20 kDa PEG,

and the mono-pegylated α-lactalbumin are shown in Table 7.1. The experiments were

performed at a high ionic strength (200 mM) to minimize electrostatic interactions. The

observed sieving coefficients were evaluated as the ratio of the solute concentration in the

Page 151: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

133

filtrate to that in the bulk solution in the stirred cell. The first column shows results for the

individual species. The data for the α-lactalbumin and the 20 kDa PEG were obtained with

the pure components while results for the mono-pegylated α-lactalbumin were for the

purified product from the pegylation reaction; the concentration of the unreacted PEG and

protein were both less than 0.05 % (g / g) while that of the multiply pegylated species was

less than 2%. The sieving coefficient of the α-lactalbumin is more than two orders of

magnitude larger than that for the 20 kDa PEG and more than one order of magnitude larger

than that for the mono-pegylated protein, consistent with the large difference in size for these

species: 2.0 nm for the protein compared to 5.1 nm for the PEG and 5.2 nm for the pegylated

protein using the correlations presented by Fee and van Alstine (2004).

The sieving coefficient data in the mixtures provided in Table 7.1 were obtained

using size exclusion chromatography to evaluate the relative contributions of the individual

components in the feed and the filtrate solutions, with the details provided in Chapter 3. The

results at the low PEG concentration (0.4 g/L PEG) are nearly identical to those obtained

with the individual species, indicating that there are no intermolecular interactions at the low

concentration used in this experiment (total concentration < 4.0 g/L). In contrast, the sieving

coefficients at the high PEG concentration (23 g/L) were significantly greater than those for

the individual (purified) species. The observed sieving coefficient for the α-lactalbumin at

high PEG concentration was actually slightly greater than one, with a value of So = 1.4,

corresponding to a “negative” rejection. The origin of this behavior is discussed in more

detail subsequently. The high PEG concentration caused more than an order of magnitude

increase in the transmission of both the 20 kDa PEG and the mono-pegylated α-lactalbumin.

These changes in the sieving coefficients were not due to any irreversible changes in the

Page 152: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

134

membrane; the hydraulic permeability values of the 30 kDa membrane before and after the

ultrafiltration experiments were within ±10%.

Table 7.1 Sieving coefficients of the unmodified α-lactalbumin, the 20 kDa PEG, and the

mono-pegylated α-lactalbumin alone and in mixtures with low (0.4 g/L) and high

(23 g/L) PEG concentrations. Data were obtained at a filtrate flux of Jv ≈ 2.3

µm/s in a pH 7, 200 mM ionic strength buffer using an unmodified 30 kDa

UltracelTM membrane.

Radius(nm) Individual

Mixture

(0.4 g/L PEG)

Mixture

(23 g/L PEG)

α-lactalbumin 2.0 0.89 ± 0.01 0.85 ± 0.01 1.4 ± 0.1

20 kDa PEG 5.1 0.02±0.02 0.01 ± 0.02 0.15 ± 0.02

Mono-pegylated 5.2 0.005±0.003 0.003 ± 0.003 0.074 ± 0.003

We hypothesized that the increase in observed sieving coefficients in the presence of

a high concentration of PEG was due to an increase in chemical potential associated with

intermolecular interactions between the PEG and other components. The solute

concentrations just inside the pore are assumed to be in equilibrium with the concentrations

in the solution immediately exterior to the membrane (Deen, 1987), with the equilibrium

partition coefficient written as:

( ) KC

C

wall

pore 21 λφ −==

(7.1)

Page 153: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

135

where Cpore and Cwall are the average solute concentrations inside and outside the pores. The

first term on the right hand side of Equation (2) describes the steric exclusion of the solute

from the region within one solute radius of the pore wall with λ equal to the ratio of the solute

(rs) to pore (rp) radii. The second term (K) accounts for the difference in the chemical

potential associated with the difference in PEG concentration between the solution inside and

outside of the pores.

Previous studies of PEG-protein interactions in aqueous two phase systems indicate

that the partition coefficient (K) for a macromolecule i between the PEG-rich and PEG-poor

phases can be approximated as (King et al., 1988):

( )[ ]pore

PEG

wall

PEGPEGi CCbK −= −exp

(7.2)

where bij is the interaction parameter. In this case, the PEG-rich phase is the solution outside

the pores while the PEG-poor phase is the dilute solution within the membrane pores. The

protein sieving coefficient can be evaluated by combining Equations (7.1) and (7.2) to give:

( )[ ]fPEGwPEGPEGiao

wi

fiCCbS

C

C,,

,

,exp −= − (7.3)

where the solute concentrations in the pore adjacent to the upstream surface of the membrane

has been approximated using the filtrate concentrations ( fiC , and fPEGC , ) (Zeman and

Zydney, 1996). Sao is the value of the actual sieving coefficient at infinite dilution where

intermolecular interactions are unimportant, i.e., where K=1.

Page 154: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

136

Figure 7.1 Observed sieving coefficients of the 20 kDa PEG, α-lactalbumin, and the mono-

pegylated α-lactalbumin as a function of the difference in PEG concentrations

between the bulk and filtrate solutions. Data obtained at a filtrate flux of 2.3

µm/s in a 200 mM ionic strength solution at pH 7 using an unmodified UltracelTM

30 kDa membrane. Dashed lines are linear regression fits. Solid curves are

model calculations discussed in more detail subsequently.

Figure 7.1 shows data for the observed sieving coefficients (So), defined as the ratio

of the solute concentration in the filtrate to that in the bulk solution, for α-lactalbumin, the 20

kDa PEG, and the mono-pegylated α-lactalbumin (in ternary mixtures) over a range of PEG

concentrations. The bulk concentrations of α-lactalbumin and the mono-pegylated protein

were approximately 0.7 and 1.7 g/L, respectively (based on the total molecular weight of the

pegylated species). The data represent results from 5 separate experiments, with the

concentration of the 3 species at each PEG concentration determined by size exclusion

Page 155: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

137

chromatography. The results are plotted as an explicit function of the difference in PEG

concentration between the bulk and pore solutions based on the form given by Equation (7.3).

This assumes that concentration polarization effects are small at the low filtrate flux used in

these experiments (Jv = 2.3 µm/s), i.e., that the PEG concentration at the membrane surface

( wPEGC , ) is essentially equal to that in the bulk solution ( bPEGC , ):

( ) bPEGPEGofPEGwPEG CSCC ,,,, 1−≈−

(7.4)

The observed sieving coefficients increase almost linearly (on the semi-log scale) with

increasing PEG concentration, consistent with the form given by Equation (7.3). The net

result is that the sieving coefficient for α-lactalbumin is somewhat larger than unity for PEG

concentrations greater than 10 g/L. This negative rejection is due to the increase in chemical

potential of the α-lactalbumin in the feed solution, with the large value for K more than

compensating for the steric exclusion in Equation (7.1). Note that the observed sieving

coefficient is approximately equal to the partition coefficient since the hydrodynamic

hindrance factor for convection is only a weak function of the solute size (Zeman and

Zydney, 1996).

The interaction parameters (bij) were evaluated directly from the slopes of the linear

regression fits to the data in Figure 7.1 (dashed lines). The calculated values of bij for the 20

kDa PEG (bPEG-PEG = 0.16 ± 0.02 L/g) and the mono-pegylated protein (bPEG-PEG1 = 0.17 ±

0.01 L/g) were both significantly greater than that for the unmodified α-lactalbumin (bPEG-αlac

= 0.027 ± 0.001 L/g), indicative of stronger intermolecular interactions. The value of the

PEG-PEG interaction parameter is slightly larger than the value calculated from the second

virial coefficient for a 23 kDa PEG determined from laser-light scattering data presented by

Page 156: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

138

Hasse et al. (1995) (bPEG-PEG = 0.11 L/g), where bii was evaluated using the following

expression (King et al., 1988) :

2Bii =1, 000bii

M i

(7.5)

where Mi is the solute (PEG) molecular weight (in g/mol), bii is in units of L/g, and Bii is in

units of mL mol/g2.

Corresponding data for the observed sieving coefficients in the presence of a low

molecular weight (1.5 kDa) PEG are shown in Figure 7.2. The data for the pegylated protein

were for α-lactalbumin pegylated with a 20 kDa PEG (in the presence of added 1.5 kDa

PEG). The sieving coefficient of the small 1.5 kDa PEG was approximately equal to one

under all conditions, thus wPEGC , was approximately equal to fPEGC , in these experiments.

The net result was that the sieving coefficient of α-lactalbumin remained essentially constant;

there was no longer any “negative” rejection of the protein at high PEG concentrations. The

solid lines in Figure 2 are the predicted values of the observed sieving coefficients given by

Equations (7.2) to (7.5) using the previously determined values of Bij . The steric exclusion

terms for α-lactalbumin and the mono-pegylated α-lactalbumin were determined from the y-

intercepts in Figure 7.1, and the observed sieving coefficient of the 1.5 kDa PEG was

assumed to be So = 0.9. The model predicts a 39% increase in the sieving coefficient of the

mono-pegylated protein over this range of PEG concentrations while the data show close to a

3.2-fold increase, although there is considerable uncertainty in this number given the inherent

errors at the very low filtrate concentrations of the mono-pegylated α-lactalbumin. Note that

it would be possible to fit the data for the mono-pegylated protein using bPEG-PEG1 = 0.48 L/g,

Page 157: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

139

although there was no independent evidence for this large an interaction parameter for

interactions between a 1.5 kDa PEG and α-lactalbumin pegylated with a 20 kDa PEG.

Figure 7.2 Observed sieving coefficients of a 1.5 kDa PEG, α-lactalbumin, and the mono-

pegylated α-lactalbumin (with a 20 kDa PEG) as a function of the PEG

concentration difference between bulk and filtrate solutions for a low molecular

weight (1.5 kDa) PEG. Data obtained at a filtrate flux of 2.3 µm/s using a pH 7,

200 mM ionic strength buffer with an unmodified 30 kDa UltracelTM membrane.

Solid lines are model calculations for α-lactalbumin, and the mono-pegylated α-

lactalbumin as described in text.

7.3.2 Concentration Polarization Effects

7.3.2.1 PEG-PEG Interactions

Figure 7.3 shows experimental data for the observed sieving coefficient for a purified

20 kDa PEG as a function of the filtrate flux. Data were obtained at both low (1.2 g/L) and

Page 158: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

140

high (14 g/L) PEG concentrations. The large increase in the sieving coefficient with

increasing filtrate flux is due to concentration polarization effects, i.e., the accumulation of a

highly concentrated region of retained species at the membrane surface (as discussed

previously in Chapter 2). The data also show a strong dependence on the PEG concentration,

with the sieving coefficients in the more concentrated solution lying well above those for the

dilute solution, particularly at low filtrate flux.

Figure 7.3 Observed sieving coefficient of a 20 kDa PEG as a function of filtrate flux at both

low (1.2 g/L) and high (14 g/L) PEG concentrations in a pH 7 and 10 mM ionic

strength buffer using an unmodified 30 kDa UltracelTM membrane. The dashed

curves are model calculations using the classical concentration polarization model

while the solid curves are those using the modified concentration polarization

model as described in the text.

Page 159: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

141

The dashed curves in Figure 7.3 are model calculations using the classical

concentration polarization model presented in Equation (2.3), shown here again for

convenience:

So =

Sa expJv

km

1− Sa( ) + Sa expJv

km

(7.6)

where km is the bulk mass transfer coefficient in the stirred cell, and So and Sa are the

observed (Cf/Cb) and actual (Cf/Cw) sieving coefficients where Cf, Cb, and Cw are the solute

concentrations in the filtrate solution, in the bulk solution, and at the membrane surface on

the retentate side (wall concentration), respectively. The dashed curves were generated with

Sao = 10-5 with km = 3.9 x 10-6 for the 1.2 g/L PEG solution and km = 2.4 x 10-6 m/s for the 14

g/L PEG solution, where the km values were determined from the correlation for mass transfer

in a stirred cell given in Chapter 2. The model is in very poor agreement with the data,

significantly over-predicting the degree of concentration polarization. It was possible to

obtain somewhat better fits by using a much smaller value of the bulk mass transfer

coefficient, but no combination of km and Sao was able to provide a good fit to the

experimental results.

The reason for the discrepancy between the data and the simple polarization model is

likely due to the effects of PEG-PEG interactions on the extent of concentration polarization.

The solute flux (Ns) in the bulk solution can be expressed as the sum of the convective and

diffusive flux as:

Ns = JvC −DC

RT

dµdz

(7.7)

Page 160: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

142

where µ is the local chemical potential, R is the ideal gas constant, T is the absolute

temperature, and D is the solute diffusion coefficient in an infinitely dilute solution.

The chemical potential of the PEG can be approximated as:

µ = µo + RT lnCPEG + RTbPEG−PEGCPEG (7.8)

where µo is the chemical potential at a reference state and bPEG-PEG is the interaction

parameter. Substitution of Equation (7.8) into Equation (7.7) yields the following one-

dimensional bulk transport equation:

JvCPEG − DPEG (1+ bPEG−PEGCPEG )dCPEG

dz= JvCPEG, f

(7.9)

where the solute flux has been set equal to the solute flux through the membrane (the product

of the filtrate flux and the filtrate concentration). Equation (7.9) can be integrated across the

concentration boundary layer yielding:

Jv

km

= (1+ bPEG−PEGCPEG, f )lnCPEG,w − CPEG, f

CPEG,b − CPEG, f

+ bPEG−PEG (CPEG,w − CPEG,b )

(7.10)

where the ratio of the diffusion coefficient to the boundary layer thickness has been set equal

to the mass transfer coefficient (km). Equation (7.10) was derived previously by Zydney

(1992) to describe protein sieving at high concentrations; it reduces to the form given by

Equation (7.6) in the limit of bPEG-PEG = 0 or very dilute PEG concentrations.

The solid curves in Figure 7.3 were developed by simultaneous solution of Equations

(7.3) and (7.10) for Cf and Cw for each value of the filtrate flux and bulk PEG concentration

using km = 3.9 x 10-6 (for the 1.2 g/L PEG solution) and km = 2.4 x 10-6 m/s (for the 14 g/L

PEG solution) as determined previously. The model calculations are in good agreement with

the experimental data at both low and high PEG concentrations using Sao = 0.00075 and bPEG-

PEG = 0.11 L/g, where the interaction parameter has been evaluated from literature data for a

Page 161: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

143

23 kDa PEG (Hasse et al., 1995). This value of the interaction parameter was also used to

generate the solid curve for the 20 kDa PEG in Figure 7.1. The model accurately captures

the observed effects of both the filtrate flux and bulk PEG concentration on the observed

sieving coefficients for the 20 kDa PEG. Note that the extent of concentration polarization in

Figure 7.1 was small but not insignificant, giving rise to the small differences between the

solid and dashed curves.

7.3.2.2 PEG-PEG1 Interactions

Figure 7.4 shows the effects of the filtrate flux on the observed sieving coefficient of

the mono-pegylated α-lactalbumin at both low (1.2 g/L) and high (14 g/L) concentrations of

the 20 kDa PEG. The sieving coefficient for the pegylated protein increases with increasing

filtrate flux due to the effects of concentration polarization, with the larger sieving

coefficients at the high PEG concentrations again due to the effects of intermolecular

interactions between the pegylated protein and the PEG.

