STRATEGIES FOR FACILITATED PROTEIN RECOVERY AFTER RECOMBINANT PRODUCTION

IN ESCHERICHIA COLI

MY HEDHAMMAR

Royal Institute of Technology Department of Biotechnology

Stockholm, 2005

2

ISBN 91-7178-176-5 © My Hedhammar School of Biotechnology Department of Biotechnology Royal Institute of Technology Albanova University Center SE-106 91 Stockholm Sweden Printed at Universitetsservice US-AB Box 700 14 100 44 Stockholm Sweden Cover picture: Model of the electrostatic potential of Zbasic. By Torbjörn Gräslund. 2000

3

My Hedhammar (2005): Strategies for Facilitated Protein Recovery after Recombinant Production in Escherichia coli. School of Biotechnology, Royal Institute of Technology, Stockholm, Sweden. ISBN 91-7178-176-5

ABSTRACT The successful genomic era has resulted in a great demand for efficient production and purification of proteins. The main objective of the work described in this thesis was to develop methods to facilitate recovery of target proteins after recombinant production in Escherichia coli. A positively charged purification tag, Zbasic, has previously been constructed by protein design of a compact three-helix bundle domain, Z. The charged domain was investigated for general use as a fusion partner. All target proteins investigated could be selectively captured by ion-exchange chromatography under conditions excluding adsorption of the majority of Escherichia coli host proteins. A single cation-exchange chromatography step at physiological pH was sufficient to provide Zbasic fusion proteins of high purity close to homogeneity. Moreover, efficient isolation directly from unclarified Escherichia coli homogenates could also be accomplished using an expanded bed mode. Since the intended use of a recombinant protein sometimes requires removal of the purification tag, a strategy for efficient release of the Zbasic moiety using an immobilised protease was developed. The protease columns were reusable without any measurable decrease in activity. Moreover, subsequent removal of the released tag, Zbasic, was effected by adsorption to a second cation-exchanger. Using a similar strategy, a purification tag with a negatively charged surface, denoted Zacid, was constructed and thoroughly characterised. Contrary to Zbasic, the negatively charged Zacid was highly unstructured in a low conductivity environment. Despite this, all Zacid fusion proteins investigated could be efficiently purified from whole cell lysates using anion-exchange chromatography. Synthesis of polypeptides occurs readily in Escherichia coli providing large amounts of protein in cells of this type, albeit often one finds the recombinant proteins sequestered in inclusion bodies. Therefore, a high throughput method for screening of protein expression was developed. Levels of both soluble and precipitated protein could simultaneously be assessed in vivo by the use of a flow cytometer. The positively charged domain, Zbasic, was shown also to be selective under denaturing conditions, providing the possibility to purify proteins solubilised from inclusion bodies. Finally, a flexible process for solid-phase refolding was developed, using Zbasic as a reversible linker to the cation-exchanger resin. Keywords: ion-exchange chromatography, protein A, Z, Zbasic, Zacid, fusion protein, proteolytic cleavage, immobilised protease, flow cytometry, inclusion bodies, solid-phase refolding

4

LIST OF PUBLICATIONS This thesis is based on the publications listed below. They are referred to in the text by their Roman number. I Gräslund, T., Ehn, M., Lundin, G., Hedhammar, M., Uhlen, M., Nygren, P.A.,

Hober, S. Strategy for highly selective ion-exchange capture using a charge-polarizing fusion partner J Chromatogr A (2002). 942, 157-66.

II Gräslund, T., Hedhammar, M., Uhlen, M., Nygren, P.A., Hober, S.

Integrated strategy for selective expanded bed ion-exchange adsorption and site-specific protein processing using gene fusion technology J Biotechnol (2002). 96, 93-102.

III Hedhammar, M., Gräslund, T., Uhlen, M., Hober, S. Negatively charged purification tags for selective anion-exchange recovery Protein Eng Des Sel (2004). 17, 779-86.

IV Hedhammar, M., Stenvall, M., Lönneborg, R., Nord, O., Sjölin, O., Brismar, H.,

Uhlén, M., Ottosson, J., Hober, S. A Novel flow cytometry-based method for analysis of expression levels in Escherichia coli, giving information about precipitated and soluble protein

J Biotechnol (2005). 119, 133-46. V Hedhammar, M., Alm, T., Gräslund, T., Hober, S.

Single-step recovery and solid phase refolding of inclusion body proteins by the use of a positively charged purification tag

Submitted, 2005

VI Hedhammar, M., Jung, H.R., Hober, S. Enzymatic cleavage of fusion proteins using immobilised protease 3C Submitted, 2005

Published articles are reproduced with the kind permission of the journals concerned. ADDITIONAL PUBLICATIONS NOT INCLUDED IN THE THESIS

Boström, M., Markland, K., Sanden, A.M., Hedhammar, M., Hober, S., Larsson, G. Effect of substrate feed rate on recombinant protein secretion, degradation and inclusion body formation in Escherichia coli. Appl Microbiol Biotechnol (2005). 68, 82-90. Hedhammar, M., Gräslund, T., Hober, S. Protein engineering strategies for selective protein purification Chemical Engineering & Technology (2005). In Press

5

Uhlén, M., Björling, E., Agaton, C., Al-Khalili Szigyarto, C., Amini, B., Andersen, E., Andersson, A-C., Asplund, A., Angelidou, P., Asplund, C., Berglund, L., Bergström, K., Cerjan, D., Ekström, M., Elobeid, A., Eriksson, C., Fagerberg, L., Falk, R., Fall ,J., Forsberg, M., Gumbel, K., Halimi, A., Hallin, I., Hamsten, C., Hansson, M., Hedhammar, M., Hercules, G., Kampf, C., Larsson, K., Lindskog, M., Lund, J., Lundeberg, J., Magnusson, K., Malm, E., Nilsson, P., Oksvold, P., Olsson, I., Ottosson, J., Paavilainen, L., Persson, A., Rimini, R., Rockberg, J., Runeson, M., Sivertsson, Å., Sköllermo, A., Steen, J., Stenvall, M., Sterky, F., Strömberg, S., Sundberg, M., Tegel, H., Tourle, S., Wahlund, E., Wan, J., Wernérus, H., Westberg, J., Wester, K., Wrethagen, U., Xu, L.L., Ödling, J., Öster, E., Hober S., and Pontén F. A human protein atlas for normal and cancer tissues based on antibody proteomics. Mol Cell Prot (2005) Aug 27, Epub ahead of print

6

ABSTRACT....................................................................................................................................3 LIST OF PUBLICATIONS...........................................................................................................4 INTRODUCTION..........................................................................................................................9 1. PROTEINS .................................................................................................................................9 2. RECOMBINANT PROTEIN PRODUCTION......................................................................11

2.1. HOSTS FOR RECOMBINANT PROTEIN PRODUCTION..............................................................11 2.1.1 Escherichia coli as Host for Recombinant Protein Production....................................14

2.1.1.1. Cultivation of Escherichia coli Cells ................................................................................. 16 3. DOWNSTREAM PROCESSING ...........................................................................................17

3.1. PROTEIN PURIFICATION .......................................................................................................17 3.1.1 Chromatography ..........................................................................................................17

3.1.1.1. Ion Exchange Chromatography......................................................................................... 20 3.1.1.1.1. Charge Properties of Proteins................................................................................... 20 3.1.1.1.2. The Stationary Phase................................................................................................. 21 3.1.1.1.3. Elution Strategies ...................................................................................................... 21 3.1.1.1.4. Features..................................................................................................................... 22

3.1.1.2. Expanded Bed Adsorption Chromatography ..................................................................... 22 3.2. PURIFICATION TAGS ............................................................................................................24

3.2.1 Protein Domains ..........................................................................................................25 3.2.1.1. Enzymes ............................................................................................................................. 26 3.2.1.2. Carbohydrate Binding Domains........................................................................................ 26 3.2.1.3. Protein Binding Domains .................................................................................................. 27 3.2.1.4. Streptavidin Binding Proteins............................................................................................ 28

3.2.2 Peptides........................................................................................................................29 3.2.2.1. Antigenic Epitopes............................................................................................................. 29 3.2.2.2. Protein Binding Peptides................................................................................................... 29 3.2.2.3. Streptavidin Binding Peptides ........................................................................................... 30 3.2.2.4. Metal Binding Peptides...................................................................................................... 31 3.2.2.5. Charged Peptides .............................................................................................................. 32

3.2.3 Functional Engineering of a Domain...........................................................................33 3.3. CLEAVAGE OF FUSION TAGS ...............................................................................................35

3.3.1 Specific Cleavage ........................................................................................................35 3.3.2 Cleavage Strategies......................................................................................................36

4. FOLDING AND MISFOLDING OF PROTEINS.................................................................38 4.1. FOLDING PATHWAYS...........................................................................................................38 4.2. FOLDING MODULATORS WITHIN CELLS...............................................................................39 4.3. INCLUSION BODIES ..............................................................................................................40

4.3.1 The Formation of Inclusion Bodies .............................................................................40 4.3.2 The Prevention of Inclusion Body Formation..............................................................41 4.3.3 Inclusion Bodies and Medical Science ........................................................................42

4.4. FOLDING SCREENS...............................................................................................................43 4.4.1 Gel Based Methods......................................................................................................43 4.4.2 Cell Response Systems ................................................................................................43 4.4.3 Reporter Systems .........................................................................................................43

4.5. IN VITRO REFOLDING ...........................................................................................................46 4.5.1 Isolation of Inclusion Bodies .......................................................................................46

7

4.5.2 Solubilisation of Inclusion Bodies ...............................................................................46 4.5.3 Refolding Strategies.....................................................................................................47

4.5.3.1. Refolding by Dilution......................................................................................................... 47 4.5.3.2. Refolding using Size Exclusion Chromatography.............................................................. 47 4.5.3.3. Solid Phase Refolding........................................................................................................ 48

4.5.3.3.1. Ion Exchange Chromatography used for Solid Phase Refolding............................... 49 4.5.3.3.2. Hydrophobic Interaction Chromatography used for Solid Phase Refolding ............. 49 4.5.3.3.3. Immobilised Metal Affinity Chromatography used for Solid Phase Refolding .......... 50 4.5.3.3.4. Affinity Chromatography used for Solid Phase Refolding......................................... 50

4.5.4 Additives......................................................................................................................50 PRESENT INVESTIGATION....................................................................................................53 5. IMPROVED PROTEIN RECOVERY BY THE USE OF CHARGED PURIFICATION

TAGS .......................................................................................................................................54 5.1. THE DEVELOPMENT OF NEGATIVELY CHARGED PURIFICATION TAGS (III) .........................54 5.2. ION EXCHANGE ANALYSIS OF Z VARIANTS (III) ..................................................................57 5.3. THE USE OF ZBASIC AND ZACID AS GENERAL PURIFICATION TAGS (I, III) ................................58

5.3.1 Zbasic .............................................................................................................................58 5.3.2 Zacid ..............................................................................................................................59

5.4. INTEGRATED PROCESS FOR SELECTIVE PROTEIN RECOVERY USING ZBASIC (II) .....................60 5.4.1 Cultivation ...................................................................................................................61 5.4.2 Purification ..................................................................................................................61 5.4.3 Cleavage ......................................................................................................................61

5.5. IMMOBILISED PROTEASE FOR CLEAVAGE OF FUSION PROTEINS (VI) ..................................62 6. NEW METHODS FOR ASSESSMENT, PURIFICATION AND REFOLDING OF NON-

SOLUBLE PROTEINS..........................................................................................................64 6.1. A FLOW CYTOMETRY BASED METHOD FOR ANALYSIS OF EXPRESSION LEVELS (IV) .........64 6.2. THE IMPACT OF ZBASIC ON PROTEIN SOLUBILITY (IV)...........................................................67 6.3. PURIFICATION UNDER DENATURING CONDITIONS (V).........................................................70 6.4. SOLID PHASE REFOLDING OF ZBASIC TAGGED FUSION PROTEINS (V) ....................................73

CONCLUDING REMARKS.......................................................................................................76 ABBREVIATIONS ......................................................................................................................78 ACKNOWLEDGMENTS ...........................................................................................................79 REFERENCES.............................................................................................................................81

8

TILL DIG SOM VILL

9

INTRODUCTION

1. PROTEINS The protein-encoding parts of the genome are transcribed and translated to build up chains of amino acid residues. Proteins are composed of 20 different amino acids with varying size, shape, charge, hydrogen bonding capacity and chemical reactivity. All proteins in all species, from bacteria to humans, are constructed from the same set of 20 amino acids. This fundamental alphabet of proteins is at least two billion years old (Stryer, 1996). The length of the chain and the specific order of the amino acids, denoted the primary sequence (Fig. 1A), are defined by the genetic code. These chains of amino acids are folded into organised structures, as illustrated in figure 1. Local folding, designated the secondary structure, is held together by hydrogen bonds between the carbonyl oxygen and the amide hydrogen of the peptide backbone into a more or less regular pattern. There are two secondary structure motifs that occur most frequently, the alpha helix (Fig. 1B)(Pauling et al., 1951) and the beta strand (Pauling et al., 1951). The overall three-dimensional protein conformation is formed through multiple interactions between these motifs and the organization is called the tertiary structure (Fig. 1C). Multimeric proteins consist of several domains and interactions between these subunits determine the quaternary structure (Fig. 1D).

Figure 1. The levels of protein structure A. Primary structure, here exemplified by a dipeptide of two amino acids. Side chain groups are depicted as R. B. Secondary structure, here exemplified by an alpha helix. C. Tertiary structure, here exemplified by a three-helix bundle. D. Quaternary structure, here exemplified by the dimeric protein insulin.

Consequently, the total structure of a protein is only determined by the primary amino acid sequence, that in turn is determined by the genetic code (Anfinsen, 1973). A protein’s biological function is then derived from its overall structure. Proteins can have widely diverse functions as the combinations of properties of amino acid residues and overall structures are nearly endless. For example, proteins play important roles such as messenger molecules, structural elements and regulators of biological processes. Given the multitude of their tasks it is not surprising that proteins are the second most abundant molecule in biology, next to water, and that even the simplest organism contains about 1000 different types of proteins (Dobson, 2004). Nearly all

10

pharmaceuticals act via proteins in the body and a large number of proteins are used therapeutically (Schmidt, 2004). As biology enters the post-genomic era, researchers have begun to entrance the exciting opportunity of investigating proteins in different high throughput experiments. Since proteins are complicated molecules there are many ongoing projects to elucidate the structure and function of the gene products. No single solution is available, but many different approaches are used; for structure determination (crystallography, NMR (Abbott, 2005)), for analysis of protein localisation (antibody based tissue profiling (Uhlen et al., 2005), reporter systems (Simpson et al., 2000)), for protein identification and analysis of global protein change and protein complexes (2D gels (Delahunty et al., 2005), LC-MS (Mann et al., 2001)) and for analysis of protein-protein interactions (two-hybrid systems (Ito et al., 2001), Tandem Affinity Purification (Gavin et al., 2002), protein microarray assays (Cahill et al., 2003)) etc. The efforts to characterize proteins of entire proteomes clearly demonstrate the importance of high throughput protein production and purification systems. Ideally, such systems should not be optimized for a single protein but allow for production and purification of a large number of proteins, irrespectively of their unique properties. The availability of proteins is also important for biomedical applications such as small molecule drug discovery and production of therapeutics and protein based vaccines.

11

2. RECOMBINANT PROTEIN PRODUCTION To isolate a protein from its natural source, a cell or tissue where it normally resides as a tiny fraction of the total protein content, is laborious and often unfruitful. In 1970, the first members of a group of enzymes that endogenously cleaves DNA helices in a specific recognition sequence were isolated (Smith et al., 1970). This enabled recombination of DNA fragments originating from different organisms (Cohen et al., 1973). Genomic DNA or cDNA specifying the protein could now be inserted into appropriately designed plasmids, called vectors, which can be introduced into host cells and induced to direct the synthesis of very large amounts of a specific protein. The first reported human gene product produced in a heterologous host cell was the 14 amino acid long growth hormone somatostatin expressed in Escherichia coli (E. coli) (Itakura et al., 1977). This new promising technique was soon used for production of several therapeutically used human proteins. In addition to increase the availability of important proteins, this new method ensured a significantly lower risk of virus contamination and allergic response (Schernthaner, 1993). Today, a large number of important proteins are easily produced thanks to recombinant DNA technology e.g. human growth hormone (Goeddel et al., 1979), human insulin (Goeddel et al., 1979), factor VIII (Wood et al., 1984) and tissue plasminogen activator (Dodd et al., 1986). Recombinant proteins can now be produced easily under controlled circumstances in well characterised host cells. Moreover, the high amount of proteins that can be achieved through overexpression increases the possibility of successful isolation of a specific protein. Recombinant protein production also opens up new possibilities, such as genetic engineering of proteins to attain desired properties.

2.1. HOSTS FOR RECOMBINANT PROTEIN PRODUCTION Today, several organisms, both prokaryotic and eukaryotic, are used as hosts for recombinant protein production. Primarily, the choice of host is based on the properties of the target protein and its future purpose. Each host organism has its own characteristics, both strengths and weaknesses. The host systems differ mainly concerning production cost, ease of use, expression levels and achieved final yield. However, as endogenous proteins are the main contaminations, downstream processing is also dependent on the host. Moreover, the quality of the protein and its proper folding, including post-translational modifications to assure biological activity, are determined by the available processes in the host cell. Prokaryotes are often attractive as hosts for recombinant production because of their cost-efficiency and suitability for large-scale cultivation. The most widely used prokaryotic host is the gram-negative enterobacteria E. coli. This is a result of the rather detailed understanding of its genetics (Blattner et al., 1997) and physiology which makes E. coli easy to manipulate and work with. Since the bacteria grows rapidly in simple media, high densities of protein producing cells can be obtain in a short period of time. However, being a gram-negative bacterium, the E. coli cell is encapsulated by two cell membranes separated by the cell wall and a periplasmic space, which complicates downstream processing of the protein product. As E. coli is the host chosen for production of the proteins used in this thesis, it will be discussed in more detail in section 2.1.1.

