M
ASTERT
HESISD
EVELOPMENT OF
C
LEANING
-I
N
-P
LACE
P
ROCEDURES FOR
P
ROTEIN
A
C
HROMATOGRAPHY RESINS
U
SING
D
ESIGN OF
E
XPERIMENTS AND
H
IGH
T
HROUGHPUT
S
CREENING
T
ECHNOLOGIES
H
ANNAT
ENGLIDENM
ASTERT
HESIS PERFORMED ATGE
H
EALTHCAREF
EBRUARY22
ND2008
LITH-IFM-EX-08/1923—SE
I
NSTITUTE OFP
HYSICS,
C
HEMISTRY ANDB
IOLOGYD
EVELOPMENT OF
C
LEANING
-I
N
-P
LACE
P
ROCEDURES FOR
P
ROTEIN
A
C
HROMATOGRAPHY RESINS
U
SING
D
ESIGN OF
E
XPERIMENTS AND
H
IGH
T
HROUGHPUT
S
CREENING
T
ECHNOLOGIES
H
ANNAT
ENGLIDENM
ASTERT
HESIS PERFORMED ATGE
H
EALTHCAREF
EBRUARY22
ND2008
S
UPERVISORSA
NNAG
RÖNBERG ANDH
ANSJ
J
OHANSSONE
XAMINERA
BSTRACT
Robust and efficient cleaning procedures for protein A chromatography resins used for production of monoclonal antibody based biopharmaceuticals are crucial for safe and cost efficient processes. In this master thesis the effect of different cleaning regimes with respect to ligand stability of two protein A derived media, MabSelectTM and MabSelect
SuReTM, has been investigated. A 96-well format has been used for preliminary screening
of different cleaning agents, contact times and temperatures. NaCl as a ligand stabilizer during cleaning-in-place (CIP) was also included as a parameter. For optimal throughput and efficiency of screening, Rectangular Experimental Design for Multi-Unit Platforms; RED-MUP, and TECAN robotic platform have been utilized. For verification of screening, selected conditions were run in column format using the parallel chromatography system ÄKTAxpressTM. In the efficiency study, where a manual preparation of CIP solutions was
compared with an automated mode performed in TECAN, the total process time ended up at eight hours versus three days respectively. However, the time measured included the learning process for the TECAN platform and for further preparations the automated mode is the superior choice. The study confirmed the higher alkaline stability of MabSelect SuRe compared to MabSelect. After exposure to 0.55 M NaOH during 24h MabSelect SuRe still retained 90% of the initial capacity. In contrast MabSelect had 60% of the initial binding capacity. When CIP with 10 mM NaOH was performed at 40 °C MabSelect reduced its capacity by half while MabSelect SuRe still had a binding capacity of 80%. The 96-well screening showed that an addition of NaCl during CIP had a significant positive effect on the stability of MabSelect, but needs to be verified on column format. The correlation between results from screening in 96-well filter plate and column format was good.
T
ABLE OF
C
ONTENTS
1 INTRODUCTION 1
2 BACKGROUND 3
2.1 BIOTECH MANUFACTURING IS GROWING UP 3
2.2 LEAN PRODUCTION IN BIOPHARMACEUTICAL INDUSTRY 4
2.3 ANTIBODIES 5
2.3.1 Evolution of Therapeutic Antibodies 6
2.4 MABS AS BIOPHARMACEUTICALS OF TODAY AND FUTURE 7
2.5 CHROMATOGRAPHY IN GENERAL AND AFFINITY CHROMATOGRAPHY ABOVE ALL 8
2.5.1 Protein A chromatography 9
2.5.2 MabSelect and MabSelect SuRe characteristics 9
2.6 BINDING CAPACITY AND TIME OF RESIDENCE 10
2.7 GENERAL DESCRIPTION OF A THERAPEUTIC ANTIBODY PRODUCTION 11
2.7.1 The Protein A capturing unit 13
2.7.2 Cleaning-in-place 14
2.7.3 Agents possible for stabilization of the protein a ligand 15
2.7.4 MAb purification using MabSelect versus MabSelect SuRe 16
2.8 HIGH THROUGHPUT SCREENING FORMAT 17
2.8.1 General description of screening in the 96-well format 19
2.8.2 High-throughput using ÄKTAxpresstm 21
2.9 INTRODUCTION TO STATISTICAL EXPERIMENTAL DESIGN 22
2.9.1 Design of Experiments - DoE 23
2.9.2 Red-Mup design – hand in hand with the 96-well format 24
2.9.3 Evaluation in Umetrics Modde v.8.0 25
2.10 AIM OF THE MASTER THESIS 28
3 MATERIALS 29
3.2 CHEMICALS 29
3.3 EQUIPMENT 30
3.4 SOFTWARE 30
4 METHODS 31
4.1 ESTABLISHING RUNNING CONDITIONS IN THE 96-WELL FORMAT PROCEDURE 31
4.2 PERFORMING THE 96-WELL METHOD 33
4.2.1 Determination of static binding capacity of chromatography resin in 96-well filter plate 33 4.2.2 Performance of Cleaning-In-Place of chromatography resin in 96-well filter plate 34
4.3 PERFORMING THE COLUMN FORMAT METHOD 35
4.3.1 Determination of dynamic binding capacity of chromatography resin in column format 35 4.3.2 Performance of Cleaning-In-Place of chromatography resin in column format 36
4.4 MANUAL VS AUTOMATED LIQUID PREPARATION USING TECAN ROBOT 36
5 RESULTS & DISCUSSION 39
5.1 ESTABLISHING RUNNING CONDITIONS IN THE 96-WELL FORMAT PROCEDURE 39
5.2 DEVELOPMENT OF THE COLUMN CHROMATOGRAPHY FORMAT 41
5.3 CIP STUDIES PERFORMED IN 96-WELL FILTER PLATE FORMAT 42
5.3.1 Temperature study 42
5.3.2 Stability Study performed in 96-well filter plate 48
5.3.3 Time study 52
5.4 EVALUATION OF ACCORDANCE BETWEEN RESULTS FROM 96-WELL AND COLUMN FORMATS 55
5.4.1 Evaluation of accordance between formats for MabSelect 56
5.4.2 Evaluation of accordance between formats for MabSelect SuRe 59
5.5 MANUAL VS AUTOMATED LIQUID PREPARATION USING TECAN ROBOT 61
6 CONCLUSIONS 63 7 FUTURE EXPERIMENTS 65 8 ACKNOWLEDGEMENTS 67 9 REFERENCES 69 10 APPENDICES 71 10.1 TEMPERATURE STUDY 71
10.2 EVALUATION OF ACCORDANCE BETWEEN RESULTS FROM 96-WELL AND COLUMN FORMATS 74
T
ABLE OF
A
BBREVIATIONS
Ab Antibody
CCC Central Composite Circumscribed
CCF Central Composite Face Centered
CV Column Volume
DBC Dynamic binding capacity
HT High throughput
HTPD High Throughput Process Development
HTS High throughput screening
Ig Immunoglobulin
IPA 2-propanol, iso-propanol
JIT Just-in-Time
LOF Lack Of Fit
mAb Monoclonal antibody
MLR Multiple Linear Regression
NaCl Sodium chloride
NaOH Sodium hydroxide
RED-MUP Rectangular Experimental Design for Multi-Unit Platforms
RSD Residual Standard Deviation
RT Room Temperature
1 I
NTRODUCTION
The therapeutic monoclonal antibodies (mAbs) have emerged as an important and rapidly growing new class of drugs for treatment of several human diseases and have outpaced traditional pharmaceuticals in both sales and regulatory approvals. Today there are more than 370 biotech products and vaccines in clinical trials targeting more than 200 diseases. Approximately 30% of these biopharmaceuticals is represented by antibodies and Fc-fusion proteins. With 17 mAb drugs on the market, new hope is brought into the future treatment of currently incurable diseases. [1], [2]
Protein A chromatography serves as the capture step in the purification of mAbs and also functions as the key volume reduction step in the process since protein A concentrates the product stream. The relatively high cost of protein A resins compared to other non-protein based chromatography resins often leads to the operational strategy that a smaller protein A column is cycled several times while purifying a batch of cell culture supernatant. To be able to recycle the resin, contaminants need to be removed between purification cycles. This cleaning-in-place (CIP) includes addition of various cleaning agents. GE Healthcare has several protein A resins in their product portfolio of which MabSelect and MabSelect SuRe are two of the most recently launched. MabSelect is based on recombinant Protein A whereas the MabSelect Sure ligand was genetically engineered during development to obtain an alkali stabilized protein A derived resin. As a result, MabSelect SuRe can be cleaned-in-place using harsh conditions like high concentrations of the inexpensive chemical sodium hydroxide.