Page 162: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

144

Figure 7.4 Observed sieving coefficients of the mono-pegylated α-lactalbumin as a function

of filtrate flux at both low (1.2 g/L) and high (14 g/L) concentrations of the 20

kDa PEG in a pH 7 and 10 mM ionic strength buffer using an unmodified 30 kDa

UltracelTM membrane. The solid curves are the numerical solution to the full

model. The dashed curves are an approximate solution as described in the text.

The solid curves in Figure 7.4 are model calculations developed using the same basic

approach as that used to describe the PEG-PEG interactions. However, in this case the local

chemical potential of the mono-pegylated protein is assumed to depend on the concentration

of the PEG (CPEG) in addition to that of the pegylated protein (C):

PEGPEGPEGo CRTbCRT 1ln −++= µµ

(7.11)

Substitution of Equation (7.11) into Equation (7.9) yields:

fvPEG

PEGPEGv CJdz

dCCb

dz

dCDCJ =

+− − 1

(7.12)

Page 163: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

145

Equation (7.12) cannot be solved analytically due to the non-linearity of the term involving

the interactions between the PEG and the pegylated protein. Instead, Equation (7.12) was

solved numerically with the actual sieving coefficient of the mono-pegylated protein given by

Equation (7.3). The model calculations are in good agreement with the experimental data at

both PEG concentrations using bPEG-PEG1 = 0.11 L/g (assumed to be equal to the value of bPEG-

PEG) with the best fit value of Sao = 0.0004. The small discrepancies between the data and

the solid curves could be due to the use of constant values of Sao, which neglects the possible

elongation of the grafted PEG chains at high filtrate flux as described by Molek and Zydney

(2006). Although it would be possible to include the steric effects provided by the

deformation/elongation in the sieving model (Davidson et al., 1986; Morao et al., 2011),

there is no obvious way to include both steric and electrostatic effects provided by the

elongation/deformation since the latter could be very complex for a non-spherical (deformed)

solute.

The same value of the interaction parameter was also used to generate the solid curve

for the mono-pegylated α-lactalbumin in Figure 7.1. Again, the small differences between

the solid and dashed curves were due to the small degree of concentration polarization. The

solid curve for the α-lactalbumin in Figure 7.1 was generated using the same approach, but

with bPEG-αlac = 0.0027 L/g based on the slope of the linear regression in Figure 7.1. The

model also accurately predicts the values of the observed sieving coefficient greater than one

(negative rejection) seen at high PEG concentrations.

Although the full numerical solution is in good agreement with the data, the required

iterative analysis is awkward for design calculations. Thus, an approximate solution to

Equation (7.12) was developed based on the similarity between Equations (7.12) and (7.9) by

Page 164: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

146

assuming that the concentration of the pegylated protein was proportional to the

concentration of the PEG:

C − Cb

Cw − Cb

=CPEG −CPEG,b

CPEG,w −CPEG,b

(7.13)

Equation (7.13) can be substituted into Equation (7.12) and directly integrated to give:

Jv

km

= 1+ bPEG−PEG1

CPEG,w −CPEG,b

Cw −Cb

C f

ln

Cw −C f

Cb −C f

+ bPEG−PEG1(CPEG,w −CPEG,b )

(7.14)

where Cf, Cb, and Cw are the concentrations of the mono-pegylated protein in the filtrate

solution, in the bulk solution, and at the membrane surface on the retentate side (wall

concentration). The dashed curves in Figure 7.4 are given by Equation (7.14); this analytical

solution is in good agreement with both the data and the full numerical calculations over the

full range of experimental conditions.

7.3.3 Diafiltration Process - PEG Removal

The intermolecular interactions discussed in the previous sections can have a

significant impact on the design and operation of a diafiltration process designed to remove

excess PEG from the reaction mixture after protein pegylation. This is particularly true when

large excesses of activated PEG are used to maximize the yield of the target pegylated

protein (Basu et al., 2006). This behavior was examined experimentally using a 1 g/L

solution of the mono-pegylated α-lactalbumin containing 15 g/L of the 20 kDa PEG. The

diafiltration was performed using a negatively-charged version of the 300 kDa UltracelTM

membrane at low ionic strength (2 mM) to obtain high retention of the negatively-charged

pegylated protein while allowing good transmission of the electrically neutral PEG.

Page 165: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

147

Experimental data for the normalized concentrations of the 20 kDa PEG and the

mono-pegylated α-lactalbumin in the retentate solution are shown in Figure 7.5 as a function

of the number of diavolumes (N), defined as the ratio of the total collected filtrate volume to

the constant feed volume in the stirred cell. The PEG concentration in the retentate drops

rapidly at the beginning of the diafiltration, but then decreases relatively slowly towards the

end of the process. This behavior is a direct result of the reduction in the PEG sieving

coefficient with decreasing PEG concentration arising from the reduction in intermolecular

interactions.

Figure 7.5 Normalized concentrations of the mono-pegylated α-lactalbumin and the 20 kDa

PEG as a function of the number of diavolumes for a diafiltration performed with

a negatively charged 300 kDa Ultracel membrane at pH 8, 2 mM ionic strength,

and a filtrate flux of 8 um/s. Solid and dashed curves are model calculations as

described in the text.

Page 166: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

148

The solid and dashed curves in Figure 7.5 represent model calculations developed

from a simple mass balance:

( ) fifi CQVCdt

d,−= (7.15)

where Ci and Ci,f are the solute concentrations in the retentate and filtrate at any point during

the diafiltration, Qf is the volumetric filtrate flow rate, and V is the retentate volume. The

dashed curves are developed assuming a constant retentate volume and a constant sieving

coefficient:

]exp[ ,io

io

i NSC

C−=

(7.16)

where ioC is the initial solute concentration in the retentate and N is the number of

diavolumes defined as the ratio of the cumulative filtrate volume to the constant retentate

volume during the diafiltration process. The sieving coefficients for the PEG and pegylated

α-lactalbumin were evaluated experimentally just before the start of the diafiltration as So,PEG

= 0.75 and So,PEG1 = 0.024. The solid curves were developed by numerical integration of

Equation (7.15) with the sieving coefficients evaluated as a function of the time-dependent

concentrations of both the pegylated protein and the PEG. The interaction parameters were

determined based on the results in Section 7.3.2.1 as bPEG-PEG = bPEG-PEG1 = 0.11 L/g, with the

sieving coefficients at infinite dilution determined from a fit to the experimental data as

Sao,PEG = 0.060 and Sao,PEG1 = 0.0008. The results from the numerical solution are in good

agreement with the experimental data over the entire diafiltration, properly capturing the

reduction in the rate of PEG removal (and the increase in the retention of the pegylated

protein) as the PEG is removed. Model calculations indicate that the PEG concentration

Page 167: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

149

could be reduced to 0.1% of its initial value after 17 diavolumes, which is nearly twice the

value of N = 9.2 given by Equation (7.16).

7.3.4 Batch Ultrafiltration

Intermolecular interactions can also have a significant effect on the behavior of

ultrafiltration systems designed for the concentration of pegylated proteins. Figure 7.6 shows

results for the batch ultrafiltration of a mono-pegylated α-lactalbumin with an initial

concentration of 2.5 g/L. Data were obtained using a 10 kDa UltracelTM membrane at a

filtrate flux of 10 µm/s, with the protein dissolved in a pH 7 buffer containing 1 mM bis-Tris

and 10 mM NaCl. The data for product loss in the filtrate (Yloss) are plotted as a function of

the volume concentration factor (VCF), with VCF = Vo/V where V is the retentate volume at

any given time during the batch ultrafiltration and Vo is the initial retentate (feed) volume.

The pegylated α-lactalbumin was strongly retained at the start of the ultrafiltration, with a

very low sieving coefficient of 0.0021. At the beginning of the process, the product loss in

the filtrate increased slowly and linearly with log(VCF), consistent with a constant value of

the sieving coefficient. However, for VCF > 6, the rate of product loss increased

significantly due to the increase in the value of the sieving coefficient of the pegylated

protein arising from the strong intermolecular interactions in the increasingly concentrated

solution of the pegylated protein.

Page 168: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

150

Figure 7.6 Filtrate product loss of mono-pegylated α-lactalbumin as a function of volume

concentration factor (VCF) at a filtrate flux of 10 µm/s. Data obtained in a pH 7

and 10 mM ionic strength buffer using an unmodified 10 kDa UltracelTM

membrane. Solid and dashed curves are model calculations as described in the

text.

The solid curve in Figure 7.6 represents the model calculation developed by

numerical integration of Equation (7.15) with the filtrate concentration given by Equation

(7.3) using bPEG1-PEG1 = 0.11 L/g and the best fit value of Sao =7.0 x10-5. The model

calculations are in good agreement with the data, properly capturing the increase in product

loss at high concentration factors. The dashed curve in Figure 7.6 is the classical analytical

solution for the batch ultrafiltration developed assuming a constant sieving coefficient

(Zeman and Zydney 1996):

oS

loss VCFY−−= 1

(7.17)

Page 169: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

151

Equation (7.17) is in good agreement with the data at low concentration factors, but

significantly under-predicts the yield loss (and thus over-predicts the yield) as the pegylated

protein becomes more heavily concentrated.

Figure 7.7 Filtrate product loss of mono-pegylated α-lactalbumin as a function of volume

concentration factor (VCF) at a filtrate flux of 10 µm/s. Data obtained in a pH 7

and 10 mM ionic strength buffer using an unmodified 30 kDa UltracelTM

membrane. Solid and dashed curves are model calculations as described in the

text.

Figure 7.7 shows results for the batch ultrafiltration of a mono-pegylated α-

lactalbumin using 30 kDa UltracelTM membrane (in contrast to the 10 kDa membrane used in

Figure 7.6) at a filtrate flux of 10 µm/s, with an initial protein concentration of 2.5 g/L in a

pH 7 buffer containing 1 mM bis-Tris and 10 mM NaCl. The product loss in the filtrate

increased linearly with log(VCF) at the beginning of the process. However, the data began to

Page 170: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

152

deviate from the prediction using a constant sieving coefficient (the dashed line with

So=0.004) as soon as VCF > 2 due to the somewhat larger pore size of the 30 kDa membrane.

The solid curve in Figure 7.7 represents the model calculations developed by numerical

integration of Equation (7.15) with the filtrate concentration given by Equation (7.3) using

bPEG1-PEG1 = 0.11 L/g and the best fit value of Sao =5.0 x10-4. The model is again in fairly

good agreement with the data, properly capturing the significant increase in product loss at

large numbers of diavolumes.

Figure 7.8 shows model calculations for the effects of the product of the initial

concentration of the mono-pegylated protein and the interaction parameter on the product

loss in the filtrate. The calculations were performed using Equation (7.15) and (7.3) with Sao

= 10-4 and Jv/km= 3. The product loss increases with increasing values of bPEG1-PEG1Co,PEG as

expected, with more than 10% loss after only 15 diavolumes for bPEG1-PEG1Co,PEG = 0.5. The

dashed line is the calculated results for b=0, i.e. in the absence of intermolecular interactions.

The product loss remains less than 1% under these conditions out to more than 50

diavolumes.

Page 171: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

153

Figure 7.8 Calculated filtrate product loss of mono-pegylated α-lactalbumin as a function of

volume concentration factor (VCF). Model calculations were performed using

Jv/km = 3 with Sao = 10-4. Solid and dashed curves are model calculations as

described in the text.

7.4 Conclusions

There is considerable interest in using membrane systems for the purification and

formulation of pegylated therapeutics. The data presented in this Chapter clearly

demonstrate that the presence of free PEG significantly increases the transmission of both the

PEG itself and any other high molecular weight species due to the increase in free energy

associated with the strong intermolecular interactions. For example, the sieving coefficient

Page 172: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

154

of the 20 kDa PEG and the mono-pegylated α-lactalbumin both increased by well over an

order of magnitude in the presence of a fairly high concentration of the 20 kDa PEG (23 g/L).

Simple theoretical models were developed for the effects of these intermolecular

interactions based on the increase in the local chemical potential of the PEG / protein on both

the sieving coefficient and the local diffusive flux. Model calculations are in good agreement

with experimental data over a range of bulk PEG concentrations and filtrate flux. In addition,

the model accurately predicts the negative rejection observed for the small α-lactalbumin at

high PEG concentrations due to the increase in the equilibrium partition coefficient.

The strong intermolecular interactions for the PEG / pegylated proteins can have a

significant impact on ultrafiltration processes for the purification and formulation of

pegylated therapeutics. In particular, the reduction in intermolecular interactions during the

course of a diafiltration process designed to remove unreacted PEG leads to a significant

reduction in the rate of PEG removal, requiring significantly more diavolumes to achieve the

same target purity. In contrast, the increase in intermolecular interactions during a batch

ultrafiltration process increases the transmission of the pegylated product leading to a

reduction in the overall product yield. This latter effect becomes increasingly significant at

the high degrees of volume reduction used to obtain highly concentrated final formulations.

The experimental results and theoretical models presented in this Chapter provide an

appropriate framework to calculate the magnitude of these phenomena and to develop

methods to optimize the performance of membrane processes for production of these

pegylated products.

Page 173: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

155

Chapter 8

Combined reaction and membrane-based separation process for

enhanced yield of protein conjugates

8.1 Introduction

As discussed in Chapter 1, one of the challenges in producing a protein – polymer

conjugate is generating a high yield of the desired conjugate, which typically involves the

attachment of only a single polymer chain to each protein. For example, the maximum yield

of a mono-pegylated protein reported by Gao et al. (2009) and Piquet et al. (2002) was only

slightly greater than 50%. Attempts to drive the reaction forward, e.g., by the use of higher

concentrations of the activated PEG, led to the formation of multiply-pegylated products that

had to be removed in a subsequent purification step. Chavez and Orpiszewski (2004) used a

sequential reaction – separation process to increase the yield of a mono-pegylated lysozyme

with slightly improved product yield. Several attempts have been made to develop

chromatographic processes in which the reaction and separation occur simultaneously in a

single column (Fee, 2003; Milunović et al., 2012). However, the final yield of mono-

pegylated product in these systems was still fairly low, and the chromatographic processes

provided relatively low throughput.

The objective of the studies described in this Chapter was to develop a combined

reaction and membrane-based separation process for the enhanced yield of a desired protein –

polymer conjugate. The next section describes a simple mathematical model for the

production of a mono-pegylated protein in the reaction-separation system. The model

Page 174: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

156

calculations are then verified using an experimental system involving pegylation of α-

lactalbumin. The results clearly demonstrate the feasibility of this combined reaction –

separation system and provide a framework for the design of novel processes for the

production of desired protein conjugates with enhanced yield.