12

Gram-positive bacteria, such as Bacillus substilus (B. substilus) (Henner, 1990, de Vos et al., 1997) and Staphylococcus aureus (S. aureus) (Iordanescu, 1975, Hammarberg et al., 1990) lacks the outer membrane which simplifies the secretion pathway and makes them suitable for expression of recombinant proteins secreted into the cell media (Sandkvist et al., 1996). Suitable production vectors are now available although genetic manipulations are not as easy as with E. coli. However, B. subtilis produces large amounts of extracellular proteases, which often leads to severe degradation of the protein of interest and thus decreased yields of the full-length gene product (Wong, 1995). Moreover, the pathogenesis of S. aureus makes it unsuitable for large scale processes (Hansson et al., 1992). Many eukaryotic proteins require post-translational modifications, such as glycosylation, phosphorylation and addition of fatty acid chains, to obtain correct fold and biological activity (Kukuruzinska et al., 1998). The enzymes that carry out these modifications are located on membranes of different sub-cellular organelles, which are only present in eukaryotic cells and thus not in bacteria. However, the level, pattern and type of glycosylation vary between different eukaryotic hosts and must therefore be considered when deciding expression system. For instance, while different types of sugar molecules are attached to hydroxyl groups of serines and threonines in mammals, lower eukaryotes can only attach mannose on these positions (Cereghino et al., 2000). Yeast is often the first choice when looking for a eukaryotic system for recombinant protein production. It is a eukaryotic organism which, because it is unicellular, retains the advantageous growth properties of bacteria (Buckholz et al., 1991) although the genetics and life cycles of yeast are considerably more complex. Yeast is further known to grow to very high cell densities and many recombinant proteins produced in yeast can be secreted from the cell, which compensates for the low expression levels of foreign genes. The budding yeast, Saccharomyces cerevisiae (baker’s/brewer’s yeast) (Brake, 1990, Gellissen et al., 1992) has been extensively used as host for production of heterologous proteins. Thanks to its long history of use in the food industry, it has got ‘generally regarded as safe’ status with the US Food and Drug Administration (FDA) (Sofer, 1997). Even though the yeast contain sub-cellular organelles and thereby is capable of performing several post-translational modifications, the carbohydrate modifications performed by yeast often differ from those in higher eukaryotic cells. N-linked glycosylation occurs at the same sites as in mammalian cells but the outer chains of the N-linked carbohydrate cores are elongated by addition of 50-100 mannose residues so that the proteins become hyperglycosylated. Some mammal proteins that are devoid of sugars in the native host can also be O-glycosylated in yeast (Jenkins et al., 1996). The methylotrophic yeasts Pichia pastoris and Hansenula polymorpha are less prone to hyperglycosylate heterologous proteins (Romanos et al., 1995) and can hence be used for production when such overglycosylations wants to be avoided. Filamentous fungi, such as Trichoderma reesei (Keranen et al., 1995), are more highly organized than yeasts and consequently have a more complex posttranslational modification apparatus. However, at present, filamentous fungi can not be regarded as a serious alterative for the production of pharmaceuticals as these expression systems have not been developed enough yet. To fully exploit the potential of filamentous fungi, their physiology and particularly the glycosylation metabolism has to be investigated and clarified in more detail (Schmidt, 2004). Another interesting system for heterologous protein production is based on the ability to introduce the gene of interest into a viral vector (Luckow et al., 1988). The Baculovirus-based expression system uses a virus from the Baculoviridae family to infect anthropoid cells. Heterologous

13

proteins can be efficiently produced in both insect cells and larvae (Geisse et al., 1996). Despite involving virus, this system is considered as safe, since the baculovirus is unable to infect vertebrates (Carbonell et al., 1985). Virus-based expression systems are a good compromise of an advanced eukaryotic system that has fewer drawbacks than mammalian cells since these cells are more stress resistant and easier to handle. Maintenance of the cells is relatively cheap, the products are secreted and the yield of correctly folded product is generally high when compared to mammalian systems, although all post-translational modifications are not identical (Sofer, 1997). Cultivated mammalian cells offer the most human-like condition, which is required for the production of some highly post-translational modified proteins with multiple subunits and complex structures. Consequently, mammalian cell systems have become the dominant system for the production of recombinant proteins for clinical applications. African Green Monkey kidney (COS) (Gluzman, 1981) or Chinese Hamster Ovary (CHO) cells (Urlaub et al., 1980) are normally used for production of recombinant therapeutic proteins and monoclonal antibodies (Hodgson, 1993). The mammalian expression vectors are often of viral origin. A foreign gene can be transiently produced from a non-integrating and non-replicating DNA plasmid. However, for retained ability of every generation to encode the product, the foreign gene has to be integrated into the host cell genome (Makrides, 1999). The generation of stable mammalian cell lines is usually laborious and time consuming due to slow growth rates, potentially inefficient transfection and the high sensitivity to bacterial infection. Most mammalian cell cultures require complex media with biological supplements. The media can be in the form of mammalian serum, which potentially can cause viral contaminations. Alternatively, an artificial media containing serum albumin and other protein factors can be used. However, these proteins need themselves to be produced before use in the media and then also be removed later in the purification process. Great efforts have been put into the development of whole organisms of transgenic animals such as mice, pigs, sheep, goats and cows to be used as bioreactors for production of large quantities of therapeutic proteins (Maga et al., 1995, Larrick et al., 2001) The main concept is to use the cells of the mammary glands as sites for expression, directing the heterologous protein to the milk of the transgenic animals (Colman, 1998). Production in the blood can also be accomplished (Echelard, 1996). Besides ethical conflicts, the cost and time-consuming development of a production animal and the fact that each animal presents its own microbiological and virological concerns has restricted the use of transgenic animals. Transgenic plants have also been used as bioreactors and been shown to be able to secrete proteins (Conrad et al., 1998). Development of cell-free protein production strategies, based on purified enzymes, is an interesting solution with attractive features. In these processes the absence of a cell membrane eliminates the harsh conditions associated with introduction of DNA into cells, lysis of cells and clearance of lysates. The most widely used of these so-called open expression systems are based on bacteria (Spirin et al., 1988), wheat germ (Nakano et al., 1994) and reticulocyte lysates (Pelham et al., 1976). The main concerns regarding cell-free systems are related to membrane clogging, the low efficiency and the somewhat questionable reproducibility of the system. Several developments in recent years have significantly improved the cell-free systems that now can preferable be used for high throughput expression of proteins at low levels that are sufficient for various analyses (Sawasaki et al., 2002).

14

2.1.1 Escherichia coli as Host for Recombinant Protein Production

Ever since the enterobacterium E. coli was isolated by a pediatrician in 1884, it has been thoroughly studied and the complete genomic sequence is now known (Blattner et al., 1997). Another advantage when choosing E. coli as expression host is that numerous phenotypic variants have been developed for different aspects of protein production. A large number of promoters for control of protein expression have also been described. The availability of commercial vector kits, the easy handling of plasmids and effective transformation allow also non-experts to perform expression experiments in this host, using shake flask cultures. Due to long tradition and attained position as the most convenient and least expensive host to cultivate, the normal approach is to first try to express the protein of interest in E. coli. Many eukaryotic proteins retain their full biological activity in a non-glycosylated form and can therefore be successfully produced in E. coli (Makrides, 1999). Nowadays, the E. coli cell is the most commonly used organism in pharma- and biopharmaceutical companies, both for production of bioproducts and for research and development (Schmidt, 2004). The gene that is to be expressed in the E. coli cells should be encoded in a circular double stranded DNA molecule, called plasmid or vector. An E. coli expression vector should, except for the gene of interest, also contain; an origin of replication, a gene that confers a selectable marker, a translational initiation region, a promoter and a transcription terminator (Sambrook, 1989). An origin of replication contains the specific DNA sequence required for replication of the vector by DNA polymerase. This sequence determines the number of identical vector molecules present in every cell. By including a selectable marker that is crucial for cell survival, one identifies transformants and ensures that only plasmid-containing cells will survive. The most common selectable markers in recombinant expression plasmids confer drug resistance to ampicillin, kanamycin, chloramphenicol or tetracyclin. In bacteria, the interaction between the rRNA in the small ribosomal unit and a short sequence in the transcript, designated the ribosomal binding site, is crucial for translation initiation. The secondary structure around the ribosomal binding site and in the sequence immediately downstream of the start codon have been described to influence the translational initiation efficiency (Hall et al., 1982, Chen et al., 1994). The promoter is a sequence where decisions about transcription take place. Promoters can be grouped into inducible and constitutive (non-inducible). Constitutive promoters, exemplified by the Staphylococcal aureus protein A promoter (Lofdahl et al., 1983), cause a constant level of transcription and subsequent translation. An inducible promoter on the other hand needs induction, often by heat or addition of a chemical, to start the transcription of mRNA. An important criterion regarding inducible promoters is the possibility to down-regulate the transcription under non-induced conditions. If the gene product is toxic, leakiness of the promoter may lead to plasmid instability and low cell growth rate. The first inducible promoter was derived from the Lac operon and is called the Lac promoter (Gronenborn, 1976). In the absence of an inducer, the Lac repressor binds to its operator situated immediately downstream from the promoter. The lactose analogue, isopropyl-P-D-thiogalactopyranoside (IPTG), inactivates the Lac repressor that thereby dissociates from its operator and enables RNA polymerase binding and subsequent transcription. Nevertheless, a Lac promoter is not completely down regulated when

15

not induced but displays some leakiness. Therefore, it is not ideal for expression of proteins that could be harmful to the host cell. The pET system also exploits IPTG induction and is yet tightly regulated (Studier et al., 1986). The pET vector has a T7 promoter that endogenous RNA polym-erases of E. coli do not recognise. Transcription of the gene is therefore not accomplished as long as the plasmid is propagated in normal E. coli cells. In order to express protein, the plasmid needs to be transformed into a strain (e.g. BL21) that carries a chromosomal T7 RNA polymerase gene under the control of a Lac promoter (Studier et al., 1986). Further suppression of transcription in non-induced cells can be obtained by inclusion of a plasmid encoding a T7 lysozyme that degrades T7 RNA polymerase (Studier, 1991). The efficient suppression together with the amplification obtained by the two step induction has made this system very frequently used. To assure efficient release of the nascent transcript from the RNA polymerase after completed transcription, a termination sequence should be located downstream of the recombinant gene. In E. coli there is a preference for the UAA stop codon (Sharp et al., 1988). By preventing transcription through the replication region and through other promoters located on the plasmid, this termination sequence has been described to enhance plasmid stability (Hannig et al., 1998). In E. coli cells the proteins are produced by the ribosome in the cytoplasm and the recombinant protein production usually reach levels as high as 10-50% of the total protein content (Lesley, 2001). However, the reducing environment of the cytoplasm prevents the formation of disulphide bonds that might be important for the structure of the protein. By using different genetic strategies on the expression vector, the proteins can be secreted. The fusion of a signal sequence N-terminal of the target protein enables many recombinant proteins of moderate size to be efficiently exported to the periplasmic space (Fuh et al., 1990, Wulfing et al., 1994). Down stream processing of periplasmic proteins is considerably facilitated, as they may be selectively extracted by osmotic shock. The target protein will also be less contaminated since the periplasm contains only 4% of the total cell protein content (Nossal et al., 1966). Product release directly into the culture media would of course simplify downstream processing even more (Hansson et al., 1994). Extracellular secretion of proteins in E. coli cells can be promoted by non-specific leakage of the cells (Hewinson et al., 1993) or by utilising the existing pathways for naturally secreted proteins (Khushoo et al., 2004). However, only a few signal sequences have been reported to promote excretion into the culture media (Abrahmsen et al., 1986, Majander et al., 2005). An issue when using E. coli as a host is that the frequencies with which codons appear in homologous E. coli genes are different from those of human origin. Consequently, E. coli cells have a lower abundance of tRNA matching certain codons and thus, human genes that contain these codons may be inefficiently expressed. One strategy to overcome these problems is to change the rare codons into codons that are more abundant in E. coli (Hu et al., 1996). Another way is to use cells that overexpress rare tRNAs, e.g. Stratagene’s BL21CodonPlus™ (Kleber-Janke et al., 2000). The limited folding machinery in E. coli cells evokes a low solubility of many over-expressed proteins. To overcome these problems, many attempts of genetic modifications have been tried. It has recently become apparent that E. coli can be manipulated to achieve at least some post-translational modifications, such as glycosylation, which has long been considered too be beyond its reach (Wacker et al., 2002, Baneyx et al., 2004, Zhang et al., 2004). Several parameters, including choice of strain, promoter, translation initiation site, codon usage, termination signals and protein localisation, that are more or less involved in outcome of the yield, need in most cases to be empirically tested (Das, 1990).

16

2.1.1.1. Cultivation of Escherichia coli Cells Different cultivation strategies can be used depending on the purpose of the target protein. A high throughput protein production of many target proteins for analysis requires multi-format cultivation of small batches in parallel. The simplest way to produce moderate amounts of a recombinant protein is through cultivation in shake-flasks. Since E. coli is a simple organism, it is also highly amenable for growth in 96-deep-well block format. Cultivation for industrial production of a pharmaceutical protein product is more preferably performed in a large reactor chamber, a fermentor, where temperature, pH, concentration of dissolved oxygen, substrate and growth rate can be readily controlled. Three modes of process operation exists for fermentors, classified depending on mass flow; batch (closed substrate and closed biomass), continuous (open substrate and open biomass) and fed-batch (open substrate and closed biomass) (Fig. 2) (Enfors, 1994).

Figure 2. Three modes of process operation for cultivation of E. coli cells A. Batch cultivation. B. Continuous fermentation. C. Fed-batch fermentation.

In batch cultivations all nutrients required are added to the bioreactor in the beginning of the process. After initial adaptation, the cell growth will be exponential until one substrate or other compound becomes inhibiting or limiting. The cells are exposed to an ever changing environment with decreasing substrate concentrations and enrichment of excreted by-products. The unrestricted growth commonly leads to unfavourable changes in the growth medium, such as oxygen limitation and pH changes. With this mode, high growth rates are the priority, rather than achieving high final concentrations or steady state conditions. A continuous culture on the other hand, is fed with complete medium at a constant feed rate. Cultivation medium is withdrawn at the same rate as the added medium, to keep a constant volume. There are several variants of continuous processes, e.g. chemostat and turbidostat, which differs in the way they are controlled (Enfors, 1994). However, this mode of process is rarely used for E. coli cultivation in industry. When the fed-batch technique was developed in the end of the 19th century it opened up possibilities to reach higher cell densities and thus increased productivity. Fed-batch fermentations are operated with a continuous inflow of substrate but without any outflow of culture medium. Substrate components are added in such a way that the concentration is growth rate limiting and the growth rate can thus be controlled via the feed rate. In this way overflow metabolism and the accumulation of inhibitory by-products can be minimized. For industrial bioprocesses the fed-batch technique is the most commonly used when cultivating bacteria and has achieved some of the highest cell concentrations reported in the literature (Knorre et al., 1991, Enfors et al., 2003, Hu et al., 2004).

17

3. DOWNSTREAM PROCESSING After production of a recombinant protein in cultivated cells is accomplished, downstream processing to isolate the protein of interest begins. The protein must be released from the cells, separated from cell debris and purified from biological contaminants. This task might require several purification steps. The aim should be to minimize the number of unit operations, maximize yield and make the total process time and cost efficient. Initially, the cells are harvested by solid-liquid separation processes such as centrifugation or filtration. Since the protein seldom can be secreted into the media, the protein must be released from cellular compartments into solution. If the protein is secreted to the periplasmic space, mild fractionation methods, like osmotic shock or lysosyme/EDTA treatment that disrupts only the outer membrane, is sufficient to reach the target protein. In order to release proteins that are produced intracellulary, successful disruption of bacteria can be done by vigorous treatment with sonication, bead-milling, French-press or high pressure homogenization. To clarify the sample from lipids and particulate matter, such as cell debris, centrifugation or filtration is performed.

3.1. PROTEIN PURIFICATION There is no single or simple way to purify all kinds of proteins. Procedures and conditions used in the purification process of one protein may result in the inactivation of another. By taking advantage of small differences in the physical and/or chemical properties of the many diverse proteins in a crude mixture, separation of the proteins from one another can be achieved. The final goal has to be considered when choosing purification method. The purity required depends on the purpose for which the protein is needed. For an enzyme that is to be used in a washing powder, a relatively impure sample is sufficient, provided it does not contain any inhibiting activities. However, if the protein is aimed for therapeutic use it must be extremely pure. Purification must then be done in several subsequent steps. In the early days of protein chemistry, the only practical way to separate different types of proteins was by taking advantage of their relative solubility. Part of a mixture was caused to precipitate through alteration of some properties of the solvent e.g. addition of salts, organic solvents or polymers, or varying the pH or temperature. Fractional precipitation is still frequently used for separation of gross impurities, membrane proteins and nucleic acids (Glatz, 1990). Under certain conditions, proteins adsorb to a variety of solid phases, preferably in a selective manner. Calcium phosphate gels have frequently been used to specifically adsorb proteins from heterogeneous mixtures (Scopes, 1994). The adsorption principle is further explored in column chromatography.

3.1.1 Chromatography Chromatography refers to a group of separation techniques that involves a retardation of molecules with respect to the solvent front that progresses through the material. The name literally means “colour drawing” and was originally used to describe the separation of natural pigments on filter papers by differential retardation (Scopes, 1994). This same principle is now commonly used for protein separation. The most common physical configuration is column chromatography, in which the stationary phase is packed into a tube, a column, through which the mobile phase, the eluent, is pumped. The degree to which the molecule adsorbs or interacts with

18

the stationary phase will determine how fast it will be carried by the mobile phase. Chromatographic separation of protein mixtures has become one of the most effective and widely used means of purifying individual proteins (Scopes, 1994). General properties of proteins are used to isolate them from other, non-protein contaminants. Minor differences between various proteins, such as size, charge, hydrophobicity and biospecific interaction (Fig. 3), are used to purify one protein from another.