To get a cost efficient biopharmaceutical production process it is necessary to optimize process parameters. To maximize yield and purity, development and optimization of production units performed in a high throughput screening (HTS) format would be of great economical advantage. The development would then be performed in small scale demanding a minimal quantity of resources where combinations of resins and buffer
conditions could rapidly be screened. A HTS system requiring small amounts of material has been designed in a 96-well filter plate format. It can become even more efficient for screening if automated with a TECAN robotic platform. The screening is further optimized using Design of Experiments, which will generate maximum information from the performed investigations. Screening conditions can thereafter be verified in a column format, comparable to the full scale purification in an industrial mAb production process. The aim of this study was to examine the stability of protein A chromatography resins MabSelect and MabSelect SuRe to different CIP agents and to develop CIP procedures. The experimental set-up involved design of experiment compatible with the 96-well filter plate format, efficient preparations of CIP solutions, efficient and informative screening of CIP conditions in the 96-well plate format including comparisons between MabSelect and MabSelect SuRe. Thereafter, selected conditions from the plate study were verified in a parallel column chromatography format using the ÄKTAxpress platform.
2 B
ACKGROUND
2.1 B
IOTECH MANUFACTURING IS GROWING UP
The biotechnology industry has faced many challenges since it all started in the late 1970s. Aiming at the ability to transform new discoveries into viable therapies that could be produced at large scale and delivered to patients around the globe there was a need to overcome many obstacles. In addition to the difficulties concerning development of a production platform based on molecular and cellular biology, biopharmaceutical manufacturing has always been subject to unpredictable clinical trials, regulatory requirements, product approvals and market demand. Also, with the need for investments in expensive manufacturing facilities, before product approval, the challenges grow in dimensions of economics, technology, and logistics.
It is promising though that biotech corporations, and entire industries, in discrete stages approach a more predictable maturity. Eventually, this development will transform novel technologies into robust and economically viable platforms. A manufacturing revival where biopharmaceuticals can be produced consistently at high yields, shifts in market demand and development can be rapidly responded to, and lower investments in production infrastructure are at hand of the biopharmaceutical industry. The fundamentals of biotechnology industry will change dramatically by facility sharing and utilization and by advances in bioprocess engineering. The production cost and flexibility will approach the cost of small-molecule pharmaceuticals.
There are mainly three strategies that will make this development true: increase production yield through process development and biological science instead of
hardware solutions; standardization of facilities and processes; broad implementation of common platform technologies.
In the early years of biopharmaceutical production scientists had to evolve into engineers, inventing, from scratch, production equipment, control systems, and analytical technology. Thereby the core technology used for production and purification of protein-based drugs varied among different companies and production plants. With the initial focus on making new products, the process of developing an efficient and standardized production platform with clear regulatory requirements were left out. As the industry grew, co-operations between regulatory agencies and companies evolved in improved and standardized biopharmaceutical manufacturing, and development of purity and safety guidelines.
How far has the biopharmaceutical industry reached during its first thirty years? It all started with the arrival of recombinant DNA technology as a promising tool for mass production of therapeutic proteins. The first proteins developed into the biopharmaceuticals were insulin, human growth hormone, hemophilia proteins, and erythropoietin, functioning by replacing or reinforce biological pathways that had become dysfunctional through disease. These proteins were followed by the introduction of monoclonal antibodies, which could function as antagonist drugs, targeting biological molecules in a variety of disease pathways. Even if both types of biopharmaceuticals are of great importance it is the monoclonal antibodies that have been the driving technology in industry for the past decade.
As the market demands for newly released products increased during the 1990s companies focused on scale, constructing more plants, adding bioreactors and equipment. Those were the days of stainless steel. However, the unpredictability of production yields, regulatory approvals, and market demand made most company struggling with over- or under-capacity. For an inflexible industry racing after new technological breakthroughs, governed by regulatory requirements, answering to a shifting market can be difficult. [3] By development and implementation of standardized facilities and processes and new platform technologies for process development the biopharmaceutical industry can continue the maturing process.
2.2 L
EAN PRODUCTION IN BIOPHARMACEUTICAL INDUSTRY
The principle of lean operations is relatively straightforward to understand. It means moving towards the elimination of all waste in order to develop an operation that is more dependable, faster, produces higher quality products and services, and above all operates at low cost. The process to achieve a lean operation is less easily explained. Just-In-Time (JIT) is one approach to start with. It provides a cost-effective production and delivery of only the necessary quantity of products at the right quality, at the right time and place, while using a minimum amount of facilities, equipment, materials, and human resources. JIT is dependant on the flexibility of the supplier and customer of every process. [4] Manufacturing facilities for production of biopharmaceuticals take up to five years to build and validate. Thus, the flexibility of this industry is quite limited when it comes to instantly meeting market demand. In addition, large capital investments are needed to get the facilities online. If the market demand has changed or new technology
has been available during this time the facilities might need reconstruction and new validations need to be made to stay competitive on the market. Although, when it comes to using a minimum amount of resources the JIT approach can be applied of standardizing facilities and processes. By standardizing manufacturing and share facilities adapting contract production, the industry will be more flexible to shifts at the market. In addition, with less than one of five drug candidates surviving clinical trials to reach market, standardizing and sharing facilities could be less of a gamble. [3]
Some JIT methodology has already been implemented to the biopharmaceutical industry through standardization. For instance, at Wyeth, in addition to biology and process technologies, improvements in manufacturing efficiency have been achieved through improvements of workflow and platform implementation. Realization of the platform was driven by the fact that most products are antibodies and thereby most products have similar requirements regarding process development, optimization can thereby become more easily translated from one product to the next. In this way raw materials, production equipment, quality procedures, fermentation technology and methods as well as batch monitoring and documentation, are generally kept consistent. [3]
Process development for the biopharmaceutical industry can be performed in the spirit of JIT using high throughput screening technologies. While the material and equipment is expensive, a screening can be performed in a 96-well filter plate, in which various conditions can be investigated simultaneously. The screening results can thereafter be verified using an automated system like ÄKTAxpressTM, claiming minimal resources in
form of time and operational control. A platform of high throughput technologies would make process development a leaner process.