8.2 Reaction--Separation System

Figure 8.1 shows a schematic of a combined reaction and membrane-based separation

system for the production of a desired protein conjugate. The outflow from the reactor is

continuously fed directly to a tangential flow ultrafiltration module, with the permeate

containing the reactants recycled back to the reactor while the retentate is collected in a

separate product tank. Since the conversion in the UF module is relatively low, the residual

reactants that enter the product tank are continuously pumped through the UF module in a

second recycle loop. Activated PEG is continuously added to the reactor to maintain a

relatively low ratio of PEG to native protein in the reactor to minimize the formation of

multiply-pegylated species (Fee and Van Alstine, 2006).

Page 175: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

157

Figure 8.1 Schematic of the reaction and membrane-based separation system

The concentrations of the products and reactants in the reaction – separation system

can be evaluated from simple mass balances on the reactor, module, and product tank:

Reactor:

Module:

Product tank

where , ,

and Ci,feed are the concentrations of solute i in the reactor, the

membrane module, product tank, and feed, respectively, with VR, VM, and Vp the volumes,

RPRiMRMiiofeedfeediRRi

RiRqCqCSqCVr

dt

CVd,,,,,

, )(−++= )1.8(

MPMiMRMiioPMPiMMi

MiMqCqCSqCVr

dt

CVd,,,,,

, )(−−+= )2.8(

MPMiPMPiRPRiPPi

PiPqCqCqCVr

dt

CVd,,,,

, )(+−+= )3.8(

RiC , MiC , PiC ,

Page 176: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

158

and qRM, qMR, qR and qfeed the volumetric flow rates. is the observed sieving coefficient

for solute i, which is equal to the ratio of the filtrate concentration to the concentration

entering the module. ri,R and ri,M are the reaction (generation) rates for species i in the

reactor and the module, respectively. The reactions are assumed to be bimolecular; thus, the

net rate of production for each species is given as:

(8.5)

(8.6)

(8.7)

(8.8)

where C represents the concentration of each species, and kn represents the rate constant in a

particular step of the reaction. The subscripts A and PEG are for the native α-lactalbumin and

the activated PEG, respectively. The subscripts P1, P2, and P3 represent the mono-, di-, and

tri-pegylated species, respectively. The conversion of the tri-pegylated species into higher

order pegylated products was not considered since the concentration of P3 remains quite low

throughout the process; an additional term could easily be included in Equation (8.6) to

account for that reaction if desired. The activated PEG is assumed to degrade by a first-order

hydrolysis (shown as the last term in Equation 8.8). The model equations were solved

numerically using Mathematica.

ioS ,

PEGPPEGAPEG CCkCCkr 1211 −= )4.8(

PEGPPEGPPEG CCkCCkr 23122 −=

PEGPPEG CCkr 233 =

PEGAA CCkr 1−=

PEGhPEGPPEGPPEGAPEG CkCCkCCkCCkr −−−−= 23121

Page 177: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

159

8.3 Materials and Methods

Experimental studies were performed with α-lactalbumin obtained from Sigma

Chemicals (St. Louis, MO). Pegylation was performed by reaction of the protein with N-

hydroxysuccinimide activated PEG with nominal molecular weight of 20 kDa (Catalog

number ME-200HS; NOF Corporation, Tokyo, Japan). In order to determine the

concentration of each species during the pegylation reaction, 100 µL samples were taken

periodically from the stirred reactor and product tank and mixed with 200 µL of 0.2 M HCl to

rapidly hydrolyze the activated NHS group on the PEG, thereby quenching the pegylation

reaction. The concentration of each species was determined by size exclusion

chromatography as described in Chapter 3.

The hydrolysis rate constant for the activated PEG (kh) was determined independently

by measuring the formation rate of the hydrolyzed NHS group, assuming a first-order

hydrolysis reaction. A small amount of activated PEG (approximately 0.02 g) was dissolved

in a buffer of interest (3 mL). The concentration of hydrolyzed NHS group as a function of

time was measured by UV-Vis spectrophotomety based on the absorbance at 260 nm

(Miron and Wilchek, 1982) using a SPECTRAMAX® PLUS384 (Molecular Devices Corp,

Sunnyvale CA).

A Pellicon XLTM tangential flow filtration module with 50 cm2 of 30 kDa UltracelTM

membrane (EMD Millipore, Bedford, MA) was used for the collection of the mono-pegylated

product (Figure 8.2). The unmodified 30 kDa membrane provides high retention of the

pegylated species and the PEG, with the unreacted protein obtained in the permeate (Molek

and Zydney, 2006). Masterflex® tubing was connected to the feed, retentate, and outlet

permeate port (next to the retentate port). The other permeate port (next to the feed inlet) was

Page 178: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

160

sealed with a barb fitting. The feed line was connected to the output line of a Masterflex

peristaltic pump (Model 7550-60; Cole-Parmer, Chicago, IL); the suction line of the pump

was connected to the feed reservoir. The membrane was initially flushed with deionized

water using a feed flow rate of 30 mL/min until at least 300 mL of water was collected from

the retentate outlet and 150 mL from the permeate outlet. The same procedure was

performed using appropriate buffer solution to precondition the membrane module prior to

the filtration experiment. After the experiments, the membrane module was flushed with

deionized water and stored at 4 oC.

Figure 8.2 Schematic of Pellicon XLTM tangential flow filtration module (image provided by

Millipore Corp.).

Page 179: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

161

8.4 Results and Discussions

8.4.1 Batch Pegylation

Figure 8.3 shows results from a batch pegylation reaction (no separation) performed

in a 1 mM bis-Tris buffer at pH 7. The reactor was initially charged with 5.0 g/L (0.35 mM)

of α-lactalbumin along with a 20 kDa PEG-NHS in a 1:1 molar ratio with the protein. The

concentrations for both reactants decreased with time as expected; the greater reduction in the

concentration of the PEG-NHS is due to the formation of the multiply-pegylated species

(denoted as PEG2 and PEG3) and the hydrolysis reaction. The final yield of the mono-

pegylated protein (PEG1) after 180 min was 53% (based on the conversion of α-lactalbumin),

with 18% of the protein present in multiply pegylated forms and 29% still unreacted. The

solid and dashed curves in Figure 8.3 are the model calculations developed by solving the

batch reaction equations (Equation 8.1 without any of the flow terms along with Equations

8.4 to 8.8), with the best fit values of the rate constants determined by minimizing the sum of

the squared residuals between the model and data (values given in Table 1 – the rate constant

for the hydrolysis reaction was determined separately from an experiment with the activated

PEG alone). The model calculations are in excellent agreement with the experimental results

for all species, providing an appropriate framework for the analysis of the combined

reaction–separation process.

Page 180: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

162

Figure 8.3 Concentration of α-lactalbumin, 20 kDa PEG, and the differently pegylated α-

lactalbumins as a function of time for a batch reaction at pH 7. Curves are model

calculations as described in the text.

Batch pegylation reactions were also performed at pH 4, 5, 6, and 8, with the best fit

values of the rate constants also shown in Table 8.1. The rate constants for the pegylation

reactions (k1, k2, and k3) increase with increasing pH due to the increase in nucleophilicity of

the lysine amino groups at a higher pH (Roberts et al., 2002). In contrast, the rate constant

for the PEG hydrolysis is greatest at pH 4, with a minimum value achieved around pH 5.

There was no evidence of any formation of the di- or tri-pegylated proteins at pH 4, although

the small value of k1 and the large value of kh would require extended reaction times and very

large quantities of PEG-NHS to effectively conduct the pegylation reaction under these

conditions.

Page 181: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

163

Table 8.1 Rate constants for the pegylation reaction of α-lactalbumin with 20 kDa PEG-

NHS in 1 mM bis-Tris (pH 6, 7, and 8) or acetate buffer (pH 4 and 5).

Symbols Reactions pH

4 5 6 7 8

kh (min-1) Hydrolysis of PEG 0.069 0.0020 0.0027 0.0028 0.010

k1 (mol-1 min-1) Mono-pegylated formation 2.2 7.2 15 40 250

k2 (mol-1 min-1) Di-pegylated formation 0 3.0 6.0 66 140

k3 (mol-1 min-1) Tri-pegylated formation 0 0 3.5 42 53

8.4.2 Combined Reaction-Separation

Production of mono-pegylated α-lactalbumin was performed using the proposed

reaction-separation system with 50 mL Pyrex® glass beakers with magnetic stir bars used as

the reactor and product tanks. The pH in the reactor and product tank were monitored using a

Thermo Orion pH meter (model 420) with a Triode pH electrode and an Orion 2 Stars pH

meter with an Ultra-Micro Combination pH electrode (Thermo Scientific, Waltham, MA),

respectively. A 30 kDa Pellicon XLTM cassette was used for the tangential flow filtration. The

flow rate between the membrane and reactor was set by a Dynamax peristaltic pump (Model

RP-1, Rainin Instrument Co., Oakland, CA). To minimize the formation of di- and tri-

pegylated protein, the pH in the membrane module and the product tank were adjusted to pH

4 by addition of small amounts of acid or base (e.g., 0.1 M HCl or 0.1 M NaOH) using

Masterflex Cartridge pumps (Model 7519-20, Cole-Parmer).

Initially the reactor was filled with 30 mL of a 4 g/L α-lactalbumin solution in a 1

mM Bis-tris and 1 mM acetate buffer with 100 mM NaCl at pH 7. The product tank was

Page 182: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

164

filled 13.5 mL of the same buffer at pH 4 (set by appropriate addition of acid). The retentate

side of the TFF module was filled with buffer from the product tank prior to the experiment.

At the beginning of the experiment, 0.5 moles of PEG per mole of protein were added to the

reactor, and the reaction was allowed to proceed for 30 min in order to enhance the

production of mono-pegylated protein before starting the separation and recycle. The pumps

were then started and additional PEG was fed to the reactor (at a concentration corresponding

to a molar ratio of 4:1 relative to the mass of initial α-lactalbumin) in 15 aliquots, once every

20 min. The reactor volume was kept constant at approximately 30 mL throughout the

process (corresponding to 20 min residence time) by adjusting the flow rate from the reactor

to the product tank (qRP in Figure 8.1) to compensate for the addition of NaOH. The total

volume of the product tank increased from 13.5 mL initially to 29.2 mL at the end of the

process due to addition of acid/base for the pH adjustment (corresponding to residence times

of 9.6 and 21 min). The UF module was operated at a constant filtrate flux of 4.8 µm/s to

reduce concentration polarization, with the feed flow rate maintained at 20 mL/min. After

330 min, all pumps were stopped. All liquid in the piping/pumps was drained into appropriate

chambers (reactor and product tank). The solution pH in each chamber was reduced to a

value of 3 by adding a small amount of 4 M HCl, ensuring no further reaction. Due to the

small production scale, 40 mL of the Bis-tris/acetate (protein-free) buffer was flushed

through the TFF module to collect any residual protein inside the module from both retentate

and permeate; the collected solution was then quenched to pH 3 by addition of 4 M HCl.

Figure 8.4 shows the concentration of mono-pegylated protein produced by the

reaction-separation system as a function of time. The solid curves are model calculations for

the product concentration in the reactor and the product tank developed by solving Equations

(8.1) to (8.8) using the rate constants in Table 8.1 with sieving coefficients of 0.72, 0.070,

Page 183: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

165

and 0.055 for the native protein, PEG, and mono-pegylated protein, respectively. These

values of the sieving coefficients were determined from the data in Chapter 7 as the

geometric mean based on the values at the high and low PEG concentrations in the system.

No attempt was made to include the variation in sieving coefficients with increasing PEG

concentration in the model, although this could easily be done in the numerical solution. The

model calculations are in good agreement with the experimental data, properly capturing the

maximum concentration of the mono-pegylated protein of 2.8 g/L after approximately 200

min. The decrease of the product concentration after 200 min was due to an increase in the

volume of the product tank (associated with the addition of acid/base), which more than

compensated for the rate of production/accumulation of mono-pegylated protein in the

system. The dashed curves in Figure 8.4 are model calculations for the corresponding batch

process (no membrane separation) using the same rate constants for different values of N, the

molar ratio of PEG to α-lactalbumin initially charged to the batch reactor. The maximum

concentration of the mono-pegylated protein for the batch reactor is only 2.1 g/L,

independent of N; increasing the molar ratio of PEG to α-lactalbumin simply shifts the

location of the maximum in the concentration of mono-pegylated protein to shorter reaction

times.

Page 184: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

166

Figure 8.4 Concentration of mono-pegylated α-lactalbumin as a function of time for the

reaction-separation system. Solid curves are model calculations for the reaction-

separation process as described in the text. Dashed curves are corresponding

model results for a batch process with different molar ratio of PEG (N) relative to

the mass of initial α-lactalbumin.

Figure 8.5 shows results for the product yield, defined as the mass of mono-pegylated

protein in the reactor or in the product tank at any given time, along with the total mass, in

each case divided by the initial mass of α-lactalbumin fed to the system. The curves are

model calculations using the same parameter values as in Figure 8.4. There is no mono-

pegylated protein in the product tank over the first 30 min since the tangential flow wasn’t

started until t = 30 min. The mass of mono-pegylated protein in the product tank increases

after t = 30 min, approaching a final value of Y = 64% after 330 min. The total yield of the

mono-pegylated protein at the end of the process was 69%, with 24% of the initial α-

lactalbumin converted to multiply pegylated proteins (not shown) while 7% was unreacted.

Page 185: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

167

Figure 8.5 Yield of mono-pegylated α-lactalbumin in the reactor, product tank, and in the

system as a whole (total yield) as a function of time. Curves are model

calculations as described in the text.

8.4.3 Model Simulations

In order to develop a better understanding of the key factors governing the

performance of the reaction-separation system, a series of model simulations were performed

to evaluate the effects of different operating parameters on the yield of the desired mono-

pegylated protein. The base case was taken as follows. The reactor and the product

tank/membrane module were operated at pH 7 and 4, respectively, with constant residence

times of 20 and 5 min (corresponding to constant reactor and product tank volumes of 30 and

7.5 mL using a feed flow rate of 20 mL/min into the membrane module). The sieving

Page 186: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

168

coefficient for the native α-lactalbumin was set to 0.9 and the membrane was assumed to be

fully retentive to the mono-pegylated α-lactalbumin, the PEG, and all multiply pegylated

species. These values are in relatively good agreement with experimental measurements in

dilute solutions; the PEG sieving coefficient through the 30 kDa membrane at high PEG

concentrations tends to be small but non-zero due to the intermolecular interactions in the

bulk solution (as discussed in Chapter 7). The initial concentration of α-lactalbumin was 4

g/L, and the activated PEG was continuously fed to the reactor at a constant rate with the

total amount of added PEG equal to four times the amount of α-lactalbumin (on a molar

basis).

The effect of the solution pH in the membrane module / product tank is examined in

Figure 8.6. The top panel shows results for the base case in which the pH in the product tank

is adjusted to pH 4 (while the reactor is at pH 7). As expected, the concentration of the native

protein in the reactor decreases as a function of process time due to the pegylation reaction.