Figure 3. Selective protein properties Examples of properties that are used to separate one protein from another. In a typical chromatography process the first step is a capturing step, where the product binds to the adsorbent while the impurities do not. Weakly bound proteins are further washed away before the conditions are changed so that the target protein is eluted. Several versions of liquid chromatography, differing mainly in the types of stationary phase, are used for protein purification. Two variants of this technique, ion exchange chromatography and size exclusion chromatography, are illustrated in figure 4. Ionic interactions are the basis for purification of proteins by Ion Exchange Chromatography (IEXC) (Fig. 4A). The separation is due to competition between proteins with different surface charges and charge distributions for oppositely charged groups on an ion exchanger adsorbent. Proteins that adsorb to the ion exchanger are eluted from the column by an increase in ionic strength. This is accomplished by addition of new ionic species such as NaCl, or by a change in pH. IEXC is the technique most frequently used in this thesis and is further discussed in the next chapter (3.1.1.1). Proteins can also be separated according to differences in the amount of exposed hydrophobic amino acids, that is, their hydrophobicity. In Hydrophobic Interaction Chromatography (HIC) the protein mixture is loaded on the column in high salt concentration to ensure that hydrophobic patches are to be exposed. Hydrophobic interactions will occur between the phenyl groups on the resin and hydrophobic regions on the proteins (el Rassi et al., 1990). Bound proteins are then eluted by decreasing the salt concentration. In Reverse Phase Chromatography (RPC) the sample is applied in an aqueous solvent, often a dilute acid that disrupt the three-dimensional structure

19

(Scopes, 1994). Consequently, the separation in RPC is according to differences in the total hydrophobicity, since all of the amino acid residues are available for interaction with the stationary phase. However, the strong interaction requires elution by organic solvents, which can be deleterious for the activity of the target protein. Since this technique usually gives very high resolution it is commonly applied for separation of small polypeptides and proteolytic fragments. RPC is also used, often as a final polishing step, for purification of sturdy proteins and in cases when an inactive protein is adequate for the purpose (Scopes, 1994).

Figure 4. Illustrations of two classical chromatographic methods A. Ion exchange chromatography. The charges of a protein are used for purification. B. Size Exclusion Chromatography. Protein size is used for fractionation.

The biological function of proteins often involves interactions with ligands e.g. enzymes and substrates. Affinity Chromatography (AC) is a very useful technique that takes advantage of the interaction between a protein and its natural specific ligand, which is covalently attached to a chromatographic matrix (Muronetz et al., 2001, Labrou, 2003, Linhult et al., 2005). Desorption is performed non-specifically, by changing the pH, ionic strength or polarity, or specifically, using a competitive ligand. AC can be very selective and may allow purification of a protein in a single step. A slightly different process setup is used in Size Exclusion Chromatography (SEC) (Fig. 4B) where separation is achieved according to molecular size or, more precisely, ease of diffusion (Bollag, 1994). The medium consists of a range of beads with slightly differing amounts of cross-linking and therefore slightly different pore sizes (Walker, 2005). The separation process depends on the different ability of various proteins to enter some, all or none of the beads. The detour through the channels in the porous beads will retard smaller molecules in comparison to larger proteins. This method can also be used to determine the relative molecular mass of a protein and for desalting a protein solution. High Performance Liquid Chromatography (HPLC) is not a new chromatographic technique, but an advancement of the technology. The flow rates in conventional chromatography are limited because of the compression of the support matrices used in the columns. These low flow rates result in diffusion and loss of resolution. New resins that can withstand high pressure have been developed. This allows for separations to be carried out under higher flow rates resulting in increased resolution.

20

3.1.1.1. Ion Exchange Chromatography The fundamental theory of Ion Exchange Chromatography (IEXC) has a very long history. One of the first examples of ion exchange purification is attributed to Moses, who purified acrid water with the aid of a special type of wood (2 Mos. 15:25). The first synthetic ion exchangers consisted of hydrophobic polymer matrices, highly substituted with ionic groups (Karlsson, 1998). Due to their low permeability these matrices had low capacities for proteins. In addition, the hydrophobic nature of the matrix denatured the proteins. It was not until hydrophilic materials of macroporous structure were introduced, in the late 1950s (Sober et al., 1958), that ion exchange chromatography of biological macromolecules became a useful separation tool (Sofer, 1995).

3.1.1.1.1. Charge Properties of Proteins Proteins are complex ampholytes that have both positive and negative charges. The isoelectric point (pI) of a protein (the pH at which the net charge is zero) depends on the proportions of ionisable amino acid residues in its structure. Positive charges are usually provided by arginines, lysines and histidines, depending of the pH of the surrounding buffer. Any free N-terminal amine will also contribute with a positive charge below pH 8. Negative charges are principally provided by aspartate and glutamate residues and the C-terminal carboxyl group. Virtually all these residues are ionized above pH 6. At higher pH values (>8) cysteines may become ionized too (Scopes, 1994). The charged groups nearly always reside on the protein surface. Exceptions are mainly metallo-proteins, where an internal metal ion often is coordinated by charged residues (Karlsson, 1998). Influences from neighbouring groups and the position in the tertiary structure will affect the pKa for the side-chain groups. The combined influence of all of the charged side chains will give the protein a varying net charge depending on the pH of the solute. Therefore, it is possible to separate proteins using either fixed positive charges on the stationary phase, anion exchanger, or fixed negative charges, cation exchanger. A protein must displace the counterions to become attached and consequently the net charge on the protein will be the same as that of the counterions displaced, thereof the term “ion exchange”. Generally, anion exchange chromatography (AIEXC) is carried out at pH values above the isoelectric point of the protein of interest, while cation exchange chromatography (CIEXC) is carried out below the isoelectric point. The pH interval in which ion exchange chromatography is carried out is restricted by the pH range in which the protein is stable. To achieve good adsorption, the pH of the buffer chosen should be at least one pH unit above or below the isoelectric point of the analytes to be separated. The pH in the microenvironment of an ion exchanger is not exactly the same as that of the applied buffer. This is due to the so called Donnan effect, that protons are attracted or expelled (depending on charge of the matrix) from the microenvironment close to the matrix, (Scopes, 1994). In general, the pH close to the matrix is up to 1 unit higher than that in the surrounding buffer in anion exchangers, and 1 unit lower in cation exchangers. Consequently, if a protein is adsorbed on a cation exchanger at pH 5, it will be exposed to pH 4, and if it has poor stability at that pH, it may be denatured. Most proteins are negatively charged at physiological pH values (pH 6 – 8) (Righetti et al., 1976) and therefore, in many applications of IEXC a first approach is to use an anion exchanger. However, for optimal separation the selection of pH should aim to, if possible, introduce as large charge difference as possible between the target protein and the contaminants. A distribution profile of the isoelectric points of a wild-type E. coli proteome showed that 95% of the intracellular proteins were negatively charged at basic pH (Kweon et al., 2002) and should therefore be most easily removed using CIEXC. It is possible to use IEXC more than once in a purification strategy since the pH of the separation can be modified to alter the charge characteristics of the sample components. Typically IEXC is used to bind the target molecule and then wash away non-bound contaminants. However, the

21

technique can also be used to bind the impurities if required. In that case the protein of interest should be found in the flow through. The interaction between a protein and an ion exchanger depends not only on the net charge and the ionic strength, but also on the surface charge distribution of the protein. If there are surface regions with a high concentration of charged groups on the protein, binding can occur even when the overall charge is the same as that on the ion exchanger. This phenomenon is exemplified by yeast phosphoglycerate kinase that binds to a cation exchanger at a pH when its net charge is slightly negative (Scopes et al., 1979). Moreover, since the chromatographic behaviour also depends on protein conformation, some structural changes can affect the separation by IEXC (Parente et al., 1984, Chicz et al., 1988). The choice of buffering ions can also influence the separation. For example, the electrostatic attraction between two oppositely charged groups is higher in a hydrophobic environment. In several instances, ammonium acetate buffers have increased the resolution. This buffer has also the advantage of being volatile and can thus easily be removed by lyophilization.

3.1.1.1.2. The Stationary Phase The properties of the ion exchanger can also influence the separation. Ion exchangers are usually classified as weak or strong. The name refers to the pKa values of their charged groups (by analogy with weak and strong acids or bases) and does not say anything about the strength of the interaction. Strong ion exchangers have pKa values outside the pH range in which it is usual to work with proteins (i.e. pH 4-10) and pH changes will therefore not change the charge of the ion exchanger. Contrary, weak ion exchangers have a limited pH range for their use. Thus, weakly ionisable proteins, requiring a very high pH or very low pH for ionization, can be separated only on a strong exchanger. On the other hand, for more highly charged proteins, weak ion exchangers are advantageous for a number of reasons, including a reduced tendency to sample denaturation, their inability to bind weakly charged impurities and the enhanced elution resolution.

3.1.1.1.3. Elution Strategies Normally, proteins with the same charge as the resin will pass through the column to waste while proteins with the opposite charge will be bound. If the sample components are only differentially retarded, and thereby separated under constant solvent composition, no changes in buffer composition are required. This is termed isocratic elution. In this way only sample volumes much smaller than the total bed volume can be applied and the resolution increases as the square root of the column length (Karlsson, 1998). More often, however, a decrease in affinity is mediated by a change in the buffer, in order to selectively release the proteins from the column. Two general methods are available, changing the pH of the eluting buffer or increasing the ionic strength by addition of NaCl. If a pH change is used, anion exchangers should be eluted by a decrease in pH to make the adsorbed proteins less negative, whereas cation exchangers are eluted by an increase in pH to make the proteins less positively charged. However, in practice pH-elution is generally not very successful. Since many proteins show minimum solubility in the vicinity of their isoelectric point (Cleland, 1993) care and precautions must be taken to avoid precipitation on the column. Moreover, unless having a very high buffering capacity, large pH changes can occur when proteins become eluted. This leads to a less good separation of individual components (Scopes, 1994). However, by employing special systems that maximize buffering capacity, it has been possible to achieve high resolution of proteins by pH elution. This technique is referred to as chromotofocusing (Sluyterman, 1978). A more common strategy to achieve elution is to increase the concentration of a non-buffering salt, such as NaCl. These ions compete with the protein for binding sites on the resin. More weakly charged proteins are eluted at lower salt strengths while the more strongly charged proteins are eluted at higher salt strengths.

22

Buffer changes required for elution can be added either in stages, by stepwise elution or continuously, by gradient elution. Stepwise elution is a serial application of several isocratic elution steps, each step consisting of about a single total bed volume of eluent. Stepwise elution is often used for recovery of a concentrated protein in the breakthrough peak of the displacing buffer (Karlsson, 1998). In gradient elution the concentration of the eluting buffer is changed continuously. First the least strongly adsorbed protein is desorbed and at somewhat higher concentration the second protein and so on. Gradient elution generally leads to improved resolution since zone sharpening occurs during elution. The separation obtained is similar to isocratic elution, but because of the continuously increasing elution power, the peaks do not become much broader as the gradient develops. Decreasing the slope of the gradient will lead to a greater separation of the solutes. However, as the slope decreases, the proteins will be more diluted. Due to extended equipment requirements, gradient elution may sometimes not be feasible for large-scale processes. The optimized gradient conditions need, in this case, to be transferred to a series of steps to first elute less retained contaminants, then elute the protein product and finally, release all tightly bound solutes in a wash step.

3.1.1.1.4. Features Ion Exchange Chromatography is one of the more powerful protein purification techniques available and is probably the most frequently used chromatographic technique for the separation of proteins, polypeptides, nucleic acids, polynucleotides, and other charged biomolecules. An advantage of this technique is that the elution normally takes place under mild conditions, so that the protein can maintain its native conformation during the chromatographic process. In general, ion exchangers are more densely substituted than other adsorbents used in protein chromatography. Its capacity for protein binding is high and its interaction with proteins due to hydrophobic or other non-ionic interactions is low. Its broad specificity also allows for removal of significant impurities such as deamidated forms, endotoxins and unwanted glycoforms (Sofer, 1995). Ion exchanger resins are very robust and can be sanitized in place and used for hundreds of cycles. Additional reasons for the success of IEXC are the straightforward separation principle and ease of performance and controllability of the method. The main disadvantage of IEXC is its limitations in selectivity.

3.1.1.2. Expanded Bed Adsorption Chromatography When purifying proteins from a complex mixture, a sequence of unit operations is normally needed (see intro to chapter 3). Each of these steps involves a potential loss of product and will add to the overall cost of the process. One way to simplify the situation is through process integration. The Expanded Bed Adsorption (EBA) technology represents a single pass operation in which target proteins are purified from crude feedstock sample, without the need for separate clarification and initial purification to remove particulate matter such as cell debris (Chase, 1994, Hjorth, 1997). The expanded bed process is operated in a manner similar to that of a packed bed process, with the main difference that the direction of liquid flow is upward, leading to an expansion of the bed. An illustration of a purification using EBA technology is shown in figure 5. The resin differs in that there is a distribution of sizes and densities of the individual particles in the gel, leading to the development of a gradient in the column, where the smallest particles are located at the top. This grading of particles reduces mixing and stabilizes the bed. In the ideal case, the net result is that the individual particles become stationary in the gel, leading to a laminar plug-flow of the sample through the column (Chase, 1994). Feedstock application and washing is done when the bed is expanded (Fig. 5A, B) to avoid that particulate matter causes clogging of the column and thereby destroyed separation power. Elution of the target protein is usually done in packed bed mode (Fig. 5C).

23

Figure 5. Process set-up for EBA purification A. Sample loading on an expanded bed. B. Washing out non-bound proteins and cell debris. C. Elution in packed bed mode.

There are several commercial adsorbents available for EBA purifications, including adsorbents for ion exchange chromatography (Johansson et al., 1996), immobilised metal ion chromatography (Noronha et al., 1999) and affinity chromatography (Chase et al., 1992). Since unclarified feedstocks are applied when using EBA technology, an important consideration is the sanitization of the chromatographic resin (Sofer, 1997). Therefore, industrial implementation of the EBA technology has predominantly been focused on ion exchange chromatography, as these resins can tolerate the harsh conditions employed in cleaning-in-place protocols necessary for efficient sanitization (Hale et al., 1994, Asplund, 2000). Although the expanded bed gives sufficient space for the particulate matter to be washed away, still problems due to cell interactions with the gel remains. Bacterial cell surfaces are negatively charged which allows them to interact with the anion exchange resin at low conductivity. Cation exchange chromatography often leads to better results. In addition to less cell debris interactions (Feuser et al., 1999), this resin also gives a better removal of contaminating DNA and endotoxins due to the negative charge of both contaminants (Garke et al., 2000).

24

3.2. PURIFICATION TAGS When purifying recombinant proteins it is of importance to minimize the number of unit operations. This can be accomplished by increasing the selectivity in each step. For purification of large amounts of a single target protein it is worthwhile to spend some time on finding conditions where the target protein can be selectively isolated. On the contrary, if many different proteins should be purified in a high throughput manner, the time spent on development of the purification process has to be minimized. In order to gain a highly selective property of each target protein, gene fusion strategies can be used. A purification tag is then genetically fused to the protein of interest and when the chimeric protein is expressed the tag allows for specific capture independent on the characteristics of the protein of interest. In figure 6, a typical affinity purification process is shown.

Figure 6. The gene fusion strategy The cell lysate, consisting of a complex mixture of molecular species, is passed over the matrix. Proteins with a specific purification tag (here in black) are captured on the ligand (here illustrated as a fish) while the rest of the molecules passes through.

The protein used as purification tag can be selectively captured due to several kinds of interactions, such as; protein-protein interactions, enzyme-substrate interactions and charge interactions. A large repertoire of purification tags is now available and each tag has its own characteristics, advantages and disadvantages. The processes used for capture and release of different tags vary in cost and complexity. Some tags require quite harsh purification conditions while other can be eluted directly into a suitable buffer. Most often, a desired high selectivity obligates harsh conditions to release the protein from the affinity matrix. The forces that can break a strong affinity interaction usually affect the protein of interest as well (Terpe, 2003). An ideal

25

purification system would combine high selectivity with the retained possibility of mild elution. Moreover, the corresponding ligand must be stable under all used conditions. The selectivity of a tag, and hence the possibly attained purity, is also affected by both the target and the initial sample contaminants from the expression host. Moreover, the fusion to a purification tag has varying effect on expression levels as well as solubility and stability of the target protein. A few tags has a selectivity that is structure independent and can thus be used even under denaturing conditions, as obligated for purification of proteins produced as inclusion bodies. Certain tags are only functional when fused to a specific terminus of the target and might thereby hinder the activity of the target protein. Sometimes the intended use of the target protein requires removal of the purification tag, especially when using large tags, since they are more likely to interfere with the function of the target protein. Consequently there are several aspects to consider. The choice of purification tag is dependent on both the target protein and its further use. Although a multitude of tags has been developed, no strategy is ideal for all expression and purification situations. Originally, native domains with specific affinity were used as purification tags to allow for selective purification. Later on, short peptides harbouring a known function as well as peptides with new functions have been designed. In the following section, the most commonly used purification tags and their important characteristics are presented. The subject is further reviewed in (Hedhammar et al., 2005). For simplicity, the protein tags are divided in two groups according to their size; protein domains (3.2.1) and peptides (3.2.2). The engineering of a protein domain into a functional purification tag is also discussed (3.2.3). Illustrations of some representative purification tags are shown in figure 7.

Figure 7. Purification tags

A. A purification tag based on a natural domain, here represented by GST binding to glutathione. B. A peptide tag, here represented by the metal binding His6 tag. C. An engineered purification tag, here represented by the positively charged Zbasic.