2.3 A
NTIBODIES
Antibodies defend mammals against infection by binding to bacteria, viruses and microbial toxins, thereby inactivating them. The binding of antibodies to foreign pathogens recruits the complement system and various types of white blood cells and this in turn leads to destruction of invading microorganisms and parasites. Antibodies are synthesized by B lymphocytes in millions of variants, each with different amino acid sequences in the variable parts and consequently different binding sites for antigens, i.e. foreign substances. Antibodies, or immunoglobulins (Ig), are one of the most abundant proteins in the blood. In mammals five classes of Igs are produced, each of which mediates a characteristic biological response due to antigen binding. The classes differentiate in amino acid sequences and number of domains in the constant region of the heavy chain [see Figure 2-1], [2]. The basic structure of an antibody molecule constitutes of four polypeptide chains, two identical light chains and two identical heavy chains, held together by disulfide bonds. The antibody consists of two identical halves, each with one antigen binding site. There are five classes of immunoglobulins; IgA, IgD, IgE, IgG and IgM of which A and G are further divided into subclasses. For human IgG these are named IgG1, IgG2, IgG3, and IgG4. [5]
Fvvariable region; antigen binding
site
Protein A binding site
CH3 CH3 CH2 CH2 CH1 CH1 VH VH VL VL CL CL - S S- - - S S- - - S S -- S S -Fc Fab Fvvariable region; antigen binding
site
Protein A binding site
CH3 CH3 CH2 CH2 CH1 CH1 VH VH VL VL CL CL - S S- - - S S- - - S S -- S S -Fc Fab
Figure 2-1: The hIgG antibody consists of two heavy and two light chains. The heavy chain on its part is built up of one variable (VH) and three constant (C1
H-C3H) regions while the light chain is composed of
one variable (VL) and one constant (CL) region.
IgM is the first antibody to be produced in an immune response after exposure to antigen. IgA is the major representative class in external secretions and intestinal mucus and serves as a first line defence against bacterial and viral antigens. IgD is the main antibody on the surface of B-cells but its role is not yet known. IgE is important in protection against parasites but also induce allergic reactions. IgG is the most stable class and has a half-life in serum of 20 days, 3-4 times longer than IgM and IgA. It is the antibody present in the highest concentration in serum and the one and only able to be transported across placenta for foetal protection. [6]
The binding antigen-antibody is reversible while mediated by the sum of relatively weak forces which are effective only in very short distances. Most large molecules, including virtually all proteins and many polysaccharides, can serve as antigens. In most cases the surface of antigens has different features, epitopes, to which antibodies interact. Heterogeneous mixtures of immunoglobulins, each specific for one epitope, are called polyclonal antibodies. Monoclonal antibodies are all identical and recognize one specific epitope. [7]
2.3.1 E
VOLUTION OFT
HERAPEUTICA
NTIBODIESIn 1975, the hybridoma technique developed by Köhler and Milstein, revolutionized the production of antibodies and mAbs could now be developed for therapeutic use. The technique, which was awarded the Nobel-prize, implies propagation of a clone from a single antibody secreting B lymphocyte and fusion with an immortal cancer cell resulting in clones of an immortal B lymphocyte, producing copies of antibodies. These cells, called hybridomas, propagate as individual clones each of which is a permanent and reliable source of monoclonal antibodies. Once a mAb has been constructed it can be used both as a probe, for tracking down and localizing its antigen and for purification of its specific antigen. The most important benefit of the hybridoma technique is that mAbs can be
produced against any protein even if it constitutes only a minor fraction of a biological sample. [5]
The evolution of monoclonal antibody engineering technology has progressed from murine to fully human antibodies where the latter dominates the R&D focus of the market. Murine antibodies stem to 100% from mice and was the first hybridoma product. During investigation of murine antibodies as therapeutic agents either imitating natural human antibodies or blocking biochemical pathways it was found causing serious immune responses in humans called HAMA (Human Anti-Mouse Antibodies). To overcome these reactions part human-part mouse antibodies, called chimeric antibodies, were developed. These consisted of 30-35% mouse sequences and 65-70% human. The chimeric mAbs proved to be by far more successful than murine, not causing immune responses in some cases. Although, the risk of immunogenicity reactions called HACA (Human Anti-Chimeric Antibodies) were still large and the engineering techniques advanced to humanized antibodies with 90-95% of human origin and 5-10% murine. Thus these antibodies have their advantages with lower immunogenecity risk there are also some limitations. The development is expensive and time consuming, and most importantly there is a risk of reduced binding affinity due to murine sequences involved in binding strength. The evolution progressed to the fully human antibodies containing no murine sequences and consequently the HAMA risk is totally eliminated. Being fully human, the antibodies will be eliminated from the human body at a slower pace, which in turn reduces the frequency and required amount of dosing. However, the differences between humanized and fully human mAb is small, the latter has not the limitations of the humanized mAbs. [8] All recombinant antibodies so far developed for therapeutic applications have been of the IgG class [2].
2.4 M
ABS AS BIOPHARMACEUTICALS OF TODAY AND FUTURE
About 20 years after launching the hybridoma technique, mAbs acceptable to the human immune system were successfully developed. There are currently more than 370 biotech products and vaccines in clinical trials targeting more than 200 diseases. Approximately 30% of these biopharmaceuticals is represented by antibodies and Fc-fusion proteins. [2] Today there are 17 mAb drugs on the market bringing new hope in future treatment of incurable diseases. An additional six drugs are in the final stages of clinical trials, and with about 60 companies active in the field there are approximately 150 new products in the pipeline. The therapeutic mAbs have emerged as an important and rapidly growing new class of drugs for treatment of several human diseases and has outpaced traditional pharmaceuticals in both sales and regulatory approvals. [1] Genentech showed sale results for two antibodies Herceptin and Avastin of $1.6 and $1.2 billion respectively. With this progress of first generation mAbs at hand Datamonitor predicts that worldwide annual sales by 2010 will reach $30 billion. [1] Cancer and its related conditions have by far the greatest focus in the development of new therapies. Infectious- and autoimmune diseases, and AIDS/HIV infection associated conditions follows. [9]
In the future the diversity between drugs might increase, facilitating the treatment of more multifaceted diseases. This will for instance progress through development of
therapeutic polyclonal antibodies that can bind to multiple epitopes and thereby target complex diseases. [1] In addition, continued research and technical advancement in gene– and cell therapy, proteomics and genomics are likely to support the development towards a more diverse range of biopharmaceuticals in the future. However, the product section of recombinant proteins and therapeutic antibodies will remain significantly unchanged at least for the rest of this decade. [10] Although, in the future the lifespan of a mAb drug might look different from a traditional pharmaceutical. If the cell line of the innovator’s plays a key role in determining the characteristics of the mAb drug it will become difficult for another manufacturer to recreate a generic version of the same mAb. It will be harder to develop generics and it is more probable that new kinds of technology, like antibody fragments, will be used to achieve the same effects as monoclonals. [1]
2.5 C
HROMATOGRAPHY IN GENERAL AND AFFINITY
CHROMATOGRAPHY ABOVE ALL
Chromatography is a group of separation techniques characterized by a distribution of molecules separated between two phases, a stationary and a mobile phase. The separation of molecules is an effect of which a particle with a high tendency to stay in the stationary phase will move through the system with a lower velocity than one that prefers the mobile phase. The most common physical arrangement of chromatography is the column chromatography where the solid phase is packed into a tube or column through which the mobile phase is pumped. The sample of molecules to be separated is applied to one end of the column. Since the molecules travel with various velocities through the packed bed they will separate and exit the column as distinct peaks which can be subsequently detected and collected.