The concentrations of mono- and multiply-pegylated proteins remain fairly constant in the

reactor due to the transport of these species into the product tank. The concentration of mono-

pegylated protein in the product tank increases throughout the process, reaching a value of

12.4 g/L after 360 min. The high concentration of mono-pegylated protein in the product tank

is due to the high ratio of reactor to product tank volumes (4:1) used in this simulation.

The bottom panel shows model calculations for a process in which the pH is constant

throughout the process (including both the product tank and reactor) at pH 7. In this case, the

concentration of the mono-pegylated protein attains a maximum value of only 5.1 g/L at

t=120 min, with the mono-pegylated protein converted to multiply pegylated species in the

product tank since the rate constants for the pegylation reactions are much higher at pH 7

than at pH 4 (Table 8.1). The final concentration of the multiply pegylated protein in the

Page 187: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

169

product tank is 15.5 g/L with almost complete conversion of the native protein to the

multiply-pegylated species due to the very low hydrolysis rate for the activated PEG at pH 7.

Figure 8.6 Concentration of mono-pegylated, multiply-pegylated, and native α-lactalbumin

as a function of process time for the reaction-separation system. Top panel is for

the product tank operated at pH 4 while the bottom panel was at pH 7.

The data in Figure 8.6 have been replotted in Figure 8.7 as the dimensionless species

mass, defined as the total mass for each species (reactor plus product tank) divided by the

Page 188: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

170

initial mass of the native protein charged to the reactor. The left panel shows results with the

pH in the product tank maintained at pH 4 (the base case), while the right panel shows results

with pH 7 throughout the system. The dimensionless mass of the mono-pegylated protein

increases to 79% at the end of the reaction when the product tank is at pH 4, with only 4.4%

of the α-lactalbumin converted to multiply-pegylated species. In contrast, the dimensionless

mass of mono-pegylated protein attains a maximum value of only 36% when the product tank

is at pH 7, with the final system containing only 2.2% mono-pegylated protein with 97% of

the α-lactalbumin converted to multiply-pegylated species.

Figure 8.7 Model calculations for the dimensionless mass of mono-pegylated, multiply-

pegylated, and native α-lactalbumin as a function of process time for the reaction-

separation system operated with the product tank at pH 4 (left panel) and at pH 7

(right panel).

The effect of the membrane selectivity, defined as the ratio of the observed sieving

coefficient for the native α-lactalbumin to that of the mono-pegylated α-lactalbumin, on the

overall yield of the mono-pegylated product is shown in Figure 8.8. Simulations were

Page 189: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

171

performed with the sieving coefficient of α-lactalbumin set equal to 0.9 with the sieving

coefficient for the mono-pegylated protein adjusted to obtain the desired selectivity. The

sieving coefficient of the PEG and multiply-pegylated proteins were maintained at zero As

expected, the yield of mono-pegylated product increases with increasing membrane

selectivity due to the greater retention of the mono-pegylated protein in the product tank. At

low selectivities, the leakage of the mono-pegylated protein through the membrane and back

to the reactor leads to further pegylation and the production of more multiply-pegylated

species. The yield of the mono-pegylated protein at infinite selectivity is 79%, with the yield

being greater than 77% for selectivities greater than 100.

Figure 8.8 Model calculations for the dimensionless mass mono-pegylated, multiply-

pegylated, and native α-lactalbumin for the combined reaction-separation system

as a function of membrane selectivity.

Page 190: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

172

The effect of the residence time in the reactor (left panel) and the product tank (right

panel) is examined in Figure 8.9 for a total reaction time of 360 min. As expected, the rate of

α-lactalbumin conversion to the pegylated species increases as residence time in the reactor

(tr) increases (with residence time in the product tank fixed at tp = 5 min). However, the use

of very long residence times increases the production of di-and tri-pegylated species, leading

to a maximum in the yield of the mono-pegylated protein of 83% at tr ≈ 30 min. The effect of

the residence time in the product tank (with tr = 20 min) is considerably different. Increasing

the volume of the product tank (i.e., increasing tp) causes a reduction in the conversion of α-

lactalbumin since more of the protein accumulates in the product tank. However, the

production of the multiply-pegylated species also decreases with increasing residence time in

the product tank due to the corresponding reduction in the concentration of PEG in the

reactor (along with the reduction in the formation of the mono-pegylated species). Note that

it would be possible to increase the conversion of α-lactalbumin into the mono-pegylated

product at small values of tp by increasing the total reaction time or by increasing the

concentration of PEG in the reactor, although the latter would also effect the membrane

selectivity due to the intermolecular interactions between the PEG and mono-pegylated

protein as discussed in Chapter 7.

Page 191: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

173

Figure 8.9 Model calculations for the dimensionless mass of α-lactalbumin, the mono-

pegylated protein, and the multiply-pegylated species in the combined reaction-

separation system as a function of the residence time in the reactor (left panel)

and product tank (right panel).

Figure 8.10 shows model results for different values of the total process time, with

the residence times in the reactor (tr = 20 min) and product tank (tp = 5 min) kept constant at

the values for the base case. In all cases, the total amount of PEG added to the reactor was

kept constant (at a value equal to four times the initial moles of α-lactalbumin); thus, the rate

of PEG addition (in moles per time) decrease with increasing process time. The net result is

that the yield of multiply-pegylated species goes through a maximum of approximately 13 %

at t = 50 min. The use of smaller process times reduces the amount of multiply-pegylated

species since the total degree of pegylation is reduced, while the use of longer residence

times reduces the formation of the multiply-pegylated species by reducing the concentration

of PEG in the reactor due to the slower rate of PEG addition. The use of a reaction time of

Page 192: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

174

1000 min leads to more than 80% conversion of α-lactalbumin into the mono-pegylated

protein with less than 2% of the α-lactalbumin converted to multiply-pegylated species.

Figure 8.10 Model calculations for the dimensionless species mass for the combined

reaction-separation system as a function of total process time for a constant

amount of PEG addition.

The previous results demonstrated showed that it was possible to obtain high yield of

the desired mono-pegylated product, but there was still a fairly large amount of unreacted α-

lactalbumin. This suggests that it might be possible to achieve even higher yield by

increasing the amount of PEG added to the reactor. The effect of the PEG addition on the

dimensionless mass of α-lactalbumin, the mono-pegylated protein, and the multiply-

pegylated species is examined in Figure 8.11. Calculations were performed assuming that

the sieving coefficient for the native α-lactalbumin was equal to 1 with infinite selectivity

relative to the mono-pegylated α-lactalbumin, the PEG, and all multiply pegylated species

Page 193: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

175

(i.e., Si = 0). The total process time was fixed as 600 min, the initial α-lactalbumin

concentration was 4 g/L, and the residence time in reactor and product tank were 20 and 5

min, respectively. As expected the amount of unreacted α-lactalbumin at the end of the

reaction decreases with increasing amount of added PEG, with a corresponding increase in

the mass of the multiply-pegylated species. The yield of the mono-pegylated product attains a

maximum value of 92% at PEG:α-lactalbumin ratio of approximately 7. Note that the high

concentrations of PEG in the product tank could reduce the membrane selectivity due to

intermolecular interactions between the PEG and mono-pegylated protein as discussed in

Chapter 7; this effect was not explored in the model simulations since the membrane was

assumed to be fully retentive to the larger molecular weight species.

.

Figure 8.11 Model calculations for the dimensionless species mass for the combined

reaction-separation system as a function of total PEG feed molar ratio (moles of

added PEG to initial moles of α-lactalbumin).

Page 194: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

176

8.5 Single-Pass Ultrafiltration Process

An alternative approach to developing a membrane-based reaction-separation scheme

for the production of a specific protein conjugate is to use a single-pass ultrafiltration module

that provides very high yield of filtrate (>90% conversion of feed to filtrate). This eliminates

the need to recycle material through the membrane module; the retentate outlet is simply

collected directly in a product tank as shown in Figure 8.12, with the permeate (containing

the unreacted PEG and protein) recycled back to the reactor. Pall Corporation has recently

commercialized an ultrafiltration module (CadenceTM) that is specifically designed for single-

pass operation using internal staging to maintain sufficient crossflow velocities (and bulk

mass transfer coefficients) even as much of the feed is converted to permeate (Casey et al.,

2011). Data for IgG ultrafiltration showed that it was possible to achieve concentration

factors of at least 25-fold, i.e., 96% conversion of the feed into permeate, using the Cadence

module.

The concentrations of each species in the single-pass ultrafiltration system can be

evaluated from simple mass balances on the reactor and UF module:

Reactor:

(8.9)

Module:

(8.10)

RMRiMRMiiofeedfeediRRi

RiRqCqCSqCVr

dt

CVd,,,,,

, )(−++=

outMiMRMiioRMRiMMi

MiMqCqCSqCVr

dt

CVd,,,,,

, )(−−+=

Page 195: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

177

Figure 8.12 Schematic of the single-pass reaction and membrane-based separation system

In this case, there is no need to quench the reaction prior to the membrane module, thus

eliminating the need to adjust the pH. The outlet from the membrane module is collected in a

product tank; we assumed no reactions in this product tank (which would typically require

the addition of a quenching agent to the tank). The base case for the model calculations

assumed that the system was operated at pH 7 with the residence time for the reactor taken as

tr = 30 min. The volume in the UF module was assumed to be 1/60 that of the reactor,

corresponding to a residence time in the UF module of 0.5 min. The sieving coefficient for

the native α-lactalbumin and PEG were both set to 1 while those for mono-pegylated and

multiply pegylated species were set to zero. The initial concentration of α-lactalbumin in the

reactor was 10 g/L with the PEG present in a 0.5:1 molar ratio. Additional PEG was

continuously fed to the reactor (at a volumetric flow rate equal to the rate at which the

Page 196: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

178

product was collected to maintain a constant reactor volume), with the amount of PEG added

during the process equal to that initially charged to the reactor (total molar ratio of PEG to α-

lactalbumin of 1.5:1). The single-pass conversion in the UF module (ratio of permeate flow

rate to inlet feed flow rate) was assumed to be 0.95; the effect of the permeate flow rate on

the reaction-separation system is examined subsequently.

Typical calculated results for the above conditions are shown in Figure 8.13. As

expected, the concentration of α-lactalbumin in the reactor (left panel) decreases as a function

of time due to the pegylation reaction. The concentrations of mono- and multiply-pegylated

proteins in the reactor remain fairly low since these species are retained by the UF membrane

and collected in the product tank. The concentration of mono-pegylated protein in the

retentate outlet (right panel) increases sharply at the beginning of the process, reaching 37.7

g/L after approximately 30 min, and then decreases to only 0.31 g/L at t = 400 min. The

concentration of multiply-pegylated species in the retentate goes through a maximum at t =

35 min, but decays much more slowly at longer times due to the continual accumulation and

formation in the UF module. The PEG concentration in the system (not shown) remains

relatively low, i.e. less than 10 g/L in both reactor and UF module, due to high PEG sieving

coefficient and constant removal via the retentate outflow.

Page 197: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

179

Figure 8.13 Model calculations for the concentration of mono-pegylated, multiply-

pegylated and native α-lactalbumin as a function of process time for the single-

pass reaction-separation system performed with the base-case conditions.

The data in Figure 8.13 have been re-plotted in Figure 8.14 as the dimensionless

species mass collected in the product tank (from the retentate outlet). The dimensionless mass

of the mono-pegylated protein increases to 73% at t = 400 min with 15% of the unreacted α-

lactalbumin and 12% multiply-pegylated species. Note that there is very little of the

unreacted α-lactalbumin or pegylated products remaining in the reactor; at t = 400 min more

than 99.5% of the initial α-lactalbumin is recovered in the product tank.

Page 198: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

180

Figure 8.14 Model calculations for the dimensionless mass of mono-pegylated, multiply-

pegylated, and native α-lactalbumin in the product tank (collected from the

retentate outflow) as function of process time for the single-pass reaction-

separation system.

The effect of the residence time in the reactor (left panel) and UF module (right

panel) on the species mass for the single-pass system is examined in Figure 8.15 for a total

reaction time of 400 min. As expected, the rate of α-lactalbumin conversion to the pegylated

species increases as the residence time in the reactor (tr) increases (with residence time in the

UF module fixed at tUF = 0.5 min). The use of very long reactor residence times increases the

production of multiply-pegylated species, leading to a maximum in the yield of the mono-

pegylated protein of 73% at tr ≈ 30 min. The effect of the residence time in the UF module

(with tr = 30 min) is shown in the right panel. Increasing tp causes a reduction in the mass of

mono-pegylated protein due to the increased conversion to the multiply-pegylated forms.

Page 199: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

181

Reducing the residence time in the UF module to 0.1 min (i.e. by reducing the hold-up

volume in the module) increased the yield of mono-pegylated protein to 75%.

Figure 8.15 Model calculations for the dimensionless mass of α-lactalbumin, the mono-

pegylated protein, and the multiply-pegylated species produced by the single-

pass system as a function of the residence time in the reactor (left panel) and

UF module (right panel).

The effects of the single-pass conversion, defined as the ratio of the permeate flow

rate to the inlet feed flow rate in the UF module (qMR/qRM), on the production of the mono-

pegylated protein in the single-pass reaction-separation system are examined in Figure 8.16.

Calculations were performed with the residence time in the reactor fixed at tr = 30 min, i.e. at

a constant feed flow rate to the UF module; the conversion was varied by changing the

permeate flow rate (qMR). Results are shown for several values of the membrane selectivity,

with the sieving coefficient of the α-lactalbumin and PEG both kept at 1, i.e., by varying the

sieving coefficient of the mono-pegylated product. The sieving coefficient of any multiply-

pegylated species was assumed to be zero (i.e., full retention). The solid curves show results

Page 200: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

182

for the full model, while the dashed curves show results assuming that there is no reaction in

the UF module (ri,M = 0). As expected, the yield of mono-pegylated protein increases with

increasing membrane selectivity due to the greater retention of the mono-pegylated product

in the UF module. The yield of mono-pegylated protein initially increases with increasing

qMR/qRM since more of the unreacted α-lactalbumin is recycled back to the reactor. However,

the yield of mono-pegylated protein decreases at very high values of qMR/qRM due to: (1) the

loss of mono-pegylated protein through the membrane (particularly at low selectivities), and

(2) the formation of multiply-pegylated protein in the very highly concentrated solution

formed within the UF module under these conditions. Note that this maximum in the yield is

not seen at infinite selectivity when there is no reaction in the membrane module since all of

the mono-pegylated protein is collected in the product tank under these conditions. The

maximum yield of the mono-pegylated protein with a membrane selectivity of 500 is 73% at

a conversion of qMR/qRM = 0.95.

Page 201: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

183

Figure 8.16 Model calculations for the yield of mono-pegylated protein as a function of

single-pass conversion for different values of the membrane selectivity. Dashed

curves are results assuming that there is no reaction in the UF module.