3.2.1 Protein Domains A majority of naturally occurring proteins perform their action by binding to other molecules. The effect of the interaction can for example be to maintain a structural function or to mediate a signal. Small globular proteins with unique binding properties inherited from nature can successfully be used as protein affinity tags. The use of intact protein domains, able to independently acquire the native structure and desired affinity, allows predictable protein purification processes.

26

3.2.1.1. Enzymes Enzymes and their binding sites are quite convenient fusion complements, if stable complexes with natural or synthetic substrates or inhibitors are formed (Flaschel et al., 1993). Historically, β-galactosidase was used to protect heterologous peptides from proteolytic degradation (Germino et al., 1984). Soon, its further use as purification tag was established with the development of thiogalactosidyl modified Sepharose (Steers et al., 1971). Unfortunately, the large size of β-galactosidase has impact on the productivity of the expression system, which limits its use as purification tag. Moreover, the active enzyme consists of a tetramer, which can disrupt the folding pattern of the target protein (Stevens, 2000). Glutathione S-transferase (GST) belongs to a family of enzymes that modifies toxic substances by adding sulphur from glutathione, leading to their excretion in the urine (Mehler, 1993). The interaction between GST and glutathione (Fig. 7A) was first utilised in 1988 by the pioneer Smith who developed what has now become one of the most commonly used purification systems (Smith et al., 1988). A GST fusion protein binds reversibly to immobilised glutathione and is competitively eluted with an excess of reduced glutathione. Although otherwise non-denaturing conditions, the use of reduced glutathione for elution might affect target proteins containing disulfides (Sassenfeld, 1990). The GST moiety has been shown to improve solubility by stabilizing the recombinant protein and in the majority of cases GST fusion proteins are soluble in aqueous solutions (Ford et al., 1991). Nevertheless, it is a common experience to observe the accumulation of partially truncated protein products when GST fusion constructs are expressed (De Marco et al., 2004). Moreover, GST is known to dimerize and this may interfere with the function of the target protein (Phizicky et al., 2003). It is important to keep the flow rate low during sample application since the binding kinetics between glutathione and GST are relatively slow. The glutathione binding site of GST is destroyed by denaturing agents and therefore the purification under non-native conditions is not feasible (Sassenfeld, 1990).

3.2.1.2. Carbohydrate Binding Domains Fusions to carbohydrate-binding proteins are attractive, as a carbohydrate based ligand is cost effective and susceptible to treatment with heat or alkali. A well-known and frequently used carbohydrate-binding protein is the maltose binding protein (MBP). MBP is encoded by the MalE gene of E. coli K12 (Duplay et al., 1988) and the product is exported to the periplasmic space where it binds specifically with affinities in the micro-molar range to maltose or maltodextrins. Theses sugars are then subsequently transported across the cystoplasmic membrane (Duplay et al., 1984). The use of MBP as a purification tag was first described in 1988 by di Guan et al (di Guan et al., 1988). The purification system is based on the strong affinity of MBP to cross-linked amylose from where it can be gently eluted with maltose (Ford et al., 1991). The low cost for amylose-based matrices and the mild conditions used for binding and elution makes this system very attractive. Moreover, MBP has been attributed chaperon-like qualities when expressed upstream of the target protein (Sachdev et al., 1998, Kapust et al., 1999). This, and the possibility of secretion of the product, has led to that fusion to the N-terminus of the target protein has been mostly used. For a successful purification, the biological activity of MBP has to be fully conserved. Consequently, denaturing agents can not be used. In the case of structural disturbance due to the sub-cloned gene, affinity purification of the fusion protein can not be performed. However, this problem can sometimes be alleviated by optimizing the linker between MBP and the target protein (Terpe, 2003).

27

3.2.1.3. Protein Binding Domains The most common interaction partner for a protein is another protein. Many protein domains with native affinities for other proteins have been explored for protein purification, since native protein-protein interactions typically give very high selectivity. However, a drawback when using protein domains as ligands is the inherent sensitivity to proteolytic degradation and the harsh conditions normally needed for elution of the target protein. Early affinity systems used matrices with immobilised antibodies for capture of the corresponding antigen (Krivi et al., 1985). Nowadays, antibodies are commonly evolved against a target protein and thereafter immobilised to be used for purification of proteins. Since it became clear that only the part of the polypeptide chain that includes the antigenic determinant was needed for efficient recovery, several fusion tags based on antigenic epitopes have been developed (see section 3.2.2.1). A problem commonly accounted for using the principle of immunoaffinity is that most antibody-antigen complexes only dissociate under quite harsh conditions. Several pathogenic bacterial species express surface proteins that are capable of specific binding to various host proteins e.g. immunoglobulin (IgG). The biological function of this action is possibly that the bacteria uses the interactions with host molecules as means for camouflaging itself to weaken the host’s immune response (Achari et al., 1992, Sauer-Eriksson et al., 1995). One of the most commonly used of these interactions is the inherent affinity of Staphylococcal Protein A (SPA) to IgG (Langone, 1982). The structure of SPA shows a highly repetitive organization, consisting of five small domains (E, D, A, B, and C) that all show independent affinity to the constant region of IgG (Fc) (Lofdahl et al., 1983, Uhlén et al., 1984, Moks et al., 1986). Due to the high selectivity, independent folding, lack of cysteines and proteolytic as well as thermal stability of SPA, it has been used in several versions as affinity tags for IgG-affinity chromatography purifications. All five domains are folded into three-helix bundles with both the N- and the C-termini exposed to the solvent, which allow the target proteins to be fused to either end (Gouda et al., 1997). However, the asparagine-glycine sequence present in all domains makes the protein sensitive to hydroxylamine treatment. In order to develop an improved single SPA domain, an engineered version, denoted Z (Fig. 11A, F, page 55), was developed (Nilsson et al., 1987). The engineered variant was based on the sequence of domain B for several reasons. First, domain B is closest to a hypothetical consensus sequence of the five IgG-binding domains (Moks et al., 1986). Moreover, it naturally lacks methionine residues and is thus resistant to cyanogen bromide treatment. As the domain B-IgG complex has been crystallized and the three dimensional structure solved (Deisenhofer, 1981), it was possible to conclude that the glycine residue in the asparagine-glycine dipeptide sequence could be changed without interfering with neither structure nor IgG binding. Both monomeric and dimeric Z domains have been successfully used as purification tags. Since the domains are able to refold after exposure to denaturants, refolding of fused target proteins are possible (Samuelsson et al., 1994, Samuelsson et al., 1996). One drawback with the protein A-IgG system is the harsh elution conditions (pH 3) needed for release of the target protein. In order to enable elution at higher pH values, the Z domain has been genetically modified. A destabilised Z variant allowed elution to be performed at pH 4.5 (Gulich et al., 2000). Another strategy to overcome this problem is to use competitive elution by a bi-functional competitor fusion protein, consisting of divalent Z fragments fused to a second affinity tail (Nilsson et al., 1994). The competitor can afterwards be efficiently removed from the eluted mixture and reused. Another bacterial surface protein also used in affinity purification is Streptococcal Protein G (SPG) that naturally is found on the cell wall of certain strains of Streptococcus. SPG is bi-functional and displays affinity both to IgG and serum albumin (SA) (Guss et al., 1986, Bjorck et al., 1987). Fortunately, the regions responsible for the binding to IgG and SA respectively, are structurally separated (Nygren et al., 1988) which facilitates separate expression of fragments with different features.

28

The C-terminal consists of three domains with IgG binding capacity and this part has been used as a ligand attached to a solid matrix in commercially available protein-based affinity chromatography media for capture of antibodies. Interestingly, the N-terminal part of the protein displays affinity to SA in three repetitive domains. Different parts of the SA binding region, denoted albumin-binding domains (ABDs) have been used as fusion tags allowing one-step SA-affinity chromatography purification. The ABDs have a compact three-helix bundle structure similar to the IgG-binding domains from SPA (Johansson et al., 2002) and are also soluble, resistant to proteolysis and possible to secrete (Nygren et al., 1988, Stahl et al., 1989, Larsson et al., 1996). Moreover, the in vivo stability of otherwise rapidly cleared proteins seems to be increased if they are fused to ABD before injected into the bloodstream of laboratory animals (Sjolander, 1993, Nilsson et al., 1997). ABD could therefore be an ideal partner for delivery of therapeutic products. The ligand used for purification of ABD fusion proteins, SA, is a non-expensive robust monomeric protein. This reduces the risks of leakage compared to resins containing IgGs, which are composed of disulphide-linked heavy and light chain subunits (Nilsson et al., 1996). Normally, elution of proteins fused to ABD is performed at a pH of 2.8, conditions that can be destructive for the target protein. However, the SPG derivates can be eluted at different pH values depending of the origin of the SA used as ligand (Nygren et al., 1990). Moreover, it is also possible to perform competitive elution, using a similar strategy as for Z fusion proteins (Nygren et al., 1990, Nilsson et al., 1994).

3.2.1.4. Streptavidin Binding Proteins Streptavidin, a tetrameric protein derived from Streptomyces avidinii, exhibits a strong affinity to the vitamin biotin (KD= 10-10-10-15) (Green, 1975). Chemical techniques for in vitro biotinylation, whereby lysine residues are modified, can not be directed and may thus disrupt the protein conformation if certain important lysine residues are modified. An alternative, site-specific in vivo biotinylation procedure would be superior. In nature only a few proteins have the ability to incorporate biotin and these proteins normally act as biotin transporters involved in metabolic carboxylate transfer processes within the cell (Samols et al., 1988). The 1.3 S subunit of Propionibacterium shermanii ( P. shermanii ) transcarboxylase (Samols et al., 1988) and the biotin carrier protein (BCCP) (Fall et al., 1973) derived from E. coli both have the ability to attach a biotin moiety when utilising the in vivo biotinylation machinery of E. coli. The reaction is catalysed by the enzyme biotin ligase (birA) in the presence of ATP (Barker et al., 1981). A polypeptide of 75 amino acids, that was long enough to serve as a suitable substrate for biotinylation, has been derived from the 1.3 S subunit of P. shermanii transcarboxylase and this sequence is termed the biotin binding domain (BBD) (Schatz, 1993). In a similar approach, a C- terminal 101 polypeptide derived from BCCP protein of E. coli has been successfully used as affinity tag (Weiss et al., 1994). The biotinylated proteins can be recovered by affinity purification on immobilised streptavidin or avidin. Very few bacterial or eukaryotic proteins are naturally biotinylated, resulting in a high level of purification for fusion proteins containing the biotin binding tail (Cronan, 1990). However, the extremely high affinity (KD~10-15) to unmodified tetrameric streptavidin or avidin demands recovery by boiling in SDS or cleavage with a site-specific protease. The more moderate affinity of biotin to modified monomeric avidin (KD~10-6-10-

7) enables recovery of the bound proteins to take place at neutral pH and low salt concentrations (Hearn et al., 2001). Another strategy to avoid harsh elution conditions has been provided by the design of photocleavable tags (Thiele et al., 1993, Olejnik et al., 1995). The general concept is that a biotinylated photoreactive group is conjugated to a tag that binds to the target protein. A biotinylated target complex can efficiently be separated from other molecules using a streptavidin matrix. Thereafter, by illumination, the biotin moiety can be removed from the target protein and the target protein is released from the affinity resin (Olejnik et al., 1995).

29

3.2.2 Peptides In terms of metabolism of the host cell, small fusion tags are preferred since the larger the added purification tag, the more energy is consumed when producing the gene product. Moreover, a small fusion tag is less likely to interfere with structure and function of the target protein and may not need to be subsequently removed (Terpe, 2003). However, the effect also depends on the location and amino acid composition of the tag (Bucher et al., 2002). Another advantage of utilizing small tags is that the cloning procedure often is facilitated since a small tag can be genetically fused to the target protein by PCR technique. The fact that a rich source of short peptides with novel binding specificities can be obtained through suitable screening and selection techniques also increases the availability of useful peptides (Nygren et al., 1994).

3.2.2.1. Antigenic Epitopes Since the pioneering immunoaffinity purification work of Krivi (Krivi et al., 1985), numerous epitope tag systems have been developed where the fused epitope can be recognized by an immobilised monoclonal antibody, specific for the epitope. The reported affinity tags represent epitopes for mono or polyclonal antibodies with strong affinities. However, their application has been restricted mostly due to the harsh conditions, normally low pH, which must be used in order to elute the target protein from the column. Moreover, the production of antibodies that are required for construction of the matrices also makes the system expensive and undesirable for large-scale applications. Exploiting the immunoaffinity principle, Hopp and coworkers developed an elegant system based on a Ca2+ dependent binding of a monoclonal antibody to an octapeptide, the FLAG tag (Hopp et al., 1988). Since the interaction depends on the presence of calcium ions, the FLAG tagged fusion protein can be released from the column by simply removing the Ca2+ ions. This can be done by applying a chelating agent such as EDTA in the elution buffer. Consequently, harsh conditions can be avoided even though dealing with an antibody antigen complex. When designing the FLAG tag, Hopp took into account that the N-terminal fusion sequence should be hydrophilic. The purpose was both to minimize interactions with the adjoining protein structure and to maximize availability for interactions with the immobilised antibody. The antigenic epitope, defined by the first four amino acids of the tag, is combined with an enterokinase protease site for subsequent cleavage of the tag from the fusion protein if necessary. One draw back with the FLAG peptide is that its interaction with the M1 antibody is functional only when the tag is located at the extreme N-terminus of the target protein. This limits the use of the M1 antibody to the purification of fusions exposing an N-terminal FLAG peptide, such as after removal of a signal peptide from a secreted FLAG fusion (Hopp et al., 1988, Prickett et al., 1989). To overcome this problem, a shorter version of FLAG has been developed, consisting of only the first 4 amino acids (Knappik et al., 1994) and another mAb (M2) capable of binding to this FLAG peptide at the C-terminus of the target protein was selected. However, this interaction is non-calcium dependent and elution from mAb M2 can therefore only be achieved by lowering pH or competitively, by using a synthetically made FLAG tag (Brizzard et al., 1994). A disadvantage of both FLAG-tag systems is that the monoclonal antibody matrix is not as stable as desired.

3.2.2.2. Protein Binding Peptides In order to further optimize affinity fusion purification systems, smaller peptides mimicking interactions known from nature has been developed. The protease subtilisin recognize and cleave a peptide bond in native RNaseA (Richards, 1955) and the product results in two tightly associated fragments, the S-peptide (1-20) and the S-protein (21-124). The strong affinity between these fragments (KD~10-9) has been exploited in the development of a fusion tag system for purification (Kim et al., 1993). A truncated version of the S-peptide is highly soluble and fusion

30

to this S-tag tag often results in increased solubility of the target protein (Hearn et al., 2001). The S-tag is functional both when attached to the N- and C-terminus of the target protein (Kim et al., 1993). Strong binding ensures a high selectivity of this system but the elution conditions are very harsh, e.g. pH 2 (Ho et al., 1996). A great advantage with this system is that the two fragments form active ribonuclease S when associated, which can be assayed with high sensitivity (Karpeisky et al., 1994). Another often utilised ligand protein is calmodulin, a small and highly acidic calcium binding protein that interacts with a number of different peptides with high affinity. A peptide of 26 amino acids derived from the C-terminus of muscle myosin light chain kinase (MLCK) has the ability to bind calmodulin with nanomolar affinity and is therefore referred to as the calmodulin binding peptide (CBP) (Stofko-Hahn et al., 1992, Neri et al., 1995). The strong interaction between calmodulin and CBP in presence of Ca2+ions makes this system highly attractive since the tight binding allows more stringent washing conditions, ensuring that few contaminating proteins will be co-purified. Due to dependence of Ca2+ for calmodulin structure, the interaction can be inhibited by eliminating the cation. This provides the possibility of a very mild elution at neutral pH using virtually any buffer containing a chelating agent. No endogenous E. coli proteins interact with calmodulin which makes this system suitable to purify recombinant proteins produced in this host. On the contrary, purification of proteins produced in eukaryotic cells is not recommended since many eukaryotic endogenous proteins interact with calmodulin (Head, 1992). CBP does not appear to affect the biological function of the target protein, probably due to its small size. However, an N-terminal location of the tag may reduce the efficiency of translation, while placement at the C-terminus often results in high expression levels (Terpe, 2003). The two-stranded α-helical coiled-coil is known as one of nature’s favourite ways of creating a dimerizing motif (Chao et al., 1998). Based on this knowledge, the interaction has been employed as the basis for a de novo designed affinity purification method, the E/K coil/coil-system (Tripet et al., 1996). The system constitutes of two peptides; the E-peptide tag (EVSALEK), fused to the target protein and the K-peptide ligand (KVSALEK), conjugated to an activated silica based matrix. Each peptide is repeated five times to assure optimal formation of a stable coiled-coil domain. Elution of the fusion protein can be achieved by lowering the pH and by simultaneously increasing the acetonitrile concentration, thus quite harsh conditions. The K-peptide ligand is small and can be easily obtained by synthetic or recombinant means which suggests that great promises exist for large-scale applications (Hearn et al., 2001). Moreover, the purification pair is shown to be resistant to proteolytic degradation and tolerant to denaturants such as urea and heat (Chao et al., 1998).

3.2.2.3. Streptavidin Binding Peptides Biotinylated domains, as discussed earlier (section 3.2.1.4), are large (75-105 aa), have quite low stability to intracellular proteolysis and is biotinylated to a low extent (Smith et al., 1998). Due to these shortcomings, efforts have been put in the development of smaller in vivo biotinylated tags. The sequences surrounding the biotinylation-sites of proteins from diverse biological sources are strongly conserved (Cronan, 1990). The upstream sequences are usually rich in proline and alanine, and usually contain segments with two proline residues separated by a residue with a small side chain (Gly, Ala, Ser). In a combinatorial library of short peptides, sequences capable of mimicking the much larger natural substrate domain were found (Schatz, 1993). A consensus sequence of 13 amino acids was found and the corresponding peptide, denoted the BioTag, could replace the larger in vivo biotinylated fusion partners earlier used (Schatz, 1993). Both N- and C-terminal fusions to the peptide have been demonstrated to be functional. In a similar approach, a 15 amino acid peptide, also known as Biotin AviTag™, was found. Still, the high affinity between streptavidin and biotin requires forceful conditions for desorption, which makes this system less attractive for preparative purposes.