In biological processes biospecific interactions might occur between proteins and low molecular weight substances, e.g. substrates and enzymes, but even more frequently between two or more biopolymers, especially proteins. Affinity chromatography is based on the phenomenon of substances’ specific affinity to adsorb to certain solid phases. One of the two interacting molecules, the ligand, is immobilized on the solid phase whereas the other molecule, the counterligand, usually a protein, is present in the mobile phase passing through the column. Affinity chromatography is in many cases a very powerful purification method due to its high specificity, particularly when the protein of interest is a fraction of a complex mixture.
A good ligand is a requirement for successful affinity chromatography which results in the fact that its characteristics are interesting parameters for method development and optimization. First the ligand must be able to form reversible complexes with the protein to be separated otherwise the protein can not be eluted and purified. In addition, to be able to receive the desired degree of product purity the specificity of the ligand must be compatible with the application in point. Another issue of great importance is that the ligand-protein complex should be stable in a constant environment. Also the complex should be easily dissociated by a simple change in the environment. Finally, the ligand should be easy to immobilize to a matrix, commonly to agarose. When having a protein-ligand pair fulfilling previous characteristics other parameters like pH and salt
concentration needs to be optimized to obtain a good purification using affinity chromatography. [11]
To facilitate the purification using affinity chromatography there are today active gels available on the commercial market. These consist of ligand immobilized matrices ready to interact with specific proteins.
2.5.1 P
ROTEINA
CHROMATOGRAPHYLigands can vary in their specificity, from extremely selective to group specific. An example of the latter is the interaction between staphylococcal protein A and immunoglobulin G, an affinity that has proven to be very helpful when put into practice in separation issues. Protein A, naturally produced in Staphylococcus aureus, is a 42 kDa large protein built up of five domains; E, D, A, B and C. The protein is a group-specific ligand having high specificity for IgG and is consequently used in purification of monoclonal IgG for clinical applications. Protein A binds to the Fc region of the antibodies and the binding site is located between the C2H and C3H region, see Figure 2-1. The
interactions are mainly hydrophobic, stabilized by hydrogen bonds.
Two protein A derived affinity media accessible on the market are MabSelectTM and
MabSelect SuReTM, produced by GE Healthcare.
2.5.2 M
ABS
ELECT ANDM
ABS
ELECTS
UR
E CHARACTERISTICSMabSelect is a family of protein A affinity media developed by GE Healthcare for purification of monoclonal antibodies. To achieve a higher compatibility with the industrial mAb production of today, protein A has been specially engineered to favor an oriented coupling resulting in an enhanced binding capacity for IgG. The specificity of recombinant protein A binding to the Fc region of IgG is similar to that of native protein A, resulting an excellent purification of monoclonal IgG in one step. [12], [13]. To elucidate, the ligand of MabSelect is composed of the five domains of recombinant protein A.
Part of the MabSelect family is MabSelect SuRe (Superior Resistance) consisting of an alkaline stablized protein A derived chromatography media. The SuRe ligand is composed of four modified B domains where the alkaline sensitive amino acids of the native B domain have been substituted by alkaline stabilized amino acids by site directed mutagenesis. [12] The ligand is at a C-terminal cysteine immobilized to a highly cross-linked agarose matrix via a chemically stable thio-ether linkage [14]. The alkali tolerance of the modified B domain, allows that the media can be cleaned-in-place with 0.1-0.5 M sodium hydroxide (NaOH). It has been shown that MabSelect SuRe can be cleaned for more than 100 CIP cycles using 0.1 M NaOH and a contact time of 15 minutes, with remained dynamic binding capacity [Figure 2-3]. MabSelect SuRe has a pH working range of 3-12, is temperature stable between 4-40 °C, and has an average agarose particle size of 85 um. The dynamic binding capacity is approximately 30 mg human IgG/mL media at a residence time of 2.4 minutes. [See section 2.6], [12]. MabSelect has a pH working range
of 3-10, is temperature stable between 4-40 °C, and has the same agarose particle size dynamic binding capacity as MabSelect Sure. [13]
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
alkaline stabilized domain protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab-cys
Zstab M X C B A D S
Zstab Zstab ZZstabstab Zstab
E
alkaline stabilized domain
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
IgG binding domains
protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E
alkaline stabilized domain protein A -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab
E -cys Zstab M X C B A D S
Zstab Zstab Zstab Zstab-cys
Zstab M X C B A D S
Zstab Zstab ZZstabstab Zstab
E
alkaline stabilized domain
Figure 2-2: Site specific mutagenesis of domain B of protein A yields an alkaline stabilized variant. A tetramer of the modified B domain, named Z domain constitutes the active ligand of MabSelect SuRe. The amino acid cysteine is the anchorage to the base matrix through a single point attachment. [15]
Number of CIP cycles
0 20 40 60 80 100 120 DB C at 1 0% b re ak th rou gh (p ol yc lo na lh Ig G ) [ % ] 0 20 40 60 80 100
15 min contact time with 0.1 M NaOH / cycle 60 min contact time with 0.1 M NaOH / cycle 15 min contact time with 0.5 M NaOH / cycle 15 min contact time with 0.1 M NaOH / cycle (conventional recombinant Protein A resin)
Number of CIP cycles
0 20 40 60 80 100 120 DB C at 1 0% b re ak th rou gh (p ol yc lo na lh Ig G ) [ % ] 0 20 40 60 80 100
15 min contact time with 0.1 M NaOH / cycle 60 min contact time with 0.1 M NaOH / cycle 15 min contact time with 0.5 M NaOH / cycle 15 min contact time with 0.1 M NaOH / cycle (conventional recombinant Protein A resin)
Figure 2-3: Dynamic binding capacity at 10% breakthrough for MabSelect SuRe packed in 5×100 mm column after cleaning-in-place with 0.1 or 0.5 M NaOH, 15 or 60 minutes contact time, for 0-120 cycles.