Figure 8.17 compares the product distribution obtained from the different

reaction/separation schemes. The pie chart for the batch reactor was based on the

experimental results in Figure 8.3. The results for the reaction-separation scheme with

recycle from the product tank, including the pH adjustment (quenching), are taken from

Figure 8.11 with the molar ratio of total added PEG to initial α-lactalbumin of 7:1. The pie

chart for the single-pass ultrafiltration scheme was based on the results from Figure 8.16

using qMR/qRM = 0.95 and a membrane selectivity of 500. The yield of mono-pegylated

protein is only 53% in the batch system compared to 73% in the single-pass UF process and

92% in the product tank recycle system. Even more striking is the reduction in the formation

of the multiply-pegylated species from 17.6% in the batch reactor to only 14.6% in the

Page 202: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

184

single-pass UF and 5.6% in the product tank recycle. The dramatic improvement in the yield

and selectivity using the product tank recycle system could significantly reduce the cost of

any subsequent purification steps.

Figure 8.17 Product distribution for the production of pegylated protein obtained from the

batch reactor and the different reaction-separation schemes.

8.6 Conclusions

The experimental and theoretical results presented in this Chapter clearly demonstrate

the feasibility of enhancing the yield of the desired mono-pegylated protein using a

membrane-based reaction-separation process. Two distinct systems were examined: one

employing a separate product tank with continuous recycle through a traditional

ultrafiltration module to remove any unreacted protein and one employing a single pass

tangential flow ultrafiltration module to directly recover the mono-pegylated product while

recycling both the protein and PEG. The product tank in the first system needed to be

Page 203: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

185

maintained at low pH to minimize the formation of multiply-pegylated species, requiring

continuous addition of acid (before the membrane module) and base (into the reactor) to

maintain the desired pH values. This quenching was not required in the single-pass

ultrafiltration system, although the module had to be operated at very high permeate flow

rates (high conversion of feed to permeate) to obtain high yields of the mono-pegylated

product.

Experimental studies were performed using the product tank recycle scheme with a

commercially available TFF module (Pellicon XLTM) having a nominal molecular weight cut-

off of 30 kDa. The final yield of the mono-pegylated protein obtained with the reaction-

separation scheme was 69%, which was a significant improvement over the 50% yield

obtained with a traditional batch reaction process under the same conditions. The

experimental data from this product tank recycle system were in excellent agreement with

model calculations developed using simple mass balances, with all of the required reaction

rate constants evaluated from data obtained in independent batch kinetic experiments.

Model calculations were used to examine the effects of several key variables on the

performance of both reaction-separation systems, including operating pH, residence times in

the reactor (and product tank), membrane selectivity, UF module conversion (in the single-

pass system), and process time. The results clearly demonstrate that it is possible to achieve

yields of the mono-pegylated product well above 80% using conditions that are easily

accessible with currently available membranes / modules. In addition, the generation of

multiply-pegylated proteins was substantially reduced, which would lead to a significant

reduction (or possible elimination) of any additional purification steps to remove these multi-

pegylated species from the final product. Note that current specifications typically require

less than 3% of multi-pegylated species in therapeutic applications Grace et al. (2001), which

Page 204: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

186

is only slightly lower than the 6% obtained using the product tank recycle system. Further

improvements in performance could be achieved by enhancing the performance of the

membrane modules, both by increasing the selectivity of the separation process and by

increasing the filtrate flux (and the conversion), particularly in the single-pass module. The

results presented in this Chapter should provide an appropriate framework for the design and

optimization of combined reaction-separation processes using membrane ultrafiltration for

enhanced yield of specific protein conjugates like mono-pegylated proteins.

Page 205: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

187

Chapter 9

Conclusions and Recommendations

9.1 Conclusions

Clinical interest in pegylated proteins has been growing due to the improved

therapeutic efficacy compared to their native counterparts. In particular, the much greater

serum circulation half-life reduces the required dosage and dosing frequency. There are a

number of challenges in producing pegylated proteins, including the low yield and difficult

purification of the desired (typically mono-pegylated) product.

This thesis provides a comprehensive study of the application of membrane systems

for the purification and production of pegylated proteins. The experimental data and

theoretical analyses clearly demonstrate the importance of steric, electrostatic, and solute-

solute intermolecular interactions on the transmission and separation of pegylated proteins

using ultrafiltration. Theoretical models accounting for these effects have been developed to

provide an appropriate framework for analyzing the ultrafiltration behavior of pegylated

proteins. An ultrafiltration/diafiltration process using a charge-modified membrane has been

developed for the purification of a mono-pegylated protein, providing high product yield with

effective removal of both residual reactants and multiply pegylated species. A combined

reaction and membrane-based separation process has been discussed, with the results clearly

demonstrating the feasibility of this process for enhancing the production yield of pegylated

Page 206: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

188

proteins. The following sections summarize the key results and conclusions from this thesis.

Recommendations for future research are also discussed subsequently.

9.1.1 Electrostatic Effects

Although several previous studies have demonstrated the importance of steric and

electrostatic interactions during protein ultrafiltration, the data obtained in Chapter 4

provided the first comprehensive study of the effects of the attached PEG chain(s) on

ultrafiltration. In contrast to ion exchange chromatography, where the grafted PEG reduces

the extent of protein binding, the data obtained in this thesis showed that electrostatic

exclusion of the pegylated protein from a charged membrane can, under some circumstances,

be greater than that for the native (unmodified) protein. For example, experiments performed

with pegylated α-lactalbumin using a negatively charged composite regenerated cellulose

membrane show more than an order of magnitude reduction in the sieving coefficient as the

ionic strength is reduced from 200 to 10 mM, compared to only a 7-fold reduction in sieving

coefficient for an acetylated version of the protein possessing a similar net charge.

The attachment of the PEG chains has three distinct effects on the protein: it increases

the effective protein size (reducing the accessibility of the space within the membrane pores),

it eliminates one of the protonatable –NH2 groups due to formation of the amide bond

(increasing the net negative charge on the protein), and it alters the electrostatic potential

field around the protein due to the reduction in ion concentration within the PEG layer.

Experimental data for the electrophoretic mobility of pegylated proteins were in good

agreement with a simple model in which the plane of shear is displaced to the outer edge of

the PEG layer, with the electrostatic potential at the outer surface of the pegylated protein

Page 207: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

189

evaluated by accounting for the ion exclusion from the PEG. An analogous model was

developed for the protein sieving coefficient, with the experimental data in good agreement

with the resulting model calculations. All model parameters were evaluated from independent

measurements: the membrane pore size was determined from sieving data obtained with the

acetylated proteins, the membrane surface charge density was determined from streaming

potential measurements, and the charge on the protein core was determined from the known

amino acid sequence and pKa values of the ionizable groups accounting for the conversion of

one or more amine groups to the corresponding amide. The model accurately describes the

key experimental observations: the reduction in sieving coefficient at high ionic strength is

due to the increase in effective size of the pegylated protein while the reduction in sieving

coefficient at low ionic strength is due to both the increase in effective size and the strong

electrostatic interactions arising from the displacement of the effective protein charge to the

outer surface of the large pegylated species. This theoretical description provides an

appropriate framework for analyzing the retention characteristics of pegylated proteins during

ultrafiltration.

9.1.2 Purification of Pegylated Proteins using Charged Membranes

Although chromatographic approaches are currently used to purify most pegylated

proteins, the low dynamic binding capacity for ion exchange chromatography and the large

column required for size exclusion chromatography lead to high purification costs. The

results obtained in Chapter 5 and 6 of this thesis provide a clear demonstration that

ultrafiltration can be used to purify a pegylated protein from both residual reactants and

multiply pegylated proteins.

Page 208: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

190

The results obtained in Chapter 5 demonstrated that it is possible to remove both

unreacted (native) protein and PEG from a desired pegylated product using a single

ultrafiltration/diafiltration step with an electrically charged UltracelTM membrane having a

relatively large pore size (300 kDa). Purification of the pegylated α-lactalbumin was

performed using a diafiltration mode in which the smaller impurities (both reactants) were

washed through the membrane by addition of diafiltration buffer. The use of the highly

charged membrane with a large pore size provided high transmission of neutral PEG, high

transmission of the native protein due to its small hydrodynamic radius, and high retention of

the mono-pegylated protein due to combination of steric and electrostatic exclusion from the

membrane pores at a very low ionic strength. The process provided greater than 90% yield

with purification factors of more than 20 relative to both the PEG and native protein. The

ability to remove the unreacted protein and PEG in a single membrane step is likely to be

more attractive than the 2-stage membrane process developed by Molek and Zydney (2007)

for the purification of pegylated proteins, and the single step process makes it feasible to

consider opportunities for coupling the separation with the pegylation reaction to increase the

overall yield of the desired mono-pegylated product (as discussed in Chapter 8).

The results presented in Chapter 6 provide the first demonstration that it is possible to

use ultrafiltration for the separation of a mono-pegylated protein from the di- and tri-

pegylated forms. In this case, the ultrafiltration/diafiltration was performed using a negatively

charged version of the 300 kDa UltracelTM membrane to exploit both the larger size and the

greater electrostatic interactions of the di-pegylated species. The mono-pegylated α-

lactalbumin was obtained in the filtrate solution while the di- and tri-pegylated protein were

highly retained by the charged membrane due to a combination of steric and electrostatic

exclusion from membrane pores. The process provided greater than 95% yield of the mono-

Page 209: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

191

pegylated α-lactalbumin with more than 20-fold purification factor for the removal of di-

pegylated protein. Although the mono-pegylated α-lactalbumin was obtained in the filtrate

solution as a dilute product solution due to dilution by the diafiltration buffer, it would be

possible to eliminate this dilution effect by using a cascade ultrafiltration system (van Reis

and Zydney, 2007) or the product could be re-concentrated using a second ultrafiltration step

as part of the final product formulation.

9.1.3 Solute-Solute Intermolecular Interactions

Although the data in Chapters 5 and 6 demonstrated the potential of using membrane

systems the purification of pegylated therapeutics, the experiments were performed at

relatively low PEG and protein concentrations. The experimental and theoretical results in

Chapter 7 examined the effects of solute-solute intermolecular interactions on the

ultrafiltration of pegylated proteins. The results clearly demonstrate that transmission of both

the PEG itself and any other high molecular weight species increases with increasing PEG

concentration due to the increase in free energy in the bulk solution associated with the

strong intermolecular interactions. For example, the sieving coefficient of the 20 kDa PEG

and the mono-pegylated α-lactalbumin both increased by well over an order of magnitude in

the presence of a fairly high concentration of the 20 kDa PEG (23 g/L).

Simple theoretical models were developed for these intermolecular interactions based

on the increase in the local chemical potential of the PEG / protein; this change in chemical

potential affects both the intrinsic solute sieving coefficient and the extent of concentration

polarization. The model calculations are in good agreement with experimental data over a

range of bulk PEG concentrations and filtrate flux. The model also accurately predicts the

Page 210: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

192

negative rejection (sieving coefficient >1) observed for the small α-lactalbumin at high PEG

concentrations arising from the increase in equilibrium partition coefficient.

The strong intermolecular interactions for the PEG / pegylated proteins can have a

significant impact on ultrafiltration processes for the purification and formulation of

pegylated therapeutics. For example, the reduction in intermolecular interactions during the

course of a diafiltration process designed to remove unreacted PEG leads to a significant

reduction in the rate of PEG removal, requiring significantly more diavolumes to achieve the

same target purity. In contrast, the increase in intermolecular interactions during a batch

ultrafiltration process increases the transmission of the pegylated product leading to a

reduction in the overall product yield. This latter effect becomes increasingly significant at

the high degrees of volume reduction used to obtain highly concentrated final formulations.

The experimental results and theoretical models presented in Chapter 7 provide an

appropriate framework to calculate the magnitude of these phenomena and to develop

methods to optimize the performance of membrane processes for production of these

pegylated products.

9.1.4 Combined Reaction-Separation Systems

One of the challenges in producing a protein – polymer conjugate is to generate a

high yield of the desired conjugate, which typically involves the attachment of only a single

polymer chain to each protein. Several attempts have been made to increase the yield, e.g.

optimization of reaction conditions to drive the reaction forward and the development of

chromatographic processes in which the reaction and separation occur simultaneously in a

Page 211: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

193

single column; however, the maximum yield of a mono-pegylated protein reported by several

studies are still fairly low, in this case only slightly greater than 50%.

The experimental and theoretical results presented in Chapter 8 demonstrated the

feasibility for enhancing the yield of the desired mono-pegylated protein using a membrane-

based reaction-separation process. Two distinct systems were examined: (1) product tank

recycle scheme with continuous recycle of the native protein through a traditional

ultrafiltration module; the product tank in this system was maintained at low pH to minimize

the formation of multiply-pegylated species, requiring continuous addition of acid (before the

membrane module) and base (into the reactor) to maintain the desired pH values, and (2)

single-pass scheme with a single-pass tangential flow ultrafiltration operated at very high

permeate conversion (ratio of feed to permeate volumetric flow rate) to obtain high recycling

rate of the native protein and high yield of the mono-pegylated product. Experimental studies

were performed using the product tank recycle scheme with a commercially available TFF

module (Pellicon XLTM) having a nominal molecular weight cut-off of 30 kDa; 69% final

yield of the mono-pegylated protein was obtained with the system, which was a significant

improvement over the 50% yield obtained with a traditional batch reaction process under the

same conditions. The experimental data from the product tank recycle system were also in

excellent agreement with model calculations developed using simple mass balances, with all

of the required reaction rate constants evaluated from data obtained in batch kinetic

experiments.

Model calculations were used to examine the effects of several key variables on the

performance of both reaction-separation systems, including operating pH, residence times in

the reactor (and product tank), membrane selectivity, and process time. The results clearly

demonstrate that it is possible to achieve yields of the mono-pegylated product well above

Page 212: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

194

80% using conditions that are easily accessible with currently available membranes. Further

improvements in performance could be achieved by enhancing the performance of the

membrane modules, both by increasing the selectivity of the separation process and by

increasing the filtrate flux (and the conversion), particularly in the single-pass module. In

addition, the results presented in Chapter 8 should provided an appropriate framework for the

design and optimization of combined reaction-separation processes using membrane

ultrafiltration for enhanced yield of other protein-polymer conjugates similar to PEG-protein

systems.

Page 213: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

195

9.2 Recommendations

The experimental results and theoretical analyses presented in this thesis provide

important insights into the ultrafiltration characteristics of pegylated proteins while

demonstrating the feasibility of using ultrafiltration processes for the production and

purification of pegylated proteins. However, there are a number of areas that would benefit

from additional investigations.

The results in Chapters 5 and 6 examined the use of a negatively charged membrane

generated by attachment of sulfonic acid groups to the base cellulose membrane. The

separation required the use of very low ionic strength solutions to fully exploit the

electrostatic interactions between the charged membrane and solutes. The use of very low

buffer concentrations could be challenging in large-scale commercial processes, both in

maintaining sufficient buffering capacity and in insuring stability of the pegylated product. It

would be very appropriate to perform additional studies using membranes with greater

surface charge density or modified with different ligand chemistries that might allow one to

operate the ultrafiltration process at somewhat higher ionic strength. For example, salt-

tolerant ligands have been developed for cation (Zhao et al., 2009) and anion exchange

chromatography (Riordan et al., 2009), providing significant protein binding (capture) at

much higher ionic strength than traditional resins. Experimental studies with ultrafiltration

membranes made with these novel ligands would provide additional insights into the nature

of the electrostatic interactions with the pegylated proteins while potentially leading to the

development of more effective membrane processes for the purification of these pegylated

therapeutics.