31

A different strategy to obtain streptavidin binding peptides is to screen for peptides that mimic biotin. The Strep tag I is a nine amino acid peptide that was selected from a random peptide library as an artificial ligand for streptavidin (Schmidt et al., 1996). This peptide mimics the structure of biotin and binds to streptavidin with good affinity via peptide-protein interactions. Still it can be eluted using mild conditions with the biotin analogue iminobiotin (Schmidt et al., 1996). The interaction is sufficiently strong for affinity purifications (KD=37µM) (Schmidt et al., 1996) but only when the Strep I tag is fused to the C-terminus of the desired protein (Maier et al., 1998). This limitation is due to a salt bridge between the carboxy group of the free C-terminus of the Strep-tag I and Arg84 of streptavidin (Schmidt et al., 1996). Employing an assay with synthetic peptide spots, an improved tag lacking this limitation was found, and denoted Strep II (Schmidt et al., 1996). Structural analysis of the complex revealed that a glutamate side chain instead could provide the salt bridge in this case. Since the affinity of the Strep II for streptavidin was slightly weaker (KD= 72 µM) than that of the original Strep I tag, an attempt to improve the capacity of streptavidin was made. Residues 44-47 in the flexible loop of streptavidin close to the binding pocket were subjected to random mutagenesis and an optimized streptavidin (StrepTactin) with high affinity (KD∼10-6) to Strep tag II was found (Voss et al., 1997). In order to find a biotin mimicking peptide with affinity enough to purify proteins from complex mixtures when utilizing only small amounts of immobilised streptavidin, an ultra high complexity peptide library was used for in vitro selection. A new, 38 amino acid long streptavidin binding peptide (SBP) with a dissociation constant to streptavidin of 2.5 nM was found (Keefe et al., 2001, Wilson et al., 2001). In order to reduce the length of the tag a new library was screened and the 9-mer peptide, Nano tag9 (KD= 17 nM) and the 15-mer, Nano tag15 (KD= 4 nM) were found (Lamla et al., 2002). For several reasons, the use of the Strep-tags has widely increased during the last years (Terpe, 2003). The purification conditions are highly variable; chelating agents, mild detergents, reducing agents and salt up to 1M can be added to the buffers. However, there are some constrains in selectivity since naturally biotinylated host proteins may be co-purified (Stahl, 2003).

3.2.2.4. Metal Binding Peptides IMAC is a powerful technique based on interactions of some amino acids with immobilised transition metal ions. Although still not fully understood, the interaction depends on the formation of coordinated complexes between transition metal ions (Ni2+, Zn2+, Cu2+, Co2+) and histidine or, although weaker, to cysteine residues. The metal ions involved in the interaction can be immobilised on chelating solid supports (Ford et al., 1991). Histidine is the amino acid that exhibits the strongest interaction with immobilised metal ion matrices, as electron donor groups on the imidazole ring in histidine readily form coordination bonds with the immobilised transition metal. Moreover, histidine represents one of the rarest amino acid in natural proteins which makes the interaction rather specific (Hearn et al., 2001). Chelating ligands are relatively inexpensive and capable of high metal ion loadings, which permit high protein-binding capacities. IMAC matrices can easily be regenerated by adding a buffer containing the specific metal ion. The ligands are very stable over a wide range of temperatures and solvent conditions. However, reducing and chelating agents must be avoided as these readily displace chelated metal ions (Yip et al., 1989). The loss of metal ions is more pronounced at lower pH values. Apart from leading to reduced adsorption capacity, metal ions that leak from the sorbent can interfere with the target protein e.g. inactivate sensitive enzymes. Histidine residues bind to metal ions only when the imidazole group is not protonated, thus at neutral or alkaline pH. Consequently, elution can be achieved by decreasing the pH to 4-5. Alternative elution of pH sensitive proteins can be done competitively using imidazole, although this may cause precipitation (Hefti et al., 2001).

32

The concept for this pseudo-affinity purification of proteins was presented by Porath (Porath et al., 1975) already in 1975 but it has found deeper interest later on, due to the evolving of genetic engineering methods. The development of a new, more stable adsorbent, nitrilotriacetic acid (NTA), was also a breakthrough for this method (Hochuli et al., 1987). Smith and co-workers were among the first to use different histidine containing peptides for purification and demonstrated high affinities for Co2+, Ni2+ and Cu2+ ions (Smith et al., 1988). A maximum of two histidines are able to bind concurrently to the free coordination sites of one immobilised metal ion (Hearn et al., 2001) but more histidine residues will enhance the probability for the correct orientation desired for coordination interactions (Chao et al., 1998). Nowadays, the hexahistididyl peptide, His6 (Fig. 7B), first employed by Hochuli, is one of the most commonly used tags for facilitated purification of recombinant proteins (Hochuli, 1988). Due to its structure independent interaction, the His6-tag will bind also under denaturing conditions, making the technique suitable for direct purification of solubilised inclusion bodies (Sherwood, 1991). Unfortunately, there are a couple of endogenous E. coli proteins that have metal binding capacity, which impair the specificity of the His6-tag. However, a high salt buffer can reduce non-specific binding and a pre-elution step with a low concentration of imidazole or EDTA can be used to elute contaminating proteins containing histidine residues (Westra et al., 2001). The repetition of the same amino acid residue, as in the His6, are known to destabilize protein tertiary structures (Hearn et al., 2001). Consequently, the His6 tag might interfere with the expression or folding of the target protein (Chao et al., 1998, Woestenenk et al., 2004). Moreover, the small and unstructured tag sometimes becomes buried within the internal structure of folded fusion proteins, and thereby the aimed subsequent purification is hindered (Kronina et al., 1999). Strong binding to IMAC resins can also occur when two histidine residues are situated at an optimal position in a structured tag e.g. on subsequent turns of an alpha helix. Several efforts has been put on constructing a more structured His peptide (Hearn et al., 2001). A naturally occurring peptide of 19 amino acids derived from chicken lactate dehydrogenase is efficiently bound to IMAC resins and has been denoted the His Affinity Tag (HAT) (Chaga et al., 1999). It has a smaller overall net charge and higher solubility than other His-tags. Purification can still be performed under native as well as denaturing conditions when using this tag. Exploring another metal-protein interaction, Thorn designed an affinity matrix based on the bis-arsenical fluorescent dye FlAsH (Fluorescein arsenical helix binder) that specifically interacts with short alpha helical peptides containing the sequence CCXXCC (Griffin et al., 1998, Thorn et al., 2000). This sequence is rarely found in proteins and thus this motif can be used selectively to bind to this arsenical dye. However, since elution is achieved using low (mM) concentrations of a dithiol, target proteins that are dependent on correct disulfide bonds are not suitable for purification with this tag, as they will have destroyed structure.

3.2.2.5. Charged Peptides As earlier discussed (section 3.1.1.1), one of the most widely used techniques in protein purification, both in small and large scale, is ion exchange chromatography (IEXC). This is mainly due to the advantageous properties of the resin. IEXC resins provide high capacity, good resolution and their robustness allows them to be cleaned and reused a large number of times. Moreover, due to simplicity of the ligands and high production quantities, IEXC resins are relatively cheap (Hearn et al., 2001). However, each time a new protein is to be purified the ion exchange purification method must be optimized to the individual charge pattern of the target protein. The incorporation of amino acids with charged side chains (e.g. Lys, Arg, Glu and Asp) makes the recombinant protein more charged and hence generates a shift in pI (Flaschel et al., 1993). Exploiting this principle, peptides containing highly charged amino acids may be used as purification tags in order to recover proteins selectively by IEXC (Sassenfeld et al., 1984, Smith et al., 1984, Sassenfeld, 1990, Stubenrauch et al., 2000, Kweon et al., 2002). In general, the use

33

of small, charged peptide purification tags are well suited for the production of therapeutic proteins as this type of fusion tag seldom perturb the stability or conformation of the target protein (Hearn et al., 2001). Although employing poly-charged peptides, charge interactions provide a low selectivity compared to affinity systems. Another limitation with this system is that a charged tag has propensity for charge-charge interactions also with other proteins. Particularly in buffer solutions with low ionic strength, this can lead to precipitation during the purification procedure (Hearn et al., 2001). The concept of using charged fusion tails for purification was pioneered by Sassenfeld and Brewer, who studied the charge contribution of positively charged amino acids located at the C-terminus of a recombinant protein (Sassenfeld et al., 1984). A penta or hexa peptide of the most basic amino acid, arginine, was fused to the C-terminus of a fusion protein. Contrary to most endogenous E. coli proteins, that have a neutral or low pI, the polyArg fusion proteins could bind to a cation exchanger at alkaline pH values. The captured proteins could then easily be eluted with a linear NaCl gradient. Charged amino acids offer a relative ease of peptide tag removal since many proteases recognizes these amino acids. After completed purification, the polyArg fusion proteins can be treated with Carboxypeptidase B to remove the polyArg tag. However, special precautions have to be taken to avoid in vivo degradation of the tag. It is often preferable to express the polyArg fusion proteins in a strain lacking the outer membrane OmpT protease (Ford et al., 1991). Unfortunately, the polyArg system often suffers in loss of expression due to high rates of plasmid instability or reversion. In order to overcome some of the degradation and plasmid instability problems using positively charged tags, peptide tags based on negatively charged amino acids have been developed (Dalboge et al., 1987, Zhao et al., 1990, Niederauer et al., 1994, Zhang et al., 2001). Unlike the positively charged tails, polyAsp tails seem to be more stable and in general more resistant to proteases (Ford et al., 1991). Since a positively charged anion exchanger is used in this case, nucleic acids from soluble cell extracts are often co-purified. To avoid this contamination, a pre-treatment step with nuclease followed by diafiltration can be introduced (Ford et al., 1991). However, the main application of charged tags has been recovery by polyelectrolyte precipitation (Glatz, 1990, Zhao et al., 1990).

3.2.3 Functional Engineering of a Domain Apart from increasing the selectivity of a protein by fusion to a purification tag, it is also possible to engineer the target protein to provide desirable properties. In a study by Egmond et al. the surface of the protease subtilisin from Bacillus lentus was engineered to include a larger number of charged amino acids (Egmond et al., 1994). It was shown that while keeping the overall charge and pI-value of the protein constant, a more charged surface allowed for increased adsorption to an ion exchange resin compared to wild-type subtilisin. Such a strategy is laborious, compared to the use of purification tags, since it both requires engineering of each individual target protein and also analysis of the resulting mutant to ascertain that its function has not been changed. However, in addition to the use of domains from nature as tags in protein purification, it is possible to endow protein tags with desired properties by protein engineering methods. This allows for tailor making of tags to suit a particular production and purification scheme.

In order to create a universal purification handle for ion exchange chromatography, which would be independent of the properties of the target protein, a domain with highly charged surface was developed by protein design (Fig. 7C) (Graslund et al., 2000). Since the use of this tag, Zbasic, is further explored in the present investigation part of this thesis, the background of this tag will be more deeply discussed. By taking advantage of an easily foldable protein domain for grafting of charges, the potential to achieve a proteolytic stable purification tag should be higher compared to the earlier published

34

unstructured charged tails (section 3.2.2.5). The scaffold for a charged domain should preferably be based on a protein that is small, monomeric, capable of independent folding, robust and easily engineered. Solvent exposed surfaces of bacterial receptors, in particular those containing helix bundle structures, can be unusually stable (Alexander et al., 1992). Moreover, theses conformations are relatively tolerant to changes in the side chains of residues that are not involved in helix-helix packing interfaces (Cedergren et al., 1993). In a previous project, a compact 58

amino acid (7 kDa) three-helix bundle protein, Z (Fig 11A, F, page 55), derived from the B domain of staphylococcal protein A (section 3.2.1.3) (Nilsson et al., 1987)) was used for the design of a charged purification handle (Graslund et al., 2000). The Z domain is not dependent on intramolecular disulfide bridges for structural stability and is also highly soluble. Moreover, Z is stable against proteolysis in a number of different hosts (Stahl et al., 1997) most likely due to its constrained alpha-helical nature. Mutational studies performed on the Z domain have shown that the three-helix bundle structure is highly tolerant to multiple, random amino acid substitutions directed to a region involved in binding to the Fc part of immunoglobulins. Due to its favourable properties, the Z domain has earlier been used as a scaffold for combinatorial mutagenesis to isolate variants with novel binding specificities, using phage display selection technology (Nord et al., 1995, Nord et al., 1997). From such libraries, Z domain-derived affinity proteins (denoted affibodies), directed to different targets, have been selected (Nord et al., 1997, Gunneriusson et al., 1999, Nord et al., 2000). The 13 residues that confers the Fc-binding involves two of the helices and covers a surface area of 800 Å2 that is similar in size to the surfaces involved in many antigen antibody interactions (Rees et al., 1994). The Z domain is slightly acidic (pI 5.16) but by introducing charged amino acids at the above mentioned positions, it would obtain a highly charged surface (Graslund et al., 2000). Distributing charges in a predicted way over a large surface into a highly ordered structure would increase the strength of the interaction with the negatively charged matrix. A proof for that the Z domain can accommodate numerous basic amino acids on its surface came from selection experiments against the highly acidic target protein EB200, derived from Plasmodium falciparum (Ahlborg et al., 1997). Panning experiments resulted in the isolation of a highly basic variant of Z, denoted ZEB4, where 6 of the 13 randomized residues were mutated into positively charged amino acids. Unfortunately, 2 of the other randomized residues were cysteines, amino acids that are undesirable in a purification handle. Therefore, the first variant of a positively charged purification tag, denoted Zbasic1 (Fig. 11B, F, page 55), was based on the selected ZEB4 variant, but with the point mutations C25S and C32S, replacing the cysteines for serines. It was further decided to design a second variant with an even higher positive surface charge. This time all residues that had been used for the construction of the affibody library were considered for exchange into charged amino acids. It was, however, thought that the substitution of some of these amino acids might lower the conformation stability of the three-helix bundle fold (Bryson et al., 1995). Two structurally important residues are H18, that possibly form the C-cap of helix one, and E25, that is involved in N-capping of helix two and may form a salt bridge to helix 3 in the wild-type domain (Olszewski et al., 1996, Tashiro et al., 1997). These two residues were left unmodified while the remaining 11 positions were subjected to exchange. Comparison of the two basic amino acids; arginine and lysine, reveals that the positive charge is delocalized over a larger van der Waals area in the urea group of arginine, whereas in lysine it is concentrated to a smaller ammonium group (Matthew et al., 1985). Moreover, arginines have the potential to more readily participate in a hydrogen bonding network since the urea group has two nitrogens that can participate in hydrogen bonds, whereas the ammonium group in lysine only has one (Matthew et al., 1986). Also the pKa of arginines in model proteins are higher (12.0) than corresponding lysines (10.4-11.1) (Creighton, 1993). Thus, in the second positively charged ion exchange handle, denoted Zbasic2 (Fig. 11C, F, page 55), arginines was incorporated in 11 positions of the former Fc binding surface of domain Z (Graslund et al., 2000).

35

3.3. CLEAVAGE OF FUSION TAGS A fusion tag can induce unwanted immunogenic response to a therapeutic protein and complicate structural and functional determination of uncharacterised proteins etc. This has hampered the use of fusion tags, especially in pharmaceutical industry. Consequently, the intended use of the target protein sometimes requires removal of the purification tag. If the amino acid sequence between the tag and the target protein includes a specific cleavage site, protelysis can occur as a result of disruption of the scissile peptide bond between two amino acids. Tag removal can either be performed by chemical or enzymatic methods (Carter, 1990, Hearn et al., 2001). Chemical methods for cleavage were developed first and are usually cost-effective and relatively easy to scale up. However, the chemical hydrolysis often requires rather harsh conditions that might denature the protein and can cause chemical modifications on the side chains (Hearn et al., 2001). Additionally, chemical cleavage sites are specified by only one or two residues and are therefore often found within the target protein, which limits chemical methods to the release of peptides or smaller proteins.

3.3.1 Specific Cleavage Even though considerably more expensive, enzymatic proteolysis is usually preferred since it can be performed under physiological conditions and generally gives more selective cleavage. A large repertoire of different proteases with varying length of the recognition sequence and specificity is now available. Digestive proteases, such as trypsin that will cleave essentially all peptide bonds in a protein where the carboxyl moiety is a lysine or arginine (Shine et al., 1980), are not ideal for this purpose since the occurrence of these amino acids generally is proportional to protein size. Regulatory proteases, such as thrombin and factor Xa that are naturally involved in blood clothing (Nagai et al., 1984, Jenny et al., 2003), will cleave only certain peptide sequences and are thus more useful for this purpose. These two particular proteases only need to recognise basic residues that are located on the N-terminal side of the scissile bond for successful cleavage, and will therefore leave native N-terminals. However, sometimes they display non-specific cleavages. A family of very specific proteases is protease 3C that has been found in several families of viruses (Miyashita et al., 1992). They recognize 7-8 amino acid sequences with small variations. An example of a sequence that is readily cleaved by protease 3C from rhinovirus and coxackie virus is LETLFQ/GP, where proteolysis occurs between the the Gln and Gly residues (Walker et al., 1994). Another choice for N-terminal fusions is enterokinase that specifically recognizes a five amino acid long peptide (D-D-D-D-K-Xi) and cleaves at the carboxyl side of lysine (Prickett et al., 1989). Exopeptidases, such as dipeptidyl aminopeptidase-1 (DAP-1), can preferably be used to remove N-terminal histidine tags containing an even number of amino acids (McDonald et al., 1969). The Sumo Protease, a highly active and specific cysteinyl protease that is a recombinant fragment of ULp1 from Saccharomyces cerevisiae, recognizes the tertiary structure of the ubiquitin-like protein SUMO rather than an amino acid sequence (Malakhov et al., 2004, Butt et al., 2005). As a result, SUMO protease never cleaves within the fused protein of interest. Most of the recognition sequences for recombinantly utilised proteases are located on the N-terminal side of the scissile bond and therefore the fusion proteins are usually constructed with the tag at the N-terminus of the target protein. Several proteases cleave without leaving any additional residue on the resulting N-terminus of the target protein. However, some proteases, e.g. protease 3C, interact with amino acids on both sides of the scissile bond, which means that the released product will not have a native N-terminus, which in some cases is a requirement. Enzymatic cleavages are sensitive to steric factors and the cleavage might be inhibited by inaccessibility of the recognition sequence within the fusion protein to be cleaved.