2.6 B
INDING CAPACITY AND TIME OF RESIDENCE
The binding capacity of an affinity adsorbent is defined as milligrams or micromoles of counterligand that can be adsorbed per milliliter of sedimented gel. However, the theoretical maximum capacity can not be achieved for several reasons and one therefore distinguishes between static and dynamic capacity. On the one hand static capacity is measured in batch experiments, which allow sufficient time for equilibration between solid and mobile phases to establish. Thus, the static capacity depends on the density of the immobilized ligand and its availability for interaction within the gel matrix. Consequently, the functional binding capacity is often lower than the binding capacity calculated from the measured ligand density. On the other hand dynamic binding capacity (DBC) of the affinity adsorbent is the binding capacity under operating conditions, i.e. in the packed affinity chromatography column during the sample application at a given flow velocity. [11]
During operation, when starting to load antibodies onto the column, the resin will bind proteins until it is saturated, or overloaded, i.e. during loading of protein A affinity columns, no antibody is present in the fluid exiting the column until the DBC is reached and the antibody begins to flow through. In the studies of this thesis, DBC was calculated as the amount antibody bound at the antibody breakthrough. Antibody breakthrough was determined when Ci/Co = 0.1, i.e. DBC at 10% breakthrough. Ci is the concentration of
antibody in the column effluent [g/L], and Co is the concentration of antibody in the load
[g/L]. When purified antibody is loaded, Ci/Co of the flow through can be monitored by UV
Figure 2-4 is a batch experiment where the capacity of MabSelect resin is plotted versus time of incubation. The gradient is large for shorter incubations but when the time is increased the amount ug antibody the resin can bind per uL gel is brought towards a state of equilibrium. The capacity is thereby depended of for batch experiments, incubation time or, for column formats, the time of residence. See Figure 2-5.
Incubation capacity uptake curve of MabSelect versus time of incubation
0 10 20 30 40 50 0 10 20 30 40 50 60 70 80 90 100 Time (min) In cu ba ti on capa ci ty (u g A b/u L gel
) Incubation capacity uptake curve of
MabSelect versus time of incubation
0 10 20 30 40 50 0 10 20 30 40 50 60 70 80 90 100 Time (min) In cu ba ti on capa ci ty (u g A b/u L gel )
Figure 2-4: Batch experiment performed in 96-well filter plate at GE Healthcare. The binding capacity of MabSelect displayed in ug antibody per uL chromatography media versus, in minutes, incubation time, i.e. the time the antibody and the ligand are allowed to integrate with one another.
DBC ( at 10% b rea kt hr ou gh ) m g /m L 1.5 2 3 4 6 8
Antibody residence time (min) 10
20 30 40 50
The dynamic binding capacity vs residence time (purifiedhIgG11.0 mg/mL)
DBC ( at 10% b rea kt hr ou gh ) m g /m L 1.5 2 3 4 6 8
Antibody residence time (min) 10 20 30 40 50 DBC ( at 10% b rea kt hr ou gh ) m g /m L 1.5 2 3 4 6 8
Antibody residence time (min) 10
20 30 40 50
The dynamic binding capacity vs residence time (purifiedhIgG11.0 mg/mL)
Figure 2-5: Dynamic binding capacity at 10% breakthrough versus residence time performed on
MabSelect SuRe using purified hIgG1, 1.0 mg/mL.
The DBC increase with the contact time between ligand and antibody.
To conclude, the binding capacities in batch mode compared to column experiment are not identical while conditions are not the same. Although, the binding capacities can become comparable by adjustment of residence time closer to a state equilibrium.
2.7 G
ENERAL DESCRIPTION OF A THERAPEUTIC ANTIBODY
PRODUCTION
In today’s biopharmaceutical industry, monoclonal antibodies are produced for a broad range of diseases. Most mAb therapies require high doses over a long period of time which results in large amounts of purified product per patient. It is therefore a true challenge to develop a manufacturing capacity meeting these demands. In this section general upstream and downstream processes in mAb production and purification will be described and also process units in need of optimization will be noted. A general process is visualized in Figure 2-6.
Virus inactivation Diafiltration Bioburden Reduction Mobile Storage Clarification Diluant Retentate Depth Filtration Protein A
AIEC Dead-end Filtration CIEC
Diafiltration Fermentation Filtration Virus inactivation Diafiltration Bioburden Reduction Mobile Storage Clarification Diluant Retentate Depth Filtration Protein A
AIEC Dead-end Filtration CIEC
Diafiltration Fermentation
Filtration
Figure 2-6: The manufacturing process of monoclonal antibodies including upstream units with cell culture to the downstream process with filtration and purification in three chromatographic steps, polishing and formulation. Chromatographic units are enclosed; first protein A, followed by anion exchange chromatography (AIEC) and cation exchange chromatography (CIEC).
In production of mAbs recombinant mammalian cell culture is the most popular expression system where Chinese hamster ovary (CHO) and murine myeloma (NS0) are the most widely used cell lines. The cell culture process starts with thawing a fraction of a working cell bank (WCB) expanded by a series of seed trains in different culture vessels. The cell culture is thereafter transferred to a production bioreactor where the cells continue to grow and expressed product accumulates in the culture broth. In order to achieve a high product titer and high cell mass the environment in the bioreactor needs to be maintained. Fed-batch where concentrated nutrients are fed to a batch is the most popular process mode due to its scalability. Bioreactors of stainless steel are still the major choice for large scale production but disposable bioreactor systems have also become available. An example is the Wave Bioreactor® which uses a plastic bag preferably during seed culture expansion. Disposable bioreactor systems have the benefits of eliminating the need of CIP and by reducing the large capital investments in stainless steel facilities. Improvements at this stage of the process can be performed by selection of highly producing cell lines, and optimization of medium composition and
bioreactor operating conditions. For enhancing the selection of high performance clones, high throughput screening technologies have been developed.
The cell culture process with high product titer generates a heavy load on the downstream process due to both non-product impurities and product-related isomers. Therefore, the downstream process is built up of various filtrating, purifying, and polishing units. The product in the culture broth is separated from the cells through continuous
centrifugation and filtration resulting in a cell-free supernatant. The following protein A affinity chromatography unit is the most efficient purification step for antibodies, resulting
a product purity of more than 98%, removing most process impurities. There are two major protein A resins available on the market: Agarose-based resin from GE Healthcare and glass bead-based resin from Millipore. The resins have different characteristics; a low non-specific binding in agarose-based media but lower backpressure in glass bead-based resins. Due to the great importance of the protein A capture step the type of resin is an option of great significance. In addition, sample loading time, pH, washes, and elution are important choices for minimizing host cell impurities (HCP) and ribosomal DNA levels. Two or three polishing steps are usually required to remove impurities like high molecular weight aggregates, trace amounts of HCP, leached protein A and potential viral contaminants. Commonly used for polishing are cation exchange chromatograpy (CIEC),
anion exchange chromatograpy (AIEC), and hydrophobic interaction chromatography
(HIC). CIEC has proven to be a good tool in order to remove product related impurities, AIEC for removal of trace amounts of viruses and impurities like DNA and endotoxins. HIC is an efficient mode to remove aggregates but since the technique generates a relatively low yield this technique is becoming less popular in antibody purification. [17]
2.7.1 T
HEP
ROTEINA
CAPTURING UNITProtein A chromatography serves as the capture step in the purification of monoclonal antibodies. It also function as the key volume reduction step in the process since protein A concentrates the product stream, from relatively dilute cell culture supernatant to the eluate at typically more than 10 g/L. It is highly selective for mAbs and yields a purity of usually more than 98%. This purification step is visualized in the chromatogram in Figure 2-7.