Page 214: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

196

The pegylated proteins used in this thesis were produced by covalent attachment of

PEG onto the lysine group(s) on the protein surface, generating a pegylated protein with

greater net negative charge due to removal of protonatable lysine(s). Although this random

pegylation has been the most widely used approach for the production of pegylated

therapeutics, other methods of conjugation generate pegylated proteins without altering the

protein’s net charge, e.g., by targeting an existing disulfide bond. It would be very interesting

to examine the behavior of this class of pegylated proteins to see if it is still possible to obtain

high resolution separations by exploiting the enhanced electrostatic interaction arising from

the exclusion of salt from the PEG layer. These studies would not only provide more insight

into the ultrafiltration behavior of different pegylated products, they would hopefully

demonstrate the potential of using ultrafiltration for the purification of pegylated proteins

produced by different methods.

Most of the purification studies in this thesis were obtained using small stirred

ultrafiltration cells with a relatively low filtrate flux (velocity) to minimize concentration

polarization effects. Commercial-scale ultrafiltration is usually operated at a higher filtrate

flux using tangential flow filtration modules, which tend to have better mass transfer

characteristics and thus less concentration polarization (at a given filtrate flux). Future studies

should be performed with these scalable tangential flow filtration modules so that the results

can be more directly used for the design and optimization of large-scale systems for the

ultrafiltration of pegylated proteins.

The simple model for solute-solute intermolecular interactions developed in Chapter

7 was expressed in terms of the second virial coefficient, which describes the long-range

interactions between the solutes. It would be very interesting to directly evaluate the second

virial coefficient for the pegylated proteins over a range of solution ionic strength and pH,

Page 215: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

197

e.g., using dynamic light scattering or osmotic pressure measurements. These results could be

compared with ultrafiltration data obtained over the same range of solution conditions,

providing a more quantitative verification use of the theoretical model developed in Chapter

7.

The results in Chapter 7 also showed how the performance of an

ultrafiltration/diafiltration process is reduced by the effects of intermolecular interactions on

the transmission of the PEG. Although not examined in this thesis, it might be possible to

improve the performance of the UF/DF process by using a salt gradient during the

diafiltration, e.g., increasing the salt concentration during the diafiltration to enhance PEG

clearance by decreasing the electrostatic exclusion of the PEG from the charged membrane

pore (associated with the distortion of electrical double layer within the pore). It would be

interesting to examine this phenomenon both experimentally and theoretically.

Previous studies by Molek and Zydney (2006) showed that the transmission of a

pegylated protein during ultrafiltration is a function of the filtrate flux due to the effects of

molecular flexibility / elongation in the converging flow field into the pore. This effect was

only seen at high flux, above the flux values examined in this thesis. Additional experimental

and theoretical studies should be performed to develop a quantitative understanding of solute

molecular flexibility/elongation on the ultrafiltration process, including the impact of this

phenomenon at high PEG concentrations where solute-solute intermolecular interactions also

have a large effect on the sieving coefficient.

There is also considerable interest in the use of protein-polymer conjugates produced

with polymers other than PEG, for example, the use of protein-polysaccharide conjugates as

vaccines. It would be beneficial to extend the studies performed in this thesis to examine the

behavior of other relevant conjugated therapeutics. These results would provide additional

Page 216: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

198

insights into the relationship between the physical/chemical properties of the conjugates and

their behavior during ultrafiltration. It would also be very interesting to examine the potential

of using a reaction-separation system, similar to that described in Chapter 8, to increase the

yield and homogeneity of these protein-polymer conjugates.

The experimental work in this thesis has been performed with α-lactalbumin as a

model protein. It would be interesting to perform further studies with actual therapeutic

proteins, for example, insulin and interferon. The differences in intrinsic properties among

different pegylated proteins, e.g., the molecular size, electrical charge, number of lysine

groups, and the size of the attached PEG, could result in different pegylation reaction kinetics

and UF separation behavior. These further studies would thus help generalize the work in this

thesis and lead to an overall framework for the design and analysis of membrane processes

for the production and purification of mono-pegylated proteins.

Finally it is important to perform an economic analysis of the membrane-based

processes developed in this thesis to see how these approaches compare with other

technologies employed for the purification of pegylated proteins, e.g. ion exchange and

hydrophobic interaction chromatography. This analysis would need to consider the

production cost associated with the various technologies as well as the differences in yield

and purity of the individual processes.

Page 217: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

199

REFERENCES

Abuchowski, A., McCoy, J.R., Palczuk, N.C., Van Es, T., Davis, F.F., 1977a. Effect of

covalent attachment of polyethylene glycol on immunogenicity and circulating life of bovine liver catalase. The Journal of Biological Chemistry 252, 3582–6.

Abuchowski, A., van Es, T., Palczuk, N.C., Davis, F.F., 1977b. Alteration of immunological

properties of bovine serum albumin by covalent attachment of polyethylene glycol. The Journal of Biological Chemistry 252, 3578–81.

Acharya, G., Lee, C.H., Lee, Y., 2012. Optimization of cardiovascular stent against

restenosis: factorial design-based statistical analysis of polymer coating conditions. PloS One 7, e43100.

Albertsson, P.A., 1986. Partition of Cell Particles and Macromolecules, 3rd ed. Wiley, New York.

Anderson, J.L., Quinn, J.A., 1974. Restricted transport in small pores. A model for steric

exclusion and hindered particle motion. Biophysical Journal 14, 130–50.

Arduini, R.M., Li, Z., Rapoza, A., Gronke, R., Hess, D.M., Wen, D., Miatkowski, K., Coots,

C., Kaffashan, A., Viseux, N., Delaney, J., Domon, B., Young, C.N., Boynton, R.,

Chen, L.L., Chen, L., Betzenhauser, M., Miller, S., Gill, A., Pepinsky, R.B., Hochman,

P.S., Baker, D.P., 2004. Expression, purification, and characterization of rat interferon-

beta, and preparation of an N-terminally PEGylated form with improved pharmacokinetic parameters. Protein Expression and Purification 34, 229–42.

Arpicco, S., Dosio, F., Bolognesi, A., Lubelli, C., Brusa, P., Stella, B., Ceruti, M., Cattel, L.,

2002. Novel Poly(ethylene glycol) Derivatives for Preparation of Ribosome-Inactivating

Protein Conjugates. Bioconjugate Chemistry 13, 757–765.

Atha, D.H., Ingham, K.C., 1981. Mechanism of precipitation of proteins by polyethylene

glycols. Analysis in terms of excluded volume. The Journal of Biological Chemistry 256, 12108–17.

Bailon, P., Berthold, W., 1998. Polyethylene glycol-conjugated pharmaceutical proteins. Pharmaceutical Science & Technology Today 1, 352–356.

Baker, D.P., Lin, E.Y., Lin, K., Pellegrini, M., Petter, R.C., Chen, L.L., Arduini, R.M.,

Brickelmaier, M., Wen, D., Hess, D.M., Chen, L., Grant, D., Whitty, A., Gill, A.,

Lindner, D.J., Pepinsky, R.B., 2006. N-terminally PEGylated human interferon-beta-1a

Page 218: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

200

with improved pharmacokinetic properties and in vivo efficacy in a melanoma angiogenesis model. Bioconjugate Chemistry 17, 179–88.

Baker, R.W., Strathmann, H., 1970. Ultrafiltration of macromolecular solutions with high-flux membranes. Journal of Applied Polymer Science 14, 1197-1214.

Balan, S., Choi, J.-W., Godwin, A., Teo, I., Laborde, C.M., Heidelberger, S., Zloh, M.,

Shaunak, S., Brocchini, S., 2007. Site-specific PEGylation of protein disulfide bonds

using a three-carbon bridge. Bioconjugate Chemistry 18, 61–76.

Basu, A., Yang, K., Wang, M., Liu, S., Chintala, R., Palm, T., Zhao, H., Peng, P., Wu, D.,

Zhang, Zhenfan, Hua, J., Hsieh, M.-C., Zhou, J., Petti, G., Li, X., Janjua, A., Mendez,

M., Liu, J., Longley, C., Zhang, Zhihua, Mehlig, M., Borowski, V., Viswanathan, M.,

Filpula, D., 2006. Structure-function engineering of interferon-beta-1b for improving

stability, solubility, potency, immunogenicity, and pharmacokinetic properties by site-selective mono-PEGylation. Bioconjugate Chemistry 17, 618–30.

Bowen, W.R., Jenner, F., 1995. Theoretical descriptions of membrane filtration of colloids

and fine particles: An assessment and review. Advances in Colloid and Interface

Science 56, 141–200.

Brocchini, S., Godwin, A., Balan, S., Choi, J., Zloh, M., Shaunak, S., 2008. Disulfide bridge based PEGylation of proteins. Advanced Drug Delivery Reviews 60, 3–12.

Burns, D.B., Zydney, A.L., 1999. Effect of solution pH on protein transport through ultrafiltration membranes. Biotechnology and Bioengineering 64, 27–37.

Burns D.B., Zydney A.L., 2000. Buffer effects on the zeta potential of ultrafiltration membranes. Journal of Membrane Science 172, 10.

Burns D. B., Zydney, A. L., 2001. Contributions to electrostatic interactions on protein transport in membrane systems. AIChE Journal 47, 1101–1114.

Caliceti, P., Veronese, F.M., 2003. Pharmacokinetic and biodistribution properties of

poly(ethylene glycol)-protein conjugates. Advanced Drug Delivery Reviews 55, 1261–1277.

Carter, P.J., 2011. Introduction to current and future protein therapeutics: a protein engineering perspective. Experimental Cell Research 317, 1261–9.

Casey, C., Gallos, T., Alekseev, Y., Ayturk, E., Pearl, S., 2011. Protein concentration with

single-pass tangential flow filtration (SPTFF). Journal of Membrane Science 384, 82–

88.

Chapman, A.P., 2002. PEGylated antibodies and antibody fragments for improved therapy: a

review. Advanced Drug Delivery Reviews 54, 531–545.

Page 219: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

201

Chavez, M. D., Orpiszewski J., 2005. Apparatuses and processes for increasing protein pegylation reaction yields, US Patent US20050089952 A1.

Chrysina, E.D., Brew, K., Acharya, K.R., 2000. Crystal structures of apo- and holo-bovine

alpha-lactalbumin at 2. 2-A resolution reveal an effect of calcium on inter-lobe interactions. The Journal of Biological Chemistry 275, 37021–9.

Clark, R., Olson, K., Fuh, G., Marian, M., Mortensen, D., Teshima, G., Chang, S., Chu, H.,

Mukku, V., Canova-Davis, E., Somers, T., Cronin, M., Winkler, M., Wells, J.A., 1996.

Long-acting growth hormones produced by conjugation with polyethylene glycol. The Journal of Biological Chemistry 271, 21969–77.

Czajkowsky, D.M., Hu, J., Shao, Z., Pleass, R.J., 2012. Fc-fusion proteins: new developments and future perspectives. EMBO Molecular Medicine 4, 1015–28.

Davidson, M.G., Suter, U.W., Deen, W.M., 1987. Equilibrium partitioning of flexible

macromolecules between bulk solution and cylindrical pores. Macromolecules 20, 1141–1146.

Deen, W. M., 1987. Hindered transport of large molecules in liquid-filled pores. AIChE Journal 33, 1409–1425.

Deiters, A., Cropp, T.A., Summerer, D., Mukherji, M., Schultz, P.G., 2004. Site-specific

PEGylation of proteins containing unnatural amino acids. Bioorganic & Medicinal

Chemistry Letters 14, 5743–5.

Dimitrov, D.S., 2010. Therapeutic antibodies, vaccines and antibodyomes. mAbs 2, 347–56.

Dimitrov, D.S., 2012. Therapeutic Proteins, in: Voynov, V., Caravella, J.A. (Eds.),

Therapeutic Proteins: Methods and Protocols, Methods in Molecular Biology Vol. 899. Springer Science+Business Media, pp. 1–26.

Doherty, D.H., Rosendahl, M.S., Smith, D.J., Hughes, J.M., Chlipala, E.A., Cox, G.N., 2005.

Site-specific PEGylation of engineered cysteine analogues of recombinant human

granulocyte-macrophage colony-stimulating factor. Bioconjugate Chemistry 16, 1291–8.

Edwards, C.K., Martin, S.W., Seely, James, Kinstler, O., Buckel, S., Bendele, A.M., Ellen

Cosenza, M., Feige, U., Kohno, T., 2003a. Design of PEGylated soluble tumor necrosis

factor receptor type I (PEG sTNF-RI) for chronic inflammatory diseases. Advanced

Drug Delivery Reviews 55, 1315–36.

Edwards, C.K., Martin, S.W., Seely, James, Kinstler, O., Buckel, S., Bendele, A.M., Ellen

Cosenza, M., Feige, U., Kohno, T., 2003b. Design of PEGylated soluble tumor necrosis

factor receptor type I (PEG sTNF-RI) for chronic inflammatory diseases. Advanced Drug Delivery Reviews 55, 1315–36.

Page 220: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

202

Fee, C. J., Damodaran, V.B., 2012. Production of PEGylated Proteins, in: Subramanian, G.

(Ed.), Biopharmaceutical Production Technology. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, pp. 199–222.

Fee, C.J., 2003. Size-exclusion reaction chromatography (SERC): a new technique for protein PEGylation. Biotechnology and Bioengineering 82, 200–6.

Fee, C.J., 2007. Size comparison between proteins PEGylated with branched and linear

poly(ethylene glycol) molecules. Biotechnology and Bioengineering 98, 725–31.

Fee, C.J., van Alstine, J.M., 2004. Prediction of the viscosity radius and the size exclusion

chromatography behavior of PEGylated proteins. Bioconjugate Chemistry 15, 1304–13.

Fee, C.J., van Alstine, J.M., 2006. PEG-proteins: Reaction engineering and separation issues.

Chemical Engineering Science 61, 924–939.

Gaberc-Porekar, Vladka Zore, Irena Podobnik, Barbara Menart, V., 2008. Obstacles and

pitfalls in the PEGylation of therapeutic proteins. Current Opinion in Drug Discovery and Development 11, 242–250.

Gaertner, H.F., Offord, R.E., 1996. Site-specific attachment of functionalized poly(ethylene glycol) to the amino terminus of proteins. Bioconjugate Chemistry 7, 38–44.

Gao, J., Whitesides, G.M., 1997. Using protein charge ladders to estimate the effective

charges and molecular weights of proteins in solution. Analytical Chemistry 69, 575–80.

Gao, W., Liu, W., Mackay, J.A., Zalutsky, M.R., Toone, E.J., Chilkoti, A., 2009. In situ

growth of a stoichiometric PEG-like conjugate at a protein’s N-terminus with

significantly improved pharmacokinetics. Proceedings of the National Academy of Sciences of the United States of America 106, 15231–6.