36

Prolines are included in many recognition sequences, which induce bends of the backbone of the amino acid chain and thereby make the target sequences more accessible to the protease. It has also been shown for Thrombin and Factor Xa that the addition of an unstructured amino acid sequence, e.g. PGISGGGGG, next to the scissile bond improves the cleavage efficiency (Guan et al., 1991). Also, small proteases usually cleave more efficiently at sites that are inaccessible for larger proteases (Walker et al., 1994). Another important issue for successful proteolysis is the temperature needed for cleavage. Some proteases (e.g. Protease 3C) are active at +4 °C and thus allow the cleavage to be performed at low temperatures. Other proteases need elevated temperatures for efficient cleavage, which can cause aggregation or denaturation of the fusion protein, which sometimes result in cleavage at secondary sites within the protein of interest.

3.3.2 Cleavage Strategies The proteolytic process can be performed using several different setups as illustrated in figure 8. After completed cleavage, strategies for removal of the released tag and the proteolytic enzyme itself must be developed. If cleavage is performed in solution (Fig. 8A) the released affinity tag can be captured by passage of the cleavage mixture over an affinity matrix, while the target protein will be obtained in the flow through. Another strategy is to perform the proteolysis on-column (Fig. 8B), while the purification tag is still adsorbed to the affinity matrix, leading to that target protein will be released. On-column proteolysis thereby also facilitates elution that might otherwise require harsh conditions. However, this strategy is limited to cases where the affinity ligand coupled to the matrix is insensitive to the protease. In both proposed schemes (Fig. 8A, B) the protease is still mixed with the target protein and must be removed if a pure target protein is required. This can be facilitated using a recombinant protease produced in fusion to an affinity tag, preferably the same as the one that is to be removed from the target protein. Both tag and protease can then be captured in a single chromatographic step (Walker et al., 1994).

Figure 8. Different process strategies for cleavage of fusion proteins The protease is symbolised as a scissor, the target protein is shown in white and the tag in dark grey. A. Cleavage in solution. B. On-column cleavage. C. Cleavage using an immobilised protease. D. Intein mediated cleavage.

Another major disadvantage with enzymatic cleavage is that the proteases are very expensive. Even though recombinant production in bacteria and use of purification tags as well to the proteases has improved the availability, the cost is still considerable. If the protease is immobilised on a solid support in a column (Fig. 8C) it can be reused several times to reduce the cost (Suh et al., 2003). The purified fusion protein is passed over the column and proteolysis occurs as the fusion protein makes its way through the proteolytic column. Moreover, while

37

immobilised onto a chromatographic matrix, the protease molecules are separated from each other which prevents intermolecular interactions. This can reduce the risk of autolysis as well as aggregation (Katchalski-Katzir, 1993, Calleri et al., 2004) and allows thus a high concentration of the protease. Many enzymes also show an improved stability while attached to a solid support (Suh et al., 2003), possibly also due to the structural rigidity imparted by multi-point linkages to the matrix. Using this setup the protease is easily separated but still a second chromatographic step is necessary to remove the released tag. Another combined strategy for elution and cleavage can be provided by the insertion a self-splicing intein element that removes itself from the target protein upon induction (Fig. 8D) (Chong et al., 1997). These intein domains normally excise themselves from a larger polypeptide precursor with subsequent joining of the two loose ends, analogous to an intron in an mRNA sequence. The first described intein was the gene product of WMA1 gene from Saccharomyses Cerevisiae (Pelletier et al., 1998) that was engineered to only release its N-terminal fusion partner (Sieber et al., 1998). After affinity capture, the target protein can be released from its handle through the proteolytic activity of the intein domain using a reducing agent such as 1,4-dithiothreitol (DTT). The advantage of using an intein compared to a protease is that it is much cheaper and does not require any second chromatographic step since there is no protease to remove. Potential drawbacks are the large size of the intein domain and that the induction often demands a reducing agent that may affect the function of the target protein. Also, this kind of cleavage may often require extended incubation times (Andrews et al., 2004).

38

4. FOLDING AND MISFOLDING OF PROTEINS Newly synthesized proteins need to fold, often into intricate and close-packed structures, in order to attain biological activity. The transfer of information from amino acid sequence to a three dimensional structure is still not understood. Folding of a protein generally appears to be a self-assembly process as first described by Anfinsen in 1973 (Anfinsen, 1973). In addition to the question of how the information for the unique native state is encoded in the amino acid sequence, there has to be a second question, namely how a protein can find such a state in a finite time (the Levinthal’s paradox) (Levinthal, 1969) .

4.1. FOLDING PATHWAYS Probably there are pathways that guide the protein to fold into its native state, where it is thermodynamically stable. There has been several models proposed to describe the folding process e.g. the nucleation model (Wetlaufer, 1973), the diffusion-collision-adhesion model (Karplus et al., 1976), the jigsaw puzzle model (Harrison et al., 1985), the frame work model (Ptitsyn, 1973), the molten globule model (Ohgushi et al., 1983) and the energy landscape and folding funnel model (Wolynes et al., 1995). None of the proposed models has become generally accepted though. From available experimental data it seems not to be possible to build a simple kinetic folding model that is applicable to all studied proteins. Although a detailed description of the folding pathway is still missing, some steps in the folding process seem to have been established. The number of possible three-dimensional conformations of a particular polypeptide is decreasing during the folding reaction due to non-covalent interactions such as van der Waals contacts, salt bonds, polar interactions, electrostatic interactions and hydrogen bonds. For most proteins, the folding involves one or more partially folded intermediates prior to the formation of the completely folded native state. The experimental tools required to probe the intermediate states along the folding pathway have only begun to become available in the past decade or so (Dobson, 2003). These intermediates have high secondary structure content, where some of the native interactions are established. However, a unique, tight packing of side chains is absent and the fully compacted conformation has thus not been reached. Although a variety of forces contribute to the stability of these intermediates, the formation of a hydrophobic core is thought to play a dominant role (Kauzmann, 1959, Tanford, 1997). This core is formed by the packing of non-polar side chains in the interior of a protein or protein domain. It is not formed because of attractive forces between hydrophobic groups, but rather by the hydrophobic effect, which is generally assumed to be entropic in nature (Dill, 1990). In general, small (<100 residues), single domain proteins can easily reach a native conformation and have fast folding kinetics (Jackson, 1998, Baneyx et al., 2004). When studying proteins with cysteine or proline residues, the rate determining step of the folding process often is that of proline isomerization or disulfide bond rearrangement (Georgiou et al., 1996). The folding of larger multidomain proteins is often more unpredictable (Cabrita et al., 2004). Incompletely folded proteins must inevitably expose at least some regions that are buried in the native state, and are therefore prone to inappropriate interactions leading to aggregation.

39

4.2. FOLDING MODULATORS WITHIN CELLS The main difference between protein folding in vitro and in vivo is that biology has evolved mechanisms to control and regulate the entire protein process within living systems. In order to facilitate folding there are some assisting proteins, foldases and chaperones, available in the cell. Foldases have a catalytic activity to accelerate potentially slow steps in the folding process and thereby prevent aggregation. One of the most important of these folding catalysts is peptidylprolyl isomerase, which increases the rate of cis-trans isomerization of peptide bonds involving proline residues. Proteins with prolyl peptide isomerase activity have been found in most organisms and in different cellular compartments (Gething et al., 1992). For many secreted proteins, disulfide bonds are important for stabilization of the tertiary structure and their assembly into multimeric structures (Seckler et al., 1992). There are a number of characterized proteins that enhance the rate of formation and reorganization of disulphide bonds in vivo. In eukaryotes, one of those is Protein Disulfide Isomerase (PDI) (Freedman, 1989, Noiva et al., 1992). The periplasmic protein DsbA, discovered in 1991 by Bardwell et al, is assumed to serve as an oxidant of secreted proteins in E. coli (Bardwell et al., 1991). In the cellular milieu there is also a stringent quality-control system to ensure that the misfolded products are either shielded from each other or targeted for degradation before they cause harm. Molecular chaperones are present to prevent aggregation of misfolded proteins. Proteases, on the other hand, are there to degrade irreversibly damaged proteins. The chaperones have not the capacity to actively refold the protein to the correct structure, but they lower the aggregation potential by repeatedly bind and release misfolded proteins (Ruddon et al., 1997). Thereby, the protein is allowed to successively reach a kinetically favourable state and finally the active conformation. The role of the chaperones is thus to improve the efficiency of the folding rather than to govern the folding process to a specific conformation. Some chaperones interact with nascent chains as they emerge from the ribosome, whereas others are involved in guiding later stages of the folding process. In the E. coli cytoplasm, protein folding involves three chaperone systems: trigger factor, DnaK-DnaJ-GrpE and GroEL-GroES (Hartl et al., 2002). The ribosome-associated trigger factor (TF) is the first to interact with nascent chains (Maier et al., 2005). TF is ATP independent and does not require any co-chaperone. The cradle of TF has a hydrophobic inner surface, which enables TF to interact with hydrophobic patches of nascent chains, thus preventing them from getting trapped in unproductive folding intermediates. Moreover, due to its unusual open shape and relatively weak affinity for the ribosome, TF is in a position to precisely control the handover of nascent chains to downstream translational folding factors. A plausible hypothesis assumes that TF may effectively coordinate the stepwise co-translational folding of multidomain proteins (Maier et al., 2005). DnaK, together with its co-chaperones, is involved in the initial binding of some aggregation prone nascent polypeptides as they proceed off the ribosome (Bukau et al., 1998). These chaperones can in cooperation with the nucleotide exchange factor GrpE, achieve the complete folding of some of these proteins. Other proteins require the additional actions of the GroEL/GroES system. GroEL is a chaperonin, and thus contains a cavity in which incompletely folded polypeptide chains can enter and undergo the final steps in the formation of their native structures while sequestered and protected from the outside world (Hartl et al., 2002). However, it is still not clear whether proteins fold to their native structure while they are bound to chaperones or whether they have to be released into solution to complete folding. Misfolded proteins are not only a target of chaperones for folding or refolding attempts, but also for proteolytic degradation (Cubarsi et al., 2001). The proteolytic susceptibility of insoluble proteins is also of interest to understand the molecular features of misfolding. Unfortunately very little is known about the digestion mechanism of insoluble proteins.

40

4.3. INCLUSION BODIES The development of techniques for expressing a protein in a simple organism such as E. coli seemed to provide an unlimited supply of any conceivable protein. However, in the cytoplasm of a simple organism as E. coli, folding of overexpressed recombinant proteins is an extraordinary challenge. Therefore, high recombinant expression levels usually end up with the desired protein trapped in large insoluble aggregates, called inclusion bodies (Marston, 1986). Several studies have shown that inclusion bodies result from the association of folding intermediates rather than the native or fully denatured form of the protein (Cleland, 1993, Villaverde et al., 2003). The aggregated state represents a kinetic trap for a folding intermediate, which otherwise would proceed to its native state (Cleland, 1993). Inclusion bodies occur naturally in both prokaryotic and eukaryotic cells, but they are especially common when foreign proteins are overproduced in bacterial systems. Most inclusion bodies are found in the bacterial cytoplasm but secreted proteins can also form aggregates in the periplasmic space (Georgiou et al., 1986). Inclusion bodies are very dense particles that can reach diameters up to the micrometer range and therefore can be seen in light microscope (Mayer et al., 2004). Interestingly, in most cases only one or two large inclusion bodies are observed in an E. coli cell (Creighton, 1994). Bacterial inclusion bodies often have been considered as homogeneous, inert depositions of insoluble protein. However, recent results indicate that protein aggregation into bacterial inclusion bodies is a specific event based on fine molecular discrimination processes (Carrio et al., 2005). Inclusion bodies seem to be more ordered than previously thought, displaying clear amyloid-like properties. Interestingly, an increase in beta-sheet content and decrease in alpha-helix content have been detected in inclusion bodies in comparison to the native product (Przybycien et al., 1994, Carrio et al., 2005).

4.3.1 The Formation of Inclusion Bodies Mechanisms governing correct folding versus inclusion body formation are still not well understood. Several attempts have been made to elucidate if there are particular properties of a protein that are connected with inclusion body formation (Wilkinson et al., 1991, Idicula-Thomas et al., 2005), however, mostly without success. In fact, there seem to be no simple generic relationship between the formation of inclusion bodies and any single factor. Analysis of how amino acid substitutions affect the inclusion body formation indicates that the folding pathway is more important than the stability of the fully folded protein itself (King et al., 1996). Probably, inclusion body formation depends on the specific folding behaviour rather than on the general characteristics of the protein (Fahnert et al., 2004). There may be several possible reasons for intracellular aggregation of recombinant proteins in E. coli cells. Most proteins are found in very low amounts in their natural sources and overexpression of even homologous proteins can be sufficient to induce the formation of inclusion bodies (Gribskov et al., 1983). This implies that one reason for the aggregation is the unnatural situation of very high concentrations of folding intermediates (Wall et al., 1995). To find out why eukaryotic proteins are more prone to aggregate in bacterial expression systems, it is interesting to evaluate differences in the process of protein synthesis in eukaryotes and prokaryotes. In eukaryotic systems, folding is in some cases initiated before the completion of protein synthesis, co-translationally, whereas the nascent chain is still attached to the ribosome. Other proteins, however, undergo the major part of their folding in the cytoplasm after release from the ribosome, whereas yet others fold in specific compartments, such as mitochondria or the endoplasmic reticulum, after trafficking and translocation through membranes (Dobson, 2003). In vivo experiments have demonstrated that co-translational protein folding is possible in prokaryotic cells as well (Nicola et al., 1999), even though post-translational folding is more frequent here. In the E. coli cytoplasm transcription and translation are tightly coupled and one protein chain is released from the ribosome every 35th second (Lorimer, 1996). The high speed of translation in

41

bacteria might disfavour co-translational folding. Inclusion body formation is dependent on the rate of expression, the rate of folding as well as on the rate of aggregation. Consequently, a high synthesis rate of a protein with slow folding characteristics may exceed the available resources of foldases, chaperones and proteases (Fahnert et al., 2004). Moreover, prokaryotic organisms do not have the necessary machinery of foldases and post-translational modifying enzymes for complete folding and processing of many eukaryotic proteins. For many recombinant proteins that are produced in prokaryotes, non-glycosylated variants can still fold to a biological active state (Creighton, 1994). Larger, more complex glycoproteins though, have problems to fold if their linked glycans are missing. Glycosylation seems to have little effect on the kinetics or thermodynamics of folding. The modifications rather seem to increase solubility and thereby reduce aggregation during and after folding (Creighton, 1994). Also, the quality-control mechanism evolved in eukaryotes involves a remarkable serie of glycosylation and deglycosylation reactions that enables correctly folded proteins to be distinguished from misfolded ones (Hammond et al., 1995). Many eukaryotic proteins of therapeutic or commercial interest have complex tertiary and quaternary structures that often require the formation of multiple disulfide bonds. Slow folding of these proteins may be expected upon cytosolic expression in the E. coli cell due to the reducing environment (Rudolph et al., 1996). However, incorrect disulphide bridges can not be the only cause of aggregation, since aggregates are formed even when a carbomethylated protein, thus with all cysteines blocked from forming disulphide bonds, are folded (Goldberg et al., 1991). Still, formation of inclusion bodies might be increased without the formation of the correct disulphide bonds, since the stability of many proteins will be lower without them (Fahnert et al., 2004). However, proteins that do not contain any cysteines at all, and thus do not have a three dimensional structure that depends on disulfide bridges, can also form inclusion bodies when overexpressed.