The cell culture supernatant can be directly loaded on the column at neutral pH and eluted at low pHs. A wash step between load and elution is often introduced to remove HCPs and other contaminants. Finally, the column is stripped, using an acidic solution of pH 2-3, and regenerated with either urea or guanidine hydrochloride or, NaOH if the resin is alkali-tolerant. The protein A ligand is to a low extent cleaved by proteases present in the cell culture supernatant and trace amounts of leached protein A is co-eluted with the mAb. Even if protein A delivers great purity it suffers from several limitations compared to i.e. ion-exchange chromatography media. The relatively high cost of protein A resin compared to non-protein based chromatography resins, often leads to the operational strategy that a smaller protein A column is cycled several times while purifying a batch of cell culture supernatant. This makes the column loading of protein A usually the rate limiting step as a large volume is loaded to a small column. Another challenge is that
elution needs to be carried out at low pH which can result in the formation of soluble high molecular weight aggregates. This in turn might lead to reduction in product yield and add to the burden of the following polishing steps in the purification process. [18]
A280 Flow pH 0.0 0.5 1.0 1.5 ml/min 4.0 6.0 8.0 10.0 12.0 0 50 100 150 min EQU ILI BRATI O N START L O AD IN G F EE D ST O C K W A SH O U T UN BO UN D MA B ST AR T COL LE CTING F RACTION STAR T C IP ST O P C IP A280 Flow pH 0.0 0.5 1.0 1.5 ml/min 4.0 6.0 8.0 10.0 12.0 0 50 100 150 min EQU ILI BRATI O N START L O AD IN G F EE D ST O C K W A SH O U T UN BO UN D MA B ST AR T COL LE CTING F RACTION STAR T C IP ST O P C IP
Figure 2-7: Purification of clarified cell culture
supernatant with 0.98 mg hIgG1 per mL run on
MabSelect SuRe packed in a 5 × 100 mm column. Sample residence time was 2.4 min. The cycle includes equilibration with loading buffer, sample loading, post load wash, elution with low pH and CIP followed by re-equilibration. CIP with 0.5 M NaOH, contact time 30 min
can be done every 10th purification cycle. Frequent CIP
every purification cycle with 0.1 M NaOH, contact time 15 min.
2.7.2 C
LEANING-
IN-
PLACECleaning-in-place is the removal of very tightly bound, precipitated or denatured substances from the chromatography resin. If not removing such contaminants they might affect the chromatographic properties of the column, decreasing binding capacity and come off in subsequent runs which results in carryover of contaminants or product between cycles. If this fouling is severe it may cause increased back pressure and reduced flow rate.
A standard CIP protocol for MabSelect, recommended by the GE Healthcare, is as follows: 1) Wash the column with 2 column volumes (CVs) of 6 M guanidine hydrochloride (10 min) or 10 mM NaOH (30 min). Other possible CIP compositions are 50 mM NaOH in 1 M NaCl (16 min) or 50 mM NaOH in 1 M Na2SO4 (16 min). 2) Wash immediately with at least 5 CVs
of sterile and filtered binding buffer at pH 7-8 with reversed flow direction. [13] One standard CIP protocol for MabSelect SuRe is as follows: 1) Wash the column with 3 CVs of binding buffer. 2) Wash with at least 2 CVs of 0.1-0.5 M NaOH (10-15 min). 3) Wash immediately with at least 5 CVs of sterile and filtered binding buffer at pH 7-8. Depending
on the nature of contaminants, i.e. the feed material, the CIP protocol of the two chromatography resins can be optimized. Concentration of NaOH, contact time and frequency are parameters interesting for optimization but as a general advice the column should be cleaned-in-place at least every fifth cycle. [19]
2.7.3 A
GENTS POSSIBLE FOR STABILIZATION OF THE PROTEIN A LIGANDCIP can be aggressive on the ligand of the chromatography resin due to site-specific chemical cleavage. It is therefore interesting to evaluate whether an addition of specific agents might stabilize the ligand resulting in maintained binding capacity. Additions of salts have been tested in previous studies performed at GE Healthcare, see Figure 2-8.
Comparison between capacity using different CIP solutions
82.0 84.0 86.0 88.0 90.0 92.0 94.0 96.0 98.0 100.0 0 50 100 150 200 250 300 350 Number of cycles Rem ain in g c apac it y (%) 50 mM NaOH+1M NaCl nr1 50 mM NaOH+1 M NaClnr2 50 mM NaOH without NaCl 100 mM NaOH+1M NaCl 10 mM NaOH+1M NaCl
Figure 2-8: Comparison between capacity after CIP-treatment of two batches with 50 mM NaOH with and without 1 M NaCl. A stronger solution (100 mM NaOH+1 M NaCl) is included as a control.
When using CIP solutions containing NaOH the stability of the MabSelect media is increased if sodium chloride (NaCl) is included. It also has the effect of weakening ionic interactions which, in theory, helps reduce contaminants. After 300 CIP cycles with 10 mM NaOH and 1 M NaCl the resin had 97% of the original binding capacity left. This, though, might be a solution too weak to have a good cleaning ability. Being more concentrated with respect to NaOH a more effective CIP was achieved using 50 mM NaOH and 1 M NaCl, which resulted after 300 cycles in remaining binding capacity of 89%. The positive effect of including NaCl was clearly demonstrated as a CIP solution containing 50 mM NaOH and no NaCl rapidly deteriorated the media: after only 61 cycles the capacity was down to 84%. Using more concentrated solutions of NaOH (still including 1 M NaCl) had a more drastic effect on the capacity. A CIP solution containing 100 mM NaOH + 1 M NaCl clearly had an impact on capacity, which was 84% after 163 CIP cycles. Being more concentrated this solution should have a better cleaning ability. This solution would be attractive if the longevity of the media is not an issue. These stronger solutions, though, might be used as CIP every 5th cycle. The same increase in binding capacity using addition
of 1 M NaCl to the CIP solution was achieved using 0.5 M Na2SO4. Thus, only half the
amount of Na2SO4 is needed to obtain the same results using NaCl.
Amino acids can have wide applications in protein research and biotechnology product development. Naturally they protect cells in various organisms from water stress like high salt concentration or freezing conditions. They are compatible with most protein’s function and activity by stabilizing but not binding to them. Most proteins are fairly unstable in aqueous solutions which might lead to unfolding and aggregation and certain amino acids can in turn be used for stabilization of proteins in solution. [20] However, arginine differs from other amino acids, although the reason for this is not well understood. Arginine does not affect the native structure of proteins, although it does not stabilize them like other amino acids do. Previous studies show that aggregation of unfolded proteins induced by increasing temperature occurs to a lower extent in the presence of arginine. [21] In addition, it has extensive use in refolding of recombinant proteins and can be an effective amino acid in production processing and characterisation of proteins, including antibodies. [20]
2.7.4 MA
B PURIFICATION USINGM
ABS
ELECT VERSUSM
ABS
ELECTS
UR
EMabSelect and MabSelect SuRe are part of the same family of recombinant protein A affinity media [See section 2.5.2]. They have the same specified affinity for IgG but differ in the successful modifications performed at the B domain of MabSelect SuRe, modifications which resulted in a more alkali-resistant ligand. Thus, the active ligand in MabSelect is the five domains of recombinant protein A, and in MabSelect SuRe a tetramere of modified B domains.