Giddings, J.C., Kucera, E., Russell, C.P., Myers, M.N., 1968. Statistical theory for the

equilibrium distribution of rigid molecules in inert porous networks. Exclusion chromatography. The Journal of Physical Chemistry 72, 4397–4408.

Grace, M., Youngster, S., Gitlin, G., Sydor, W., Xie, L., Westreich, L., Jacobs, S., Brassard,

D., Bausch, J., Bordens, R., 2001. Structural and biologic characterization of pegylated recombinant IFN-alpha2b. Journal of Interferon & Cytokine Research 21, 1103–15.

Hamidi, M., Azadi, A., Rafiei, P., 2006. Pharmacokinetic consequences of pegylation. Drug Delivery 13, 399–409.

Harris, J. Milton, Martin, N.E., Modi, M., 2001a. Pegylation- a novel process for modifying pharmacokinetics. Clinical Pharmacokinetics 40, 539–551.

Page 221: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

203

Hasse, H., Kany, H.-P., Tintinger, R., Maurer, G., 1995. Osmotic Virial Coefficients of

Aqueous Poly(ethylene glycol) from Laser-Light Scattering and Isopiestic Measurements. Macromolecules 28, 3540–3552.

Hoyle, P.C., 1991. Polymer-modified proteins: preclinical development and regulatory strategies. Advanced Drug Delivery Reviews 6, 219–233.

Jevsevar, S., Kunstelj, Menci, Porekar, V.G., 2010. PEGylation of therapeutic proteins.

Biotechnology Journal 5, 113–28.

Kang, J.S., Deluca, P.P., Lee, K.C., 2009. Emerging PEGylated drugs. Expert Opinion on

Emerging Drugs 14, 363–80.

Katre, N. V, Knauf, M.J., Laird, W.J., 1987. Chemical modification of recombinant

interleukin 2 by polyethylene glycol increases its potency in the murine Meth A

sarcoma model. Proceedings of the National Academy of Sciences of the United States of America 84, 1487–91.

Kawashima, K., Takeshima, H., Higashi, Y., Hamaguchi, M., Sugie, H., Imamura, I., Wada,

H., 1991. High efficacy of monomethoxypolyethylene glycolconjugated l-asparaginase

(PEG2-ASP) in two patients with hematological malignancies. Leukemia Research 15,

525–530.

Keating, M.J., Holmes, R., Lerner, S., Ho, D.H., 1993. L-asparaginase and PEG

asparaginase--past, present, and future. Leukemia & Lymphoma 10 Suppl, 153–7.

King, R.S., Blanch, H.W., Prausnitz, J.M., 1988. Molecular thermodynamics of aqueous two-

phase systems for bioseparations. AIChE Journal 34, 1585–1594.

Kinstler, O., Molineux, G., Treuheit, M., Ladd, D., Gegg, C., 2002. Mono-N-terminal

poly(ethylene glycol)-protein conjugates. Advanced Drug Delivery Reviews 54, 477–85.

Kinstler, O.B., Brems, D.N., Lauren, S.L., Paige, A.G., Hamburger, J.B., Treuheit, M.J.,

1996. Characterization and stability of N-terminally PEGylated rhG-CSF. Pharmaceutical Research 13, 996–1002.

Knop, K., Hoogenboom, R., Fischer, D., Schubert, U.S., 2010. Poly(ethylene glycol) in drug

delivery: pros and cons as well as potential alternatives. Angewandte Chemie (International ed. in English) 49, 6288–308.

Kozlowski, A., Charles, S.A., Harris, J. Milton, 2001. Development of Pegylated Interferons

for the Treatment of Chronic Hepatitis C. BioDrugs 15, 419–429.

Kunitani, M., Dollinger, G., Johnson, D., Kresin, L., 1991. On-line characterization of

polyethylene glycol-modified proteins. Journal of Chromatography A 588, 125–137.

Page 222: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

204

Kurnik, R.T., Yu, A.W., Blank, G.S., Burton, A.R., Smith, D., Athalye, A.M., Van Reis, R.,

1995. Buffer exchange using size exclusion chromatography, countercurrent dialysis,

and tangential flow filtration: Models, development, and industrial application.

Biotechnology and bioengineering 45, 149–57.

Kusterle, M., Jevsevar, S., Porekar, V.G., 2008. Size of Pegylated Protein Conjugates Studied

by Various Methods. ACTA CHIMICA SLOVENICA 55, 594–601.

Kwon, B., Molek, J.R., Zydney, A.L., 2008. Ultrafiltration of PEGylated proteins: Fouling and concentration polarization effects. Journal of Membrane Science 319, 206–213.

Leader, B., Baca, Q.J., Golan, D.E., 2008. Protein therapeutics: a summary and pharmacological classification. Nature reviews. Drug Discovery 7, 21–39.

Lebreton, Bénédicte, Brown, A., Van Reis, Robert, 2008. Application of high-performance

tangential flow filtration (HPTFF) to the purification of a human pharmaceutical

antibody fragment expressed in Escherichia coli. Biotechnology and Bioengineering 100, 964–74.

Lee, D.L., Sharif, I., Kodihalli, S., Stewart, D.I.H., Tsvetnitsky, V., 2008. Preparation and

characterization of monopegylated human granulocyte-macrophage colony-stimulating

factor. Journal of Interferon & Cytokine Research 28, 101–12.

Li, W., Zhan, P., De Clercq, E., Lou, H., Liu, X., 2013a. Current drug research on

PEGylation with small molecular agents. Progress in Polymer Science 38, 421–444.

Mayolo-Deloisa, K., González-Valdez, J., Guajardo-Flores, D., Aguilar, O., Benavides, J.,

Rito-Palomares, M., 2011. Current advances in the non-chromatographic fractionation

and characterization of PEGylated proteins. Journal of Chemical Technology & Biotechnology 86, 18–25.

Mehta, A., Len, A., Lebreton, Benedicte, Wolk, B., Fogle, J., Tse, M.L., Fontes, N., Van

Reis, Robert, Shrestha, R., 2008. Purifying therapeutic monoclonal antibodies. Chemical Engineering Progress 104, S14–S20.

Mehta, A., Zydney, A.L., 2006. Effect of membrane charge on flow and protein transport during ultrafiltration. Biotechnology Progress 22, 484–92.

Mehta, A., Zydney, A.L., 2005. Permeability and selectivity analysis for ultrafiltration membranes. Journal of Membrane Science 249, 245–249.

Mei, L.-H., Lin, D.-Q., Zhu, Z.-Q., Han, Z.-X., 1995. Densities and Viscosities of

Polyethylene Glycol + Salt + Water Systems at 20 .degree.C. Journal of Chemical & Engineering Data 40, 1168–1171.

Page 223: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

205

Menon, M.K., Zydney, A.L., 2000. Determination of Effective Protein Charge by Capillary

Electrophoresis: Effects of Charge Regulation in the Analysis of Charge Ladders. Analytical Chemistry 72, 5714–5717.

Milunović, T., Kunstelj, Menči, Fidler, K., Anderluh, G., Gaberc Porekar, V., 2012. Coupling

purification and on-column PEGylation of tumor necrosis factor alpha analogue.

Analytical Biochemistry 430, 105–7.

Miron, T., Wilchek, M., 1982. A spectrophotometric assay for soluble and immobilized N-hydroxysuccinimide esters. Analytical Biochemistry 126, 433–435.

Mochizuki, S., Zydney, A.L., 1993. Theoretical analysis of pore size distribution effects on membrane transport. Journal of Membrane Science 82, 211–227.

Molek, J.R., 2008. Ph.D. dissertation: Ultrafiltration of PEGylated Proteins. The Pennsylvania State University.

Molek, J.R., Zydney, A.L., 2006. Ultrafiltration characteristics of pegylated proteins. Biotechnology and Bioengineering 95, 474–82.

Molek, J.R., Zydney, A.L, 2007. Separation of PEGylated alpha-lactalbumin from unreacted precursors and byproducts using ultrafiltration. Biotechnology Progress 23, 1417–24.

Moosmann, A., Christel, J., Boettinger, H., Mueller, E., 2010. Analytical and preparative

separation of PEGylated lysozyme for the characterization of chromatography media. Journal of Chromatography A 1217, 209–15.

Morão, A.M., Nunes, J.C., Sousa, F., Pessoa de Amorim, M.T., Escobar, I.C., Queiroz, J.A.,

2011. Ultrafiltration of supercoiled plasmid DNA: Modeling and application. Journal of

Membrane Science 378, 280–289.

Morar, S.A., Schrimsher, J.L., Chavez, Mark D., 2006. PEGylation of Proteins: A Structural

Approach. BioPharm International 19, 34-+.

Mosbah, I.B., Franco-Gou, R., Abdennebi, H.B., Hernandez, R., Escolar, G., Saidane, D.,

Rosello-Catafau, J., Peralta, C., 2006. Effects of polyethylene glycol and hydroxyethyl

starch in University of Wisconsin preservation solution on human red blood cell aggregation and viscosity. Transplantation Proceedings 38, 1229–35.

Nelson, D.L., Cox, M.M., 2008. Lehninger Principles of Biochemistry, 4th ed. W.H. Freeman & Company, New York.

Nguyen, D.P., Lusic, H., Neumann, H., Kapadnis, P.B., Deiters, A., Chin, J.W., 2009.

Genetic encoding and labeling of aliphatic azides and alkynes in recombinant proteins

via a pyrrolysyl-tRNA Synthetase/tRNA(CUA) pair and click chemistry. Journal of the

American Chemical Society 131, 8720–1.

Page 224: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

206

Ohshima, H., 2002. Modified Henry function for the electrophoretic mobility of a charged

spherical colloidal particle covered with an ion-penetrable uncharged polymer layer. Journal of Colloid and Interface Science 252, 119–25.

Opong, W.S., Zydney, A.L., 1991. Diffusive and convective protein transport through asymmetric membranes. AIChE Journal 37, 1497–1510.

Pabst, T.M., Buckley, J.J., Ramasubramanyan, N., Hunter, A.K., 2007. Comparison of strong

anion-exchangers for the purification of a PEGylated protein. Journal of Chromatography A 1147, 172–82.

Pasut, G., Veronese, F.M., 2007. Polymer–drug conjugation, recent achievements and general strategies. Progress in Polymer Science 32, 933–961.

Payne, R.W., Murphy, B.M., Manning, M.C., 2011. Product development issues for PEGylated proteins. Pharmaceutical Development and Technology 16, 423–40.

Permyakov, E.A., Berliner, L.J., 2000. alpha-Lactalbumin: structure and function. FEBS Letters 473, 269–74.

Phillips, M.W., Bolton, G., Krishnan, M., Lewnard, J.J., Raghunath, B., 2007. Virus

Filtration Process Design and Implementation, in: Abhinav A . Shukla , Mark R . Etzel,

and S.G. (Ed.), Process Scale Bioseparations for the Biopharmaceutical Industry. Boca Raton, pp. 333–365.

Piquet, G., Gatti, M., Barbero, L., Traversa, S., Caccia, P., Esposito, P., 2002. Set-up of large

laboratory-scale chromatographic separations of poly(ethylene glycol) derivatives of the

growth hormone-releasing factor 1-29 analogue. Journal of Chromatography A 944, 141–8.

Pujar, N.S., Zydney, AL, 1997. Charge Regulation and Electrostatic Interactions for a

Spherical Particle in a Cylindrical Pore. Journal of Colloid and Interface science 192, 338–49.

Pujar, N.S., 1996. Ph.D. Dissertation: Electrostatic and Electrokinetic Interactions During

Protein Filtration Using Semi-permeable Membranes. The University of Delaware, Newark.

Pujar, N.S., Zydney, A.L., 1994. Electrostatic and Electrokinetic Interactions during Protein

Transport through Narrow Pore Membranes. Industrial & Engineering Chemistry

Research 33, 2473–2482.

Rao, S., 2006. Ph.D.Dissertation: Protein separation using affinity ultrafiltration with small charged ligands. The Pennsylvania State University, University Park.

Reichert, J.M., 2011. Antibody-based therapeutics to watch in 2011. mAbs 3, 76–99.

Page 225: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

207

Riordan, W., Heilmann, S., Brorson, K., Seshadri, K., He, Y., Etzel, M., 2009. Design of salt-

tolerant membrane adsorbers for viral clearance. Biotechnology and Bioengineering 103, 920–9.

Roberts, M.J., Bentley, M.D., Harris, J M, 2002. Chemistry for peptide and protein PEGylation. Advanced Drug Delivery Reviews 54, 459–76.

Rohani, M.M., 2011. Ph.D.Dissertation: Effect of electrostatic interactions during protein

ultrafiltration: effect of ligand chemistry and protein surface distribution. The Pennsylvania State University, University Park.

Rosendahl, M.S., Doherty, D.H., Smith, D.J., Carlson, S.J., Chlipala, E.A., Cox, G.N., 2005.

A long-acting, highly potent interferon alpha-2 conjugate created using site-specific PEGylation. Bioconjugate Chemistry 16, 200–7.

Russell, E., Wang, A., Rathore, A.S., 2007. Harvest of a Therapeutic Protein Product from

High Cell Density Fermentation Broths: Principles and Case Study, in: Abhinav A .

Shukla , Mark R . Etzel, and S.G. (Ed.), Process Scale Bioseparations for the Biopharmaceutical Industry. CRC Press, Boca Raton, pp. 1–58.

Sato, H., 2002. Enzymatic procedure for site-specific pegylation of proteins. Advanced Drug

Delivery Reviews 54, 487–504.

Seely, J.E., Richey, C.W., 2001. Use of ion-exchange chromatography and hydrophobic

interaction chromatography in the preparation and recovery of polyethylene glycol-linked proteins. Journal of Chromatography A 908, 235–41.

Seely, J.E., Scott, B., Green, P., Richey, C., 2005. Making site-specific PEGylation work. Biopharm international 18, 30.

Shah, T., 2013. Bioconjugates: The Adaptable Challenge. Biopharm International 26, 34–38.

Sharma, U., Carbeck, J.D., 2005. Hydrodynamic radius ladders of proteins. Electrophoresis 26, 2086–91.

Sharma, U., Negin, R.S., Carbeck, J.D., 2003. Effects of Cooperativity in Proton Binding on

the Net Charge of Proteins in Charge Ladders. The Journal of Physical Chemistry B 107, 4653–4666.

Sherman, M.R., Saifer, M.G.P., Perez-Ruiz, F., 2008. PEG-uricase in the management of

treatment-resistant gout and hyperuricemia. Advanced Drug Delivery reviews 60, 59–68.

Smith, F.G., Deen, William M, 1983. Electrostatic effects on the partitioning of spherical

colloids between dilute bulk solution and cylindrical pores. Journal of Colloid and

Interface Science 91, 571–590.

Page 226: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

208

Smith, F.G., Deen, William M., 1980. Electrostatic double-layer interactions for spherical colloids in cylindrical pores. Journal of Colloid and Interface Science 78, 444–465.

Smith, K. A., Colton, C. K., Merrill, E.W., Evans, L.B., 1968. Convective transport in a batch

dialyzer: determination of the true membrane permeability from a single measurement. Chemical Engineering Progress Symposium Series 64, 45.