4.3.2 The Prevention of Inclusion Body Formation There are several ways to prevent aggregation of a protein in vivo. A traditional approach involves a reduced synthesis rate of the target gene product, in order to increase the chance of folding into a native state before it has the chance of aggregating with other folding intermediates. This can for example be achieved by using a vector with a low copy number, a weaker promoter or a weaker ribosome binding site (Galloway et al., 2003). Controlled transcription can also be accomplished with a low concentration of inducer (Baneyx et al., 2004). An alternative strategy is to decrease the temperature at which the recombinant protein is expressed (Schein, 1988, Chalmers et al., 1990, Strandberg et al., 1991). The use of low expression temperatures has the combined advantages of slowing down transcription and translation rates, as well as reducing the strength of the endothermic hydrophobic interactions that contribute to protein misfolding (Cleland, 1993). Although promising in theory, the presence of a small number of rare codons does not affect the rate of protein synthesis very much in practice (Fahnert et al., 2004). In the case of multiple consecutive rare codons situated near the N-terminus of a coding gene sequence, some beneficial effects on the protein expression have been observed though (Imamura et al., 1999). Changes in the fermentation media composition, e.g. addition of non-metabolisable sugars such as sucrose and raffinose, can also affect inclusion body formation (Georgiou et al., 1986). In E. coli only 10-20% of the host proteins receive help by chaperones for their folding and thus the amount of available natural chaperones is limited (Ewalt et al., 1997). Accordingly, a reduced inclusion body formation can be obtained by co-expression of chaperones and foldases (Hockney, 1994, Wall et al., 1995). The beneficial effect of an increase in the intracellular concentration of TF, DnaK-DnaJ (with or without GrpE) and GroEL-GroES is well documented and a number of plasmids vectors are available (Baneyx et al., 2004). However, contradictory findings were recently reported, showing that binding of TF and DnaK to nascent firefly luciferase chains redirects the

42

folding of this protein from an efficient eukaryotic co-translational mode to a slower post-translational pathway that is accompanied by aggregation (Agashe et al., 2004). Obviously, certain multidomain proteins that have evolved to take advantage of co-translational folding do not benefit from chaperone overproduction (Baneyx et al., 2004). Aggregation is most prominent for heterologous proteins containing disulfide bonds in their native state. Stable disulphide bonds do not form in the cytoplasm of E. coli, owing to the reducing redox conditions provided by the combined action of thioredoxins and glutaredoxins systems. Identification of the members of these systems and subsequent elucidation of their roles in disulfide bond reduction, has made it possible to manipulate the E. coli cytoplasm in some strains to be less reducing and thus more suitable for production of oxidized proteins (Derman et al., 1993, Baneyx et al., 2004). If optimization of the expression conditions is unfruitful, another strategy is to try to alter the protein of interest itself. Protein engineering using strategies like molecular evolution (Mosavi et al., 2003, Ito et al., 2004) or rational protein design (Mitraki et al., 1992, Murby et al., 1995) can be used to enhance the solubility of a protein. Fusion to a protein that has high internal solubility can also improve the solubility. Several proteins have been used for this purpose with varying success e.g. NusA (Davis et al., 1999), MBP and Theoredoxin (Sachdev et al., 1998). The fusion tag is thought to rapidly fold to a soluble domain, which may accelerate or stabilise the folding of the rest of the molecule, thereby acting as an intramolecular chaperone (Fedorov et al., 1997).

4.3.3 Inclusion Bodies and Medical Science The biological relevance of protein aggregation has been shown to be much higher than previously realized. In eukaryotic cells, protein aggregation can lead to the formation of deposits, amyloids, which very much resemble inclusion bodies. Misassembly of proteins into insoluble amyloid deposits is now considered as the fundamental cause behind some debilitating human disorders such as Alzheimer’s disease, diabetes, cystic fibrosis, transmissible spongiform encephalopathies, liver cirrhosis and many neurodegenerative diseases (Dobson, 2003, Carrio et al., 2005). However, the ability of polypeptide chains to form amyloid structures is not restricted to the relatively small number of proteins associated with recognized clinical disorders; it rather seems to be a generic feature of polypeptide chains (Dobson, 2003). A convincing evidence for this is that amyloids can be formed in vitro by many other peptides and proteins, including such well-known molecules such as myoglobin, and also by homopolymers such as polythreonine or polylysine (Dobson, 2001, Fandrich et al., 2002). Even though the ability to form amyloid fibrils seems to be generic, the propensity to do so under given circumstances can vary remarkably between different sequences (Dobson, 2003). Proteins able to form amyloid fibrils in vitro are usually also produced as inclusion bodies in E. coli. Overall, the study of bacterial systems should be seriously considered when exploring the molecular basis of protein misfolding and aggregation. Bacterial inclusion bodies provides a fast, simple and biologically relevant experimental model that could serve to better understand biological situations which are difficult to approach experimentally, such as protein deposition in amyloid diseases. Understanding the protein aggregation phenomenon might thus eventually provide valuable tools for developing protein aggregation antagonists to control several diseases. The recently realized fact that protein aggregation can be reverted even from poorly structured elements such as inclusion bodies (Carrio et al., 2001) encourages further studies. Also, controlled proteolysis of insoluble polypeptides might eventually represent an alternative approach to remove pathogenic deposits of proteins (Cubarsi et al., 2001).

43

4.4. FOLDING SCREENS Although there have been several attempts to increase the yield of soluble proteins, these strategies do not result in the same degree of success for all polypeptides (Wall et al., 1995, Nishihara et al., 2000). Moreover, when screening for the best conditions for folding of a specific protein, it is difficult to determine the yield of native protein obtained if no biological assay is available. Recently, several new protein solubility screens have been developed that do not require any structural or functional knowledge about the target protein (Waldo, 2003). However, these methods, as described below, do not provide information about obtained structure or activity of the target, only whether it is soluble or not.

4.4.1 Gel Based Methods If the soluble and non-soluble parts of the cell content are physically fractionated, the proteins in these fractions can be directly separated by SDS-PAGE. Thereafter, protein quantification can be made by densitometric scanning of coomassie stained or immunologically visualised gels (Knaust et al., 2001, Cornvik et al., 2005). Immunological detection increases the sensitivity but on the other hand, this method requires specific antibodies and lengthy cycles of binding, washing and probing. The major drawbacks of these methods are the difficulties concerning total cell disruption and separation without disturbing the in vivo distribution. To avoid cell disruption, the expression can be performed using a cell-free expression system instead of the classical bacterial expression system. In vitro expression systems also facilitate the control of several parameters such as having the same amount of plasmid DNA template (Busso et al., 2004). Recent reports claim that some results from screening in vitro expression can be valid for in vivo expression as well (Busso et al., 2004). However, obviously the contribution of folding moderators inside of the cell can not be assessed in such systems.

4.4.2 Cell Response Systems The complicated living cell must be considered as the most accurate model. As mentioned earlier, cells tend to respond to over-expression using several stress response systems. Certain genes are specifically activated when misfolded proteins are expressed. The corresponding responsive promoter can be cloned in front of a gene product with an easily measured function. In this way protein misfolding can be measured in vivo using the own bacterial response to protein misfolding (Lesley et al., 2002). The main disadvantage of the stress-response system is that the stress reporters give a positive signal to low solubility and no signal when a lot of soluble protein is expressed.

4.4.3 Reporter Systems In order to get a strong positive response, the target protein can be expressed as a genetic fusion to a reporter protein with an easily detected function or biological activity. These approaches are simple to implement but might perturb the inherent solubility of the test protein to which it is fused. The solubility of test proteins expressed in the cells as N-terminal fusions with chloramphenicol acetyltransferase (CAT) was positively correlated with the survival of E. coli cells on media containing the antibiotic chloramphenicol (Maxwell et al., 1999). Growing the cells on media containing progressively higher levels of chloramphenicol provides a selective pressure. Since the CAT protein is an obligate trimer, the fusion may lead to formation of higher-order aggregates of otherwise soluble multimeric proteins. It remains to be seen whether the CAT fusion method is applicable to a wider range of targets.

44

In the LacZα (β-galactosidase α-peptide) solubility reporter assay, a peptide of about 100 amino acids, LacZα, is fused to the test protein (Wigley et al., 2001). If the fusion protein remains soluble and the tag is not hidden, the LacZα peptide can bind inactive LacZω, supplied by the host in trans, or in vitro, and thereby complementation will restore β-galactosidase activity. It was expected that the small LacZα tag would have less effect on the folding and solubility of the test protein compared with a larger protein such as CAT. Moreover, since the complementation depends only on the accessibility of the tag in soluble form, this system might be especially useful for monitoring slow protein aggregation processes. However, it later became clear that many partially soluble proteins are insoluble when expressed as LacZα fusions (Waldo, 2003), probably due to the inherent poor solubility and folding characteristics of the split protein used as reporter tag. The most widely used reporter tag is the green fluorescent protein (GFP) derived from the jellyfish Aequorea Victoria (Waldo et al., 1999) (Fig. 9A). Recombinant expression of this gene gives a fluorescent product in a variety of organisms, including E. coli. The first application of GFP and its inherent fluorescence was as a reporter of protein expression (Chalfie et al., 1994). However, it has been shown that GFP recovered from inclusion bodies in E. coli lacks the mature chromophore (Fig. 9B), suggesting that protein destined for inclusion bodies aggregates prior to productive folding of GFP (Reid et al., 1997). Consequently, monitoring fluorescence of intact cells becomes an easy tool to quantify the solubility of expressed proteins (Waldo et al., 1999). Since GFP was isolated from an organism whose natural habitat is in the ocean, the correct folding seems to be temperature dependent (Lim et al., 1995). Several mutants of GFP have been constructed in order to circumvent this limitation (Siemering et al., 1996, Kimata et al., 1997). When expressed alone, most GFP variants folds properly at or below 30°C. However, if an aggregation-prone protein is fused to GFP, aggregation of the recombinant protein domain also leads to a decrease or loss of fluorescence (Waldo et al., 1999). The use of a C-terminal GFP fusion (i.e., protein-GFP) provides that the protein of interest is translated first. Therefore, the folding kinetics of the protein of interest likely determines whether the entire fusion protein folds correctly or not. If the fusion protein aggregates, it apparently quenches the fluorescence signal. One of the major benefits with GFP is that the chromophore is formed auto-catalytically from the primary amino acid sequence. Formation of the chromophore is accomplished through the cyclisation of serine-dehydrotyrosine-glycine within a hexapeptide that is placed in the central cavity of the barrel structure of GFP (Cody et al., 1993). The protein does not require substrates or cofactors to fluoresce, which makes quantification possible without need for invading the cells (Chalfie et al., 1994). Estimates of the time for spontaneous chromophore formation of newly synthesised GFP range from 30 min (Waldo et al., 1999) to 4h (Crameri et al., 1996). Bacterial colonies can be inspected by eye directly on plates and single cells can be thoroughly analyzed in the microscope. GFP fluorescence is efficiently excited with a standard 488 nm argon laser (Maksimow et al., 2002) and is thus easily monitored using fluorescence based equipment. The distribution of fluorescence levels in cell populations can be determined using multi-parameter flow cytometric analysis, usually performed on a Fluorescence Activated Cell Sorter (FACS) (Winson et al., 2000). In figure 9C, the instrument setup for a flow cytometer is illustrated. Flow cytometry is a powerful tool to gather information about living bacteria with single-cell resolution (Maksimow et al., 2002) and provides a convenient assay for GFP fusion proteins within cells. Flow cytometric measurements can be performed directly in bacterial suspensions without disruption of the cells. During the time each cell is present within the laser beam, on the order of ten microseconds, GFP is excited, which makes fluorescent measurements very sensitive (Nolan et al., 1999). The flow cytometer is a quantitative instrument, and the number of cells analysed is recorded. Consequently, time consuming and potentially error prone dilution steps, which are needed for e.g. spectroscopic methods, are avoided. A typical flow cytometric run comprise 10 000 cells and thus simultaneously provides data about the distribution over the entire population, which

45

is valuable information for optimisation of cultivation conditions. Apart from fluorescence, a flow cytometer simultaneously measures other physical characteristics of the single cell as it flow in a fluid stream through the beam of light. The measured light scattering, side-angle (SSC) and forward-angle (FSC), gives indications about physical cell characteristics like internal structure, shape and relative size (Shapiro, 1995). This technique can thus provide valuable knowledge for protein production.

Figure 9. GFP mediated folding screens A. Correctly folded and fluorescent GFP. B. Misfolded non-active GFP. C. Flow cytometric analysis of GFP fusion proteins

46

4.5. IN VITRO REFOLDING If optimizations of in vivo production of properly folded proteins in E coli do not succeed, there is still a second option to acquire well folded recombinant proteins. Several strategies are available to increase the yields of soluble and active protein by in vitro refolding of proteins aggregated in inclusion bodies. Sometimes it is even advantageous to have the target protein destined to inclusion bodies. Expression of a target protein as inclusion bodies can be directly observed by phase contrast microscopy, which avoids the need for initial expression identification by electrophoresis methods (Li et al., 2004). Another example is that when producing a toxic protein it is actually preferable if the deleterious activity is not presented to the host. Under appropriate conditions, the recombinant protein deposited in inclusion bodies constitutes about 50% or more of the total cellular protein (Panda, 2003). Since inclusion bodies have a relatively high density, they can easily be isolated from the cellular proteins by centrifugation (Misawa et al., 1999). Thereby a rather pure product can easily be achieved (Babbitt et al., 1990). It was formerly believed that aggregated proteins retained totally protected from proteolysis. Recently it has been proven that even aggregated proteins are partly susceptible to proteolytic digestion both in vitro (Cubarsi et al., 2001) and in vivo (Corchero et al., 1997). However, inclusion body proteins probably have an increased protection from proteolytic degradation compared to the soluble counterpart. In the refolding process the main components are isolation, washing and solubilisation of the inclusion body prior to transfer of the protein to conditions that favour the correct folding. Different unit operations involved in this process are discussed below.

4.5.1 Isolation of Inclusion Bodies Cells containing inclusion bodies are usually disrupted by high-pressure homogenization or by a combination of mechanical, chemical and enzymatic methods (Burgess, 1996). Direct lysis and solubilisation provides initially a more cost-effective process but also renders more contaminating proteins. Unsolubilsed inclusion bodies have a relatively high density and can therefore be isolated from the resulting cell extracts by centrifugation, filtration or SEC. As the cell membranes contain a number of proteolytic enzymes, it is of great importance to extract these contaminants or at least inhibit their activity. The most generally applicable method for selective extraction is washing of the inclusion bodies with a buffer containing a chelating agent (e.g. EDTA), a neutral detergent (e.g. Triton X-100) and a chaotrop (e.g. urea) (Babbitt et al., 1990). An alternative method is gradient centrifugation, which offers a more precise separation but is technically more complex (Villaverde et al., 2003).

4.5.2 Solubilisation of Inclusion Bodies Inclusion bodies contain aggregated intermediates which form bonds such as van der Waals interactions, hydrogen bonds, disulfide bonds, etc. with each other. These bonds are typically difficult to selectively disrupt since the intermolecular interactions are analogous to intramolecular bonds formed within the native protein. Denaturants such as guanidine hydrochloride or urea are thus utilized to disrupt these aggregates, as these compounds can unfold the protein chains into flexible and disordered structures. GdnHCl is usually preferable to urea for two reasons. First, GdnHCl is a rather strong chaotroph, which allows solubilisation of extremely tightly bound inclusion bodies that are resistant to solubilisation by urea. Second, urea solutions may contain isocyanate, leading to carbamylation of free amino groups of the polypeptide, especially upon long term incubation at alkaline pH values (Hagel et al., 1971). In most cases, solubilisation involves total denaturation of the protein. However, if the protein has substantial amount of native

47

conformation while trapped in the inclusion bodies, it is possible to choose a weaker solubilising buffer that does not disrupt the already formed structures. In the case of cysteine containing proteins, the dissolved inclusion bodies usually form non-native intramolecular and intermolecular disulphide bonds which reduce the solubility. Initially, these bonds have to be disrupted by reducing agents such as DTT, GSH, cysteine, cystamine or β-mercaptoethanol (Misawa et al., 1999). Moreover, the renaturation buffer has to be supplemented with a mixture of the reduced and oxidised forms of low molecular weight thiol reagents such as glutathione, cysteine and cysteamine to provide the appropriate redox potential to allow formation and reshuffling of disulphides (Lilie et al., 1998).

4.5.3 Refolding Strategies The first, and most obvious, step to obtain correctly folded proteins after solubilisation of inclusion bodies is removal of the excess denaturant to create an appropriate environment where proteins can fold spontaneously (Vallejo et al., 2004). What determines the cellular milieu is the solvent, pH, ionic strength, presence of other components such as metal ions or prosthetic groups, other proteins such as chaperones, total protein concentration, temperature etc. This is a complicated environment that is difficult to assess for refolding in vitro. As a first attempt to enhance the refolding yield, easily controllable environmental factors such as pH, denaturant concentration, temperature, protein concentration, reducing agents etc. can be carefully optimized for each individual protein (Ahn et al., 1997, Misawa et al., 1999). Unfortunately, each protein is likely to fold by a different pathway for each condition and it is therefore difficult to develop general rules for refolding (Cleland, 1993). Moreover, protein refolding is not a single reaction since folding competes with other, higher-order aggregation reactions leading to inactive proteins (De Bernardez Clark et al., 1998, Tsumoto et al., 2003). With increasing concentration of denatured polypeptide chains, the rate of the folding reaction remains constant, whereas the rate of aggregation increases. Suppression of the aggregation reaction is thus the main concern to enhance the refolding yield. A straightforward way is to conduct protein refolding at very dilute solutions <<0.1mg/ml. Different strategies used for refolding are depicted in figure 10.

4.5.3.1. Refolding by Dilution To dilute the solubilised protein in a non-denaturing renaturation buffer combines a way to remove the denaturant in the same time as the protein is gradually less concentrated (Fig. 10A). However, this requires large vessels of renaturation solution from where the product has to be concentrated afterwards, parameters that should be avoided in cost effective bioprocessing (Dong et al., 2004). A major improvement of this technique was the development of a method where the solubilised, denatured protein is added in pulses into the refolding buffer (Katoh, 2000). Since it is only the concentration of non-native polypeptides, and not that of native proteins, that is critical, a second batch of denatured proteins can be added to the same refolding mixture when proteins from the first batch are folded. Solvent exchange for removal of the excess denaturant can also be done gradually by the classical methods dialysis (Fig. 10B) or diafiltration. However, the inherent low process rate and the low yields obtained using membrane-based methods due to protein binding to the membranes restricts these applications on a large scale (Clark, 2001).

4.5.3.2. Refolding using Size Exclusion Chromatography Denaturant removal can also be accomplished using chromatographic techniques. Size exclusion chromatography (SEC) allows the protein to escape the denaturing buffer and then either fold to the native state and continue to elute or precipitate to large aggregates that consequently stops migrating (Fig. 10C). These aggregates can subsequently be resolubilised by the following

48

denaturant front, which gives them a new chance to escape and fold (Werner et al., 1994). However, the sharp decrease in denaturant concentration during the feed loading stage inevitably causes aggregate formation. This problem can partly be solved by an additional small volume of denaturant solution immediately after the sample injection (Liu, 2003). The SEC method will only work optimally in those cases where folding is much faster than the gel-filtration process time. However, a new trend is the application of SEC refolding in continuous processes, which can also contain a recycling system for aggregated protein (Jungbauer et al., 2004).