Previous studies have shown that NaOH is an effective tool in removal of proteins and nucleic acids. It is also effective for inactivation of most viruses, bacteria, fungi, yeasts and endotoxins. Sodium hydroxide is widely accepted in industrial manufacturing as a cleaning agent thanks to its effectiveness and low cost and it is common practice to add sodium chloride to the NaOH solution. The ability of protein and nucleic acids elimination depends on the type of media, sample and contaminants. If the substances to be removed are for instance lipids bound to a protein, higher concentrations of sodium hydroxide may be required. Regarding to the inactivation of viruses and prions tests run at Q-ONE Biotech Ltd. showed the inactivation of eight different viruses using 0.1 and 0.5 M NaOH, a few of these where highly resistant non-enveloped viruses. In addition, NaOH has been proven to be an effective agent in inactivation of prions giving raise to the BSE (bovine spongiform encephalopathy) disease in cattle. [22] Other alternate agents for elimination of Bacillus spores are for instance peracetic acid, chlorine dioxide hydrogen peroxide and gamma radiation etc but the affect on MabSelect and MabSelect SuRe has not yet been examined. [23] MabSelect SuRe is designed to withstand cleaning in 0.1-0.5 M NaOH while MabSelect demands alternative cleaning solutions like 6 M guanidine hydrochloride (HCl) in 10 mM NaOH [19], [13]. An estimation of the cost of usage of guanidine HCl as CIP is summarized in Figure 2-9. The column volume is 230 L and for each CIP 2 CVs is needed resulting in 460 L of guanidin HCl for each CIP cycle. Since 140 purification cycles and CIP every fifth, results in about 13000 L CIP chemicals, the total cost of using guanidine HCl is about £1,400,000 per campaign and £100/L or $50/L. [24] The total cost includes purchase, handling and disposal. The numbers origin from 2004
but still the difference in total cost of CIP using guanidin HCl versus sodium hydroxide is significant. The cost of using NaOH is estimated to about $3/L. In addition, guanidin HCl claims extra care during handling while it is a viscous crystal forming chemical. Consequently, NaOH as an effective CIP agent to a low cost is in advantage of MabSelect SuRe.
Figure 2-9: The cost (in £) of CIP using guanidine HCl versus the number of purification cycles. CIP is
performed every 5th purification cycle which gives 28 CIP cycles. The CV is 230 L and each CIP cycle takes 2
CVs, i.e. 460 L. For 140 purification cycles about 13000 L of CIP is needed resulting in a cost per L of about £100 or $50. [24].
2.8 H
IGH THROUGHPUT SCREENING FORMAT
Protein A affinity chromatography is today widely used in the purification of monoclonal antibodies. To maximize yield and purity in a production process, development and optimization of production units performed in a high throughput screening (HTS) format would be of great economical advantage. The development would then be performed in small scale, demanding a minimal quantity of resources where combinations of resins and buffer conditions could rapidly be screened. GE Healthcare has developed a High Throughput Process Development (HTPD) platform constituting of a 96-well filter plate which enables High Throughput Screening of process conditions. The format can become even more resource efficient if automated with a robotic platform. [25] The screening is
further optimized using design of experiments, which will generate maximum information from the performed investigations.
The process development using HTS tools can be described as a variant of the PDCA-cycle, first described by the quality guru W.E Deming. PDCA stands for Plan-Do-Check-Act and is a model of an improvement cycle. Influences from PDCA produced the more experimental version; the DMAIC cycle (Define-Measure-Analyze-Improve-Control) which became popular by the Six Sigma approach. Process development and optimization starts with planning the study, defining the problem area and designing the experimental set-up. Next step is to perform the study according to the experimental plan. Screening can be performed in a 96-well filter plate, utilized by a robotic platform like TECAN. The third step is to analyze what the first two steps resulted in, if the response was the expected and to analyze which conditions are the more interesting to proceed investigating. The chosen screening conditions are further evaluated using statistical software and if the experimental set-up needs to be narrowed or if the design was not sufficient, another cycle for further optimization or new experimental design can be performed in the filter plate. When having interesting and practically doable conditions, the set-up should be verified in a column format. [4], [26]
Optimize
Plan
Predict
Perform
Analyze
Verify
Evaluate
Pl at e form at Co lu m n f or m at X2 X1 X3 -1 1 -1 1-1 1 OptimizePlan
Predict
Perform
Analyze
Verify
Evaluate
Pl at e form at Co lu m n f or m at X2 X1 X3 -1 1 -1 1-1 1 OptimizePlan
Predict
Perform
Analyze
Verify
Evaluate
Pl at e form at Co lu m n f or m at X2 X1 X3 -1 1 -1 1-1 1Figure 2-10: A version of the PDCA-cycle (Plan-Do-Check-Act) used in High Throughput Process Development. The PDCA is performed in screening format using a 96-well filter plate and chosen conditions verified in column format. [26]
2.8.1 G
ENERAL DESCRIPTION OF SCREENING IN THE96-
WELL FORMATAs described above GE Healthcare has developed a HTPD platform constituting of a 96-well filter plate which enables HTS of process conditions. The 96-96-well filter plates filled with small volumes of chromatography resin can be used for determination of the static capacity of the resin by overloading the resin in the wells with protein solution and measure the amount of protein that is unbound; flows through. [27] The difference between loaded sample and flow through equals the amount of bound protein. The bound protein can also be eluted and by determination of the concentration of the eluted fraction the total amount of bound protein can be calculated.
Figure 2-11: Procedure for determination of binding capacity of chromatography resin in 96-well format. [27]
When investigations of buffer conditions are performed in the 96-well format a large number of various buffers can be evaluated. A manual preparation of these buffers is time consuming with the risk of physical effects due to a monotonous way of working. A prominent robotic platform has been developed by TECAN which offers a variety of applications regarding liquid handling. In cooperation with Millipore® an application note
presents a total automated procedure involving protein precipitation and filtration. A robotic arm moves the 96-well filter plates between locations of sample load and filtration, mixing during incubation is performed at a micro plate shaker and removal of loaded liquids is performed using vacuum. The application note promises the total processing time of 96 samples to about 18 minutes. [28]
Picture 2-1: TECAN is an automated system with numerous applications in liquid handling.
A co-operation between GE Healthcare and TECAN has resulted in a complete solution for making protein expression screening efficient and robust. The solution promises reliable one-step purification of tagged proteins from plate preparation and sample application to elution and analysis. Once conditions from the screening are optimized scale-up can be performed using the high throughput chromatography system ÄKTAxpressTM. [29]
2.8.2 H
IGH-
THROUGHPUT USINGÄKTA
XPRESSTMÄKTAxpress is a chromatography system where a number of modules can be run in parallel. This enables concurrent purification of multiple samples which increases the throughput. The system consists of a computer to which up to 12 modules can be connected [See Picture 2-2]. The software, named UNICORNTM, provides unattended purification with full documentation and evaluation. Not only is ÄKTAxpress valuable in protein purification but also for establishment of DBC and runs of CIP protocols. Buffers and samples are introduced into the system through sample and buffer inlets and flown through the column according to the protocol in point. Detection is performed on line by the UV cell which is measuring the absorbance of liquid passing through. The run data is sent to the computer and visualized in a chromatogram shown on screen. [See Picture 2-2, Picture 10-1]
The great benefit of a high throughput system like ÄKTAxpress is that, when various conditions need to be examined, the modules can be run simultaneously either following the same or different protocols. Each module downloads, when starting the run, the protocol from the system computer. No changes can be made when started the protocol and a possible computer breakdown will not affect the run. This makes the system more robust and useful while a possible failure in one module will not affect the others.