Smith, M., 1970. Molecular weights of proteins and some other materials including

sedimentation, diffusion and frictional coefficients and partial specific volumes, in: Handbook of Biochemistry. The Chemical Rubber, Co., Cleveland, pp. C3–C292.

Stoner, M.R., Fischer, N., Nixon, L., Buckel, S., Benke, M., Austin, F., Randolph, T.W.,

Kendrick, B.S., 2004. Protein-solute interactions affect the outcome of ultrafiltration/diafiltration operations. Journal of Pharmaceutical Sciences 93, 2332–42.

Tan, Y., Sun, X., Xu, M., An, Z., Tan, X., Han, Q., Miljkovic, D.A., Yang, M., Hoffman,

R.M., 1998. Polyethylene glycol conjugation of recombinant methioninase for cancer therapy. Protein Expression and Purification 12, 45–52.

Thordarson, P., Le Droumaguet, B., Velonia, K., 2006. Well-defined protein-polymer

conjugates--synthesis and potential applications. Applied Microbiology and

Biotechnology 73, 243–54.

U.S. Food and Drug Administration, 2013. Vaccines licensed for immunization and

Distribution in the US with supporting documents, viewed 5 June 2013, from

http://www.fda.gov/BiologicsBloodVaccines/Vaccines/ApprovedProducts/ucm093830.htm

van Reis, R, Brake, J.., Charkoudian, J., Burns, D.B, Zydney, A.L, 1999. High-performance

tangential flow filtration using charged membranes. Journal of Membrane Science 159, 133–142.

van Reis, R, Saksena, S., 1997. Optimization diagram for membrane separations. Journal of Membrane Science 129, 19–29.

van Reis, R., Zydney, A.L., 2007. Bioprocess membrane technology. Journal of Membrane Science 297, 16–50.

Veronese, Francesco M, Pasut, Gianfranco, 2005. PEGylation, successful approach to drug delivery. Drug Discovery Today 10, 1451–8.

Veronese, Francesco M., Caliceti, P., Pastorino, A., Schiavon, O., Sartore, L., Banci, L.,

Scolaro, L.M., 1989. Preparation, physico-chemical and pharmacokinetic

characterization of monomethoxypoly(ethylene glycol)-derivatized superoxide dismutase. Journal of Controlled Release 10, 145–154.

Page 227: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

209

Viaene, A., Volckaert, G., Joniau, M., Baetselier, A., Cauwelaert, F., 1991. Efficient

expression of bovine alpha-lactalbumin in Saccharomyces cerevisiae. European Journal of Biochemistry 202, 471–477.

Vilker, V.L., Colton, Clark K, Smith, K. A, 1981. The osmotic pressure of concentrated

protein solutions: Effect of concentration and ph in saline solutions of bovine serum

albumin. Journal of Colloid and Interface Science 79, 548–566.

Walsh, G., Jefferis, R., 2006. Post-translational modifications in the context of therapeutic proteins. Nature Biotechnology 24, 1241–52.

Willauer, H.D., Huddleston, J.G., Rogers, R.D., 2002. Solute Partitioning in Aqueous

Biphasic Systems Composed of Polyethylene Glycol and Salt: The Partitioning of

Small Neutral Organic Species. Industrial & Engineering Chemistry Research 41, 1892–1904.

Working, P.K., Newman, M.S., Johnson, J., Cornacoff, J.B., 1997. Safety of poly(ethylene

glycol) and poly(ethylene glycol) derivatives, in: Harris, J.M., Zalipsky, S. (Eds.), Poly(ethylene Glycol). American Chemical Society, Washington, DC, pp. 45–57.

Youn, Y.S., Na, D.H., Yoo, S.D., Song, S.-C., Lee, K.C., 2004. Chromatographic separation

and mass spectrometric identification of positional isomers of polyethylene glycol-

modified growth hormone-releasing factor (1-29). Journal of Chromatography A 1061, 45–49.

Yun, Q., Yang, R.E., Chen, T., Bi, J., Ma, G., Su, Z., 2005. Reproducible preparation and

effective separation of PEGylated recombinant human granulocyte colony-stimulating

factor with novel “PEG-pellet” PEGylation mode and ion-exchange chromatography. Journal of Biotechnology 118, 67–74.

Zeman, L.J., Zydney, A.L., 1996. Microfiltration and Ultrafiltration: Principles and Applications. Marcel Dekker, New York.

Zhang, M., Zhang, Y., Zhu, S., Wu, L., Dou, H., Yin, C., 2007. Synthesis and

Chromatographic Separation of Monomethoxypolyethylene Glycol Modified Insulin. Separation Science and Technology 42, 789–801.

Zhao, G., Dong, X.-Y., Sun, Y., 2009. Ligands for mixed-mode protein chromatography: Principles, characteristics and design. Journal of Biotechnology 144, 3–11.

Zinman, B., 2013. Newer insulin analogs: advances in basal insulin replacement. Diabetes, Obesity & Metabolism 15 Suppl 1, 6–10.

Zydney A.L., 1997. Stagnant film model for concentration polarization in membrane systems. Journal of Membrane Science 130, 7.

Page 228: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

210

Zydney, A.L., Pujar, N.S., 1998. Protein transport through porous membranes: effects of

colloidal interactions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 138, 133–143.

Zydney, A.L., 1992. Concentration effects on membrane sieving: development of a stagnant

filmmodel incorporating the effects of solute-solute interactions. Journal of Membrane

Science 68, 183–190.

Page 229: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

211

APPENDIX

A.1 Evaluation of Protein Sieving Coefficient

This section presents the computer program (Software: Mathematica 7.0) for the

evaluation of sieving coefficients for pegylated proteins as a function of solution ionic

strength. As discussed in Chapter 2 and Chapter 4, the actual sieving coefficient for a protein

is dependent on both steric and electrostatic interactions between the protein and the

membrane pore. The electrostatic energy of interaction for a protein partitioning into a

charged cylindrical pore was evaluated using the model developed by Smith and Deen

(1980). The surface charge density at the outer surface of a pegylated protein was also

evaluated using the decay potential through the PEG layer as discussed in Section 4.3.3.

(*Solute*) rs =53.1; (*pegylated protein Radius (Angstrom)-- from CJ Fee correlations*) rpro=19.9;(*radius of protein core, Angstrom)*) (*Solution*) z1 = 1; (*charge of ion 1*) z2 = -1; (*charge of ion 2*) η = 0.001*10-20; (*viscosity - N*s/A^2 *) (*Membrane*) rpore=83;(*radius of membrane pore calculated from membrane hydraulic permeability*) ε =0.5; (*porosity - Obtain estimate from manufacturer*) qp = -2.7*10-23; (*surface charge of pore determined by zeta potential Coulomb/A^2*) δ = 10000; (*membrane thickness, Angstrom)*) z=.20; (*ratio of standard deviation to pore size*) σ = z*rpore;(*Standard Deviation for pore size*) F = 96500; (*Faraday's constant, Coulombs/mol*) R = 8.314*1010;(*Universal Gas Constant, Newton-Angstrom/mole-Kelvin*)

Page 230: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

212

T = 298; (*Kelvin*) εr = 80; (*relative permittivity of water*) εnot = 8.854*10-32; k=1.3807*10-13; (*Boltzmann's Constant, Newton-Angstrom/Kelvin*) a1 = -73./60.; a2 = 77293/50400; a3 =-22.5083; a4 = -5.6117; a5 = -0.3363; a6 = -1.216; a7 = 1.647; b1 = 7/60; b2 = -2277/50400; b3 = 4.0180; b4 = -3.9788; b5 = -1.9215; b6 = 4.392; b7 = 5.006; Ionic = 1; IonResults = Table[{ phiKc=0; phiKd = 0; Ionic = Ionic*3, (*to vary ionic strength*) c1 = Ionic*10-30;(*positive ions moles/A3*) c2 = Ionic*10-30 ; (*negative ions, moles/A3*) κ = ((F^2*(z1^2*c1+z2^2*c2))/(εnot*εr*R*T))0.5; k1 = ((rpro/rs)^2)*(1+κ*rs)/(1+κ*rpro)*Exp[-0.38*κ*(rs-rpro)];(*k1 is the decay function due to electrostatic alternation associated with PEG layer*) qs = k1*(-4.98)* 1.6*10-19/(4*π*rpro^2); (*qs is the surface charge density at the outer edge of a pegylated protein*) (*-4.98 is the net charge of the pegylated alpha lactalbumin core, which was calculated elsewhere (Excel Spreadsheet) using charged regulation theory*) Do[ λ = rs /r;

∑∑=

=

− +

−+−=

7

3

32

1

2/52 )1(1)1(24

9

g

g

g

g

g

gs bbK λλλπ ;

Page 231: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

213

∑∑=

=

− +

−+−=

7

3

32

1

2/52 )1(1)1(24

9

g

g

g

g

g

g aaKt λλλπ ;

P = (1/(r*(2*π)0.5)*(Log[1 +(z)^2])-0.5*Exp[-(Log[r/(σ/z)] + Log[1+(z)^2]/2)2/(2*Log[1+(z)2])]);(*Pore size distribution*) m= NIntegrate[(BesselK[1,(((κ*r)^2 + x^2 )0.5)]/BesselI[1,(((κ*r)^2 + x^2)0.5)]),{x,0,1.3}];(*Function M0*) h= (1 + κ*r*λ)*�-κ*r*λ - (1 - κ*r*λ)*�κ*r*λ;(*Function h*) σs = (F*r*qs)/(εnot*εr*R*T); σp = (F*r*qp)/(εnot*εr*R*T); As = (4*π*λ4*κ*r*Exp[κ*r*λ]*m)/(1 + κ*r*λ); Asp = (4*π2*λ2)/BesselI[1,κ*r]; Ap = (π2*h)/(( κ*r)2*(BesselI[1,κ*r])2); V = (As*σs2 + Asp*σs*σp + Ap*σp2)/(π*κ*r*(1 + κ*r*λ)*�(- κ*r*λ)-m*h);(*electrostatic energy of interactions*) c = r*εnot*εr*R^2*T^2*v/F2;

j =(1-λ)^2*Exp[-c/(k*T)]*(Ks/(2*Kt)*(2-((1-λ)^2*Exp[-c/(k*T)]))) ;(*PhiKc*) n=6*π*(1-λ)^2/Kt*Exp[-c/(k*T)]; (*PhiKd*) phiKc1=(j*p*(r^4)); phiKd1=(n*p*(r^4)); phiKd=phiKd1+phiKd; phiKc=phiKc1+phiKc, {r,rs+0.1,500}]; DenR = NIntegrate[rp^4/(rp*(2*π)0.5)*(Log[1+(z)^2])-0.5*Exp[-(Log[rp/(σ/z)]+Log[1+(z)^2]/2)2/(2*Log[1+(z)2])],{rp,1,500}]; Printphikc = phiKc; Printphikd = phiKd; SFinal = phiKc/DenR,(*asymptotic sieving coefficient*) Diffuse = 1.38*10^-23*298/(6*π*µ *rs/10000000000); (*diffusion coefficient in m2/s*) Jv = 8*10-6; (*flux, m/s*) ω = 600*2*π/60;

Page 232: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

214

µ = 0.001; (*viscosity in kg/m*s*) Area = 0.00041; (* membrane area m^2 *) Beta1 = 0.23; aStirred = 0.567; bStirred = 0.33; Ro = 1000; (*density of solution - kg/m^3*) Vel = Jv/ε; (*velocity of fluid through pores in m/s*) Rey = Ro*rStir^2*ω/(µ);

Sc = µ /(Ro*Diffuse); rStir = 0.0125; (membrane radius, m) Sh = Beta1*Rey^aStirred*Sc^bStirred; kMass = Sh*Diffuse/rStir; Pe = (Vel*δ/(Diffuse*10000000000))*(phiKc/phiKd), (*Peclet number*) SFinal2=phiKc/DenR*E^Pe/(phiKc/DenR+E^Pe-1), (*actual sieving coefficient*) So=SFinal/((1-SFinal)*Exp[-Jv/kMass]+SFinal)(*Observed Sieving Coefficient*) }, {StepResults,1,5}]; Print["Ionic strength / S_asymptotic / Peclet num / Sa / So "] TableForm[IonResults] Export["IonicStrength.csv",IonResults]

Page 233: PURIFICATION AND PRODUCTION OF PEGYLATED PROTEINS USING

VITA

Krisada Ruanjaikaen

EDUCATION

The Pennsylvania State University, University Park, PA

• Doctor of Philosophy in Chemical Engineering Aug 13

Chulalongkorn University, Bangkok, Thailand

• Bachelor of Engineering in Chemical Engineering, First Honors May 06

RESEARCH / WORK EXPERIENCE

Doctoral Research, The Pennsylvania State University Jan 09 to June 13

Advisor: Dr. Andrew L. Zydney

Thesis title: Purification and production of pegylated proteins using membrane processes.

Industrial Experience

Process Engineer (Siam Cement Chemicals, Rayong, Thailand) May 06 to Jan 07

Provided on the-floor technical support for the manufacturing team and led process-related

investigations. Led projects on the modification of heat exchangers and refrigerator systems for

debottlenecking of the production capacity of low density polyethylene (LDPE) resins.

Teaching Experience

Teaching Assistant (The Pennsylvania State University)

• Unit Operation Laboratory Aug 12 to Dec 12

• Chemical Process Design Jan 09 to May 09

SELECTED AWARDS

John R. and Jeanette Dachille Mcwhirter Graduate Scholarship, 2008

Walter R. and Aura Lee Supina Graduate Fellowship, 2008

Siam Cement Group’s Outstanding Academic Achievement Award, 2005

PUBLICATIONS

Journal Papers

• Ruanjaikaen K, Zydney AL. 2013. “Intermolecular interactions during ultrafiltration of

pegylated proteins”. Biotechnology Progress. In Press.

• Ruanjaikaen K, Zydney AL. 2011. “Purification of singly pegylated α-lactalbumin using

charged ultrafiltration membranes”. Biotechnology and Bioengineering 108: 822-829.

• Molek JR, Ruanjaikaen K, Zydney AL. 2010. “Effect of electrostatic interactions on

transmission of pegylated proteins through charged ultrafiltration membranes”. Journal of

Membrane Science 353: 60–69.

Conference Presentations

• Ruanjaikaen K, Zydney AL. “Role of intermolecular interactions on ultrafiltration of

pegylated proteins”. American Institute of Chemical Engineers Annual Meeting, Pittsburgh,

Nov 2012.

• Zydney AL, Ruanjaikaen K., “Purification and production of protein conjugates using

membrane systems”. Recovery of Biological Products XV, Stowe, Aug 2012.

• Ruanjaikaen K, Zydney AL. “Recent advances in purification of pegylated proteins using

charge- modified ultrafiltration membranes”. American Chemical Society National Meeting,

San Diego, March 2012.

• Ruanjaikaen K, Zydney AL. “Purification of pegylated proteins using high performance

tangential flow filtration”. North American Membrane Society Meeting, Washington DC,

May 2010.