Figure 10. Different strategies used for in vitro refolding Denatured protein is illustrated as dark grey spots while folded proteins are shown as grey peanuts. Lighter areas symbolise non-denaturing conditions while darker areas symbolise denaturing conditions. A. Refolding by dilution. B. Refolding by dialysis. C. SEC mediated refolding. D. Solid phase refolding.

4.5.3.3. Solid Phase Refolding Refolding using a chromatographic process is attractive because it is easily automated using commercially available chromatography systems. Moreover, this step can often be combined with simultaneous purification (Li et al., 2004). A promising chromatographic alternative is the solid phase refolding (also called matrix-assisted refolding) (Fig. 10D) that was initially proposed by Creighton (Creighton, 1985). The protein, unfolded by a denaturant, is reversibly adsorbed to a solid support and upon removal of the denaturant, the bound protein is allowed to refold and, due to the spatial separation of the individual polypeptide chains, aggregation should not occur, even at high protein concentrations. In this way the target protein can be concentrated, instead of diluted, during refolding. Due to its origin of selective binding, this method also enables purification and refolding to be achieved simultaneously. Conditions needed for optimal protein refolding seem to be protein dependent. Slow removal of the denaturant often results in the exposure of intermediate species over a long period of time and hence the propensity for the protein to aggregate is increased (De Bernardez Clark et al., 1998). On the other hand, forcing the protein to adopt its conformation in a limited time frame may also increase its chances of misfolding. Hence, how fast the denaturant concentration is reduced and how long the protein molecules are exposed to intermediate denaturant concentration will determine the rate of folding and the degree of flexibility of the folding intermediates. Solid phase refolding offers an easily controllable milieu regarding buffer composition. The rate of denaturant removal can thus be performed in any suitable way for each individual target protein. Moreover, the final concentration of buffer components can be controlled carefully and the conditions can be adjusted before elution. Different features have been used for solid phase refolding; charges (Creighton, 1985), hydrophobicity (Geng, 1992), metal ion affinity (Rogl et al., 1998) and affinity (Berdichevsky et al., 1999).

49

4.5.3.3.1. Ion Exchange Chromatography used for Solid Phase Refolding Ion exchange chromatography is the most frequently used technique for chromatographic processes (Janson, 1998). Since these matrices are stable and selective even under denaturing conditions, this technique can be used for solid phase refolding as well. Denatured proteins can generally be adsorbed onto ion exchangers at suitable pH and low ionic strength if non-ionic denaturants such as urea are used (Rudolph et al., 1996). By decreasing the concentration of the denaturant gradually, the adsorbed proteins are allowed to refold while adsorbed on the media. Once the refolding is completed, the folded protein is eluted by traditional salt-gradient elution. The electrostatic binding of the protein to the resin is not expected to interfere greatly with refolding, since globular proteins generally have their ionic groups on the surface in the native structure. However, the hydrophobic interaction between the solid support and the polypeptide chain should also be taken into consideration, since an elevated ionic strength promotes hydrophobic interactions. Refolding should be carried out under weakly dissociating conditions, where the protein to be refolded fluctuates between bound and unbound states in the resin (Tsumoto et al., 2003). This can be accomplished by the addition of NaCl or changing the pH. An elegant approach is used in the “two buffer system” (Li et al., 2002), where a decreasing urea concentration is introduced in parallel with an increasing ionic strength gradient, allowing the protein to simultaneously structurally rearrange and elute during its migration down through the column. This allows a single protein molecule to refold gradually in each protein’s optimal urea concentration in the gradient (Li et al., 2004). Considering the importance of denaturant concentration as well as pH in refolding, another strategy, using a dual gradient, was introduced (Li et al., 2002). Elution was started by a gradual decrease in the urea concentration combined with a gradual increase of pH of the elution buffer. The dual-gradient strategy provides an incremental change of the solution environment for the protein refolding and favours formation of the correct disulfide bonds (Li et al., 2004). Moreover, this strategy provides an improved elution by avoiding high salt concentrations and thus avoiding stronger hydrophobic interactions that might lead to aggregation. In order to further minimise constraints of the target protein, solid phase refolding using the own charge of the protein can be avoided by the use of a polycationic fusion tag, e.g. a ten-lysine tag (Kweon et al., 2004).

4.5.3.3.2. Hydrophobic Interaction Chromatography used for Solid Phase Refolding

Hydrophobic interaction chromatography (HIC) has also been successfully used for protein refolding with concomitant removal of contaminating proteins during the renaturation process. Unfolded proteins are applied to the hydrophobic matrix on the column at high salt concentrations. During migration through the column, the protein will pass through several steps of adsorption and desorption, controlled by the salt concentration and hydrophobicity of the intermediates, finally resulting in the formation of the native structure. As successful polypeptide folding also is dependent on undisturbed hydrophobic interactions, the hydrophobic interaction with the matrix should not be too strong as it might prevent proper protein refolding. Several HIC refolding studies have used a silica-based material crafted with polyethylene glycol (PEG) with a hydrophobic end group (Jungbauer et al., 2004). Thereby, PEG also could be responsible for the enhanced refolding yield, as it has successfully been used to improve refolding yields in other applications and is considered to be an enhancing additive (Kim, 2000, Jungbauer et al., 2004).

50

4.5.3.3.3. Immobilised Metal Affinity Chromatography used for Solid Phase Refolding

Since multivalent binding of the protein to a solid phase can hinder the folding reaction, directed immobilisation of denatured His6-tagged fusion proteins on an IMAC resin is a promising alternative. Primary, this strategy requires that the His6 tag must neither block the protein’s folding capability nor have any effect on its native structure. Moreover, to provide a stable matrix, the refolding conditions must be carefully optimized. Neither strong reducing agent nor chelating agents can be used since these compounds readily displace chelated metal ions on the charged IMAC column (Yip et al., 1989). In addition to resulting in a lowered adsorption characteristics of the matrix, lost trace amount of metal ions might catalyse unwanted oxidation reactions leading to mispaired disulphide bonds. Moreover, the long-term stability of the final formulation can be significantly reduced by metal ions since they can act as catalysts for cleavage reactions (Gaberc-Porekar et al., 2001) and also interfere with some enzymes (Misawa et al., 1999). High concentration of buffer components containing strong electron donating groups (e.g. arginine) should also be avoided (Kweon et al., 2004). This has limited a broad application of solid phase refolding using IMAC, because reduction and oxidation of cysteine residues and the addition of the above mentioned components are sometimes indispensable.

4.5.3.3.4. Affinity Chromatography used for Solid Phase Refolding To assure that the captured target protein is able to refold without being hindered by interaction to the matrix it can be fused to a purification tag. The use of a selective tag ensures that the interaction with the matrix is distanced from the target protein and provides the protein freedom to refold in the bulk. This strategy of course demands that the affinity domain has an interaction that is denaturant-resistant. For example, the cellulose binding domain (CBD) retains its ability to bind specifically to a cellulose matrix in 6 M urea and has successfully been used for solid phase refolding (Berdichevsky et al., 1999).

4.5.4 Additives In order to mimic the conditions for in vivo folding, the rate of refolding versus misfolding or aggregation can be manipulated by several additives. The exact mechanism by which an additive affects the folding of a given protein is difficult to assess. Additives may influence both the solubility and the stability of the unfolded protein, folding intermediates as well as the final native state. Furthermore, these additives may affect the rate of folding, misfolding or precipitation. There are certain additives that mainly increase the flexibility or solubility of proteins, while others enhance structure formation. An aggregation suppressor reduces side chain interactions while a folding enhancer in principle enhances these kinds of interactions (Tsumoto et al., 2003). Probably, L-arginine is nowadays the most commonly used refolding agent (Clark, 2001). Even though L-arginine contains a guanidine group, it does not destabilize the native folded structure as much as GdnHCl (Lin et al., 1996). However, the beneficial effect of L-arginine on refolding probably originates from an increased solubisation of folding intermediates (Lilie et al., 1998). Numerous other low molecular weight additives such as detergents, sugars, polymers, protein-stabilizing agents such as glycerol or even low residual concentrations of denaturants have been shown to improve refolding yields. Efficient protein refolding at relatively high concentrations has been achieved by using mixtures composed of detergents (e.g. TritonX, lauryl-maltoside) and phospholipids (Zardeneta et al., 1994, Zardeneta et al., 1994). These mixed micelles contain both polar and non-polar moieties that can interact with various sites on the folding intermediates, thereby preventing misfolding and aggregation. Another combinatorial strategy is the use of a detergent followed by the addition of cyclodextrin (Singh et al., 2002). This method has been

51

claimed to be analogous to a molecular chaperone system in terms of function. The in vitro addition of natural chaperones has also been successful for the refolding of various proteins (Thomas et al., 1997). However, a routine application is limited by their cost, the relatively high chaperone concentration required (at least equimolar to the target protein) and the need for their removal after the refolding procedure (Buchner et al., 1992). However, to somehow overcome these limitations, the folding catalysts can be immobilised onto chromatographic support so that the column behaves like a catalytic folding reactor (Altamirano et al., 1997). A three components chaperone system with GroEL, DsbA and peptidyl-prolyl isomerase immobilised on an agarose gel has been successfully used for oxidative refolding (Altamirano et al., 1999). In some cases, specific cofactors, e.g. metal ions, stabilizes the protein already at the level of folding intermediates and thereby prevent off-pathway reactions. In a similar way, antibodies have been shown to facilitate refolding of the corresponding antigen (Carlson et al., 1992). The antibody may facilitate the formation of the correct native-like epitope and, thereby, guide the protein along the proper folding pathway. If the antibody could retain or recover its binding to the antigen at reasonably high denaturant concentrations, immobilised antibodies would be a useful tool to assist in the refolding.

52

53

PRESENT INVESTIGATION The main objective of this thesis has been to develop methods to facilitate product recovery of target proteins after recombinant production in Escherichia coli. Previously, a purification tag with a highly positively charged surface has been constructed by protein design. A compact three-helix bundle domain, Z, from the bacterial receptor protein A was used as a scaffold and the resulting handle was denoted Zbasic (Graslund et al., 2000). As this protein domain is further explored in several of the articles (I, II, IV, V, and VI) that this thesis is based on, the background of Zbasic has been discussed in section 3.2.3. Using a similar strategy as for the development of Zbasic, a purification tag with a highly negatively charged surface, denoted Zacid, was constructed and thoroughly characterised (III). Both these charged domains, Zbasic and Zacid, were investigated for general use as fusion partners to different target proteins (I, III). Moreover, the positively charged purification tag Zbasic was investigated for further use. Cation exchange chromatography in an expanded bed mode was utilised for isolation of Zbasic fusion proteins from unclarified Escherichia coli homogenates (II). Integrated strategies for purification, site-specific cleavage and subsequent removal of the fusion tag were developed, both for cleavage in solution (II) and by the use of an immobilised protease (VI). Since the formation of inclusion bodies is a major problem for recombinant protein production in Escherichia coli, a method for in vivo detection of soluble and non-soluble protein product levels was developed (IV). The use of Zbasic for purification of inclusion body proteins under denaturing conditions was also investigated and a process for solid-phase refolding was developed (V).

54

5. IMPROVED PROTEIN RECOVERY BY THE USE OF CHARGED PURIFICATION TAGS 5.1. THE DEVELOPMENT OF NEGATIVELY CHARGED

PURIFICATION TAGS (III) Gene fusion strategies, where the protein of interest is genetically fused to a purification tag, have been proven to be very useful to selectively purify proteins using a general scheme despite different biochemical characteristics. Earlier, a highly charged purification tag, Zbasic (Fig. 11B, C, F), has been constructed for selective IEXC purification (section 3.2.3). For this, a compact 58

amino acid (7 kDa) three-helix bundle protein, Z (Fig 11A, F), derived from the B domain of staphylococcal protein A (section 3.2.1.3) (Nilsson et al., 1987) was used as a scaffold. Using a similar strategy as for the development of Zbasic, corresponding purification tags with a negative surface charge were also constructed (Fig. 11D, E, F). Although glutamic acid residues in proteins have slightly higher pKa-values than aspartic acids (Forsyth et al., 2002), glutamic acid was chosen to be used since the longer carbon chain of glutamic acid is thought to make it more resistance to hydrolysis (Creighton, 1993). Moreover, glutamic acid is found to have higher propensity to form alpha helixes than aspartic acid (O'Neil et al., 1990) and are therefore less likely disrupt the scaffold. To rationally build a negative surface, the residues previously identified as suitable for mutation (section 3.2.3) were closely examined. Two of them are basic (R27 and K35) and were thus the first to be exchanged. Both of these residues are situated in helix two and it was therefore decided that two more residues (F13 and LI7) in the other helix should be selected as targets for mutagenesis for the first negatively charge tag. This gave rice to a variant, denoted Zacid1, with six glutamic acids on the former Fc binding surface, since two of the 13 residues naturally were glutamic acids (Fig. 11D, F). A second variant was also constructed, in which all the previously mentioned 11 residues were designed to be glutamic acids (Fig. 11E, F). Restoring of the structural important E25 residue was not a problem this time, since this residue already was negatively charged. The constructed charged variants, Zacid1 and Zacid2, as well as their parental scaffold structure Z and the basic variants Zbasic1 and Zbasic2 (section 3.2.3), could be expressed in high amounts using E. coli as host organism. Expression was performed in fusion to a C-terminal hexahistidine tag to allow for IMAC purification, if purification by their intended means should have been proven difficult.

55

Fig. 11 Electrostatic models and amino acid alignments of the Z variants Models of the electrostatic potentials of the Z variants; A: Zwt. B: Zbasic1. C: Zbasic2. D: Zacid1. E: Zacid2. Amino acid alignments of the Z variants are shown in the middle (F).

56

In order to characterize the secondary structure of the different Z variants, their absorption of far-ultraviolet circular polarized light was measured between 190 and 250 nm (Fig. 12). The circular dichroism (CD) spectrum collected for the Zwt scaffold revealed a molecule with an alpha-helical structure (Fig. 12), which also is in accordance with the determined NMR-stucture of the Z domain (Tashiro et al., 1997). For comparison, the CD data for the basic Z variants (Graslund et al., 2000) also are presented in figure 12. These plots had shown a similarity of the spectra that implies they have secondary structure content comparable to the parental molecule, despite the number of positively charge amino acids introduced, although with a somewhat lower alpha helicity (Graslund et al., 2000). However, the CD spectra that were collected for the acidic variants differed and showed far more unstructured domains, resembling a random coil-like conformation (Fig 12). Hence, these data suggest that multiple negative charges on the surface of the Z molecule are more deleterious to the structure than positive charges. The buffers used in the experiment were of low conductivity. A possible explanation for the difference in structure could be that the negative charges repel each other and destabilize the structure. In an attempt to shield the charges and thereby stabilize the structure, increasing amounts of KCl were added to the acidic variants prior to the collection of new CD spectra (Fig. 12). Interestingly, the alpha helicity of the Zacid variants could be partly recreated by the addition of salt to the buffer. In 3 M KCl or more, the spectrum of Zacid1 was similar to that obtained for the parental Z scaffold. The structure of the most acidic variant, Zacid2, was also somewhat stabilized, although only to a lower extent. Even at 4.5 M KCl, which is close to the solubility maximum of KCl in water, the structure did not adopt the same degree of alpha helicity as the parental Z domain. As can also be seen in figure 12, the signal below 205 nm is disturbed by the high amounts of KCl. Due to the destabilised secondary structure of the acidic purification tags, high levels of proteolysis could be expected. Therefore, proteolytic analyses of the different Zacid protein products were performed. Interestingly, no degradation could be detected.

Figure 12. Circular dichroism spectra of the Z variants between 190 and 250 nm Overlay plot of CD spectra for Zwt, Zbasic1, Zbasic2, Zacid1 and Zacid2. Additional spectra for Zacid1 and Zacid2 collected after addition of KCl. All Z variants are produced with a C-terminal His6 tag and a Trp leader sequence consisting of 18 amino acids.

57

5.2. ION EXCHANGE ANALYSIS OF Z VARIANTS (III) In order to evaluate the use of the newly constructed purification tags, their adsorption characteristics on ion exchange resins were investigated. Purified Z variants were loaded onto ion exchange columns at different pH conditions and subsequent elution was performed using a NaCl gradient (Fig. 13). For comparison, a chromatogram showing the elution profiles of Zbasic1 and Zbasic2 (Graslund et al., 2000) are shown in figure 13A. At low pH values the basic variants were eluted from a cation exchanger at much higher conductivity than the wild-type variant (Graslund et al., 2000). At elevated pH values, only the engineered variants could efficiently be adsorbed on the chromatographic resin. Zbasic1 and Zbasic2 were readily adsorbed even at pH values in the 9 to 11 range respectively, even though the parental Z domain is not adsorbed at all at pH 9 (Graslund et al., 2000).

Figure 13. Ion exchange analysis of the Z variants A. Cation exchange chromatogram of Zwt, Zbasic1 and Zbasic2 at pH 9.

B. Anion exchange chromatogram of Zwt, Zacid1 and Zacid2 at pH 6.

Even though the negatively charged mutants are unstructured in low salt environment, they were tightly adsorbed to an anion exchange resin and were easily eluted by merely increasing the salt concentration. A comparison between the wild-type Z domain and the engineered variants actually demonstrated a dramatic difference in chromatographic performance. At pH 7 the acidic variants were eluted at much higher salt concentrations (34 mS/cm and 48 mS/cm respectively) than the wild-type variant (8 mS/cm). At pH 6 the Z domain was detected in the flow through, whereas both Zacid1 and Zacid2 were quantitatively adsorbed on the chromatographic resin (Fig 13B). Actually, Zacid2 was able to bind to an anion exchanger even at pH values down to 5. The high conductivity needed for elution of the engineered domains indicates that the proteins still are able to adopt an ordered structure that displays a well-defined charged surface when binding to the matrix. This hypothesis was evaluated by analyzing the adsorption