However, an electric failure will stop the runs while no back-up source of energy is accessible. Another aspect of high throughput is that the set-up time is reduced while the modules can be started one by one when preparations of each module are done. This optimizes the system utilization. The set-up time is defined as the time it takes to change over the process from one activity to the other. In ÄKTAxpress, this gain in time is particularly obvious when different protocols with different running times are processed. [4]
Picture 2-2: ÄKTAxpress system here with four modules. It is possible to connect up to 12 modules to one computer from which protocols are created and runs supervised and controlled. The modules can be run simultaneously either following different or the same protocols.
2.9 I
NTRODUCTION TO STATISTICAL EXPERIMENTAL DESIGN
How can the optimal performance of a system be found? Which factors have significant effects and which combination of these factors will give the best results of the system? Traditionally, experiments were performed by the approach “Change one thing at a Time” – the COST approach. The factors are then investigated separately until no further improvement of the system is accomplished. One great disadvantage of this approach is that no possible interactions between the factors can be discovered, even though the effect is highly significant for the result of the system. A consequence of this is the risk of failure in defining the true optimum of the system. Figure 2-12 shows a system where the circular markings are factorial levels and the black centre indicates the optimum of the system. The topographic curves indicate there are interaction effects in the system, effects that the COST approach will not find. [30]
a) b)
Figure 2-12: Experiments performed according to a) COST approach and b) DOE. Utilizing COST approach – “Change One Separate thing at a Time” it is not possible to detect any interactions between factors. The curvature in figure a) indicates that the factors are not independent, thus interaction effects exists. The true optimum of the system might therefore be missed using the COST approach. DOE is an approach considering interaction effects and the mathematical model will show the direction of how to reach the optimal settings of the factors of the system.[31]
While experiments are very costly and time consuming it is essential to reduce the quantity by producing a design consisting of the most interesting set of trials. A design of experiment (DoE) selects a number of experiments that will provide the maximum information of the system it describes. The number of trials can be tailored to meet the specific demands of the actual case. Thus, DoE reveals the optimal conditions of the system and which factors are influential. Thereby it provides a reliable basis for decision-making. [30-32]
2.9.1 D
ESIGN OFE
XPERIMENTS-
D
OE
When developing a new process or optimizing a current one a sophisticated technique called Design of Experiments (DoE) can be utilized. For more complex processes having several inputs affecting the process, the output, simple experiments are usually not sufficient but an experimental plan, a design is required. Since the inputs can vary both independently and in interaction with one another, the relationship of the output to the inputs need to be investigated. An example from the manufacturing industry is how temperature and flows can vary independently but also by interaction creating a pressure factor affecting the output. The traditional experimental approach, called COST approach by Umetrics [See chapter 2.9], where one factor at a time is varied and the output studied is usually unsuccessful in process optimization since the interaction effects are left out. On the contrary, utilizing DoE technique, non-linear, interaction and quadratic relationships can be examined. The outcome after having designed, performed and analyzed the results of a well defined DoE is a mathematical process model that predicts the response of all output variables for any combination of inputs. After analysis of each significant factor the model could, not only be used for optimization of the system, but also to troubleshoot the process if deviations occur in the future. [32]
To summarize, there are three types of problems to which design of experiments is applicable. First screening issues considering which factors are significant and in what ranges they influence. Second DOE is a good approach to find the optimum of a system. Third is to make the proper adjustments of the factors to guarantee robustness of the system. Design of experiments is utilized in many different industries where it is a popular approach concerning development of new products and processes, screening of important factors and minimization of production costs. [30]
2.9.2 R
ED-M
UP DESIGN–
HAND IN HAND WITH THE96-
WELL FORMATThe Rectangular Experimental Design for Multi-Unit Platforms (RED-MUP) is widely used for experimentation in for instance biochemistry, microbiology and pharmaceutical development. RED-MUP are custom designs developed for 96 (8x12), 384 (16x24) and 1536 (32x48) runs, which makes it very useful when designing for screening in 96-well filter plates. The RED-MUP design consists of two sub-designs corresponding to the vertical and the horizontal directions of the plate, i.e. 8 and 12, respectively, for a 96-well filter plate. The total design is made by multiplying the two sub-designs together. Hence, this total design supports a model with all interactions between the factors in the designs, plus the main effects from each design, and, if supported by the sub-designs, also interactions and quadratic effects.
When specifying the RED-MUP design it is important to distribute the factors over the two sub-designs so that the actual experimental protocol remains simple and practically doable. Another thing to keep in mind during the designing process is to choose designs that allow interaction and quadratic effects to be evaluated if expected. Also the final design needs to make chemical, biological and engineering sense. If interaction effects are expected between two factors these should be split into two sub-designs. This will allow the interaction effect to be estimated, regardless of choice of sub-design. [33]. The sub-designs used in the studies of this thesis were Full Factorial 2-level-, multi level-, Central Composite Circumscribed (CCC) and Central Composite Face Centered (CCF) designs. The creation of a multi level design is simple where the number of levels and their magnitude are chosen freely. Full Factorial 2-level design is shown in Figure 2-13 and the CCF design in Figure 2-14. By moving the red marbles placed at high and low levels of each factor, outside the experimental square, a CCC design is achieved.
A
B
A
B
Figure 2-13: Full Factorial 2-level design where the experiments performed are shown with black marbles. At least one but more often three experiments are executed in the centre point.
A
B
A
B
A
B
Figure 2-14: A CCF design where the experiments performed are shown with marbles. A CCC design is achieved by moving the face centred marbles, placed at high and low levels of each factor, outside of the square.
2.9.3 E
VALUATION INU
METRICSM
ODDE V.8.0
After an experimental design has been produced and the study executed, the achieved responses can be evaluated in MODDE, using different tools. The Summary of Fit plot in Figure 2-15 summarizes values describing how well the produced model predicts future data. In the 1st column to the left is R2 adjusted, which is the fraction of variation of the
response explained by the model with adjustment for the number of degrees of freedom. R2 adjusted is a measure of fit, i.e. how well the model fits the data. A large R2 adjusted is a necessary condition for a good model, but is not sufficient. One can have poor models, models that can not predict data, with a large R2 adjusted. This is particularly true when having few degrees of freedom for the residuals. A poor R2 adjusted origins from poor reproducibility, meaning poor control of experimental error or poor model validity, which might be an indication of an incorrect model. There is also R2, which is the fraction of variation of the response explained by the model but with no adjustment for the degrees of freedom. Since screening in 96-well filter plates using RED-MUP design delivers a great number of degrees of freedom R2 adjusted is used in the evaluations in section 5.
The 2nd column, Q2, is the fraction of the variation of the response predicted by the model
according to cross validation, i.e. Q2 estimates how good the model predicts new data. Thus, a useful model should have a large Q2 value. One will get a poor Q2 when the reproducibility is bad, meaning poor control of the experimental error and/or poor model validity, meaning the model might be incorrect. If the R2 adjusted value is good and the model validity is moderate a bad Q2 usually origins from insignificant terms in the model. R2 adjusted is always a value between 0 and 1, but Q2 can be negative for very poor models. Good models having both R2 adjusted and Q2 values close to 1 generally have excellent predictive power. [33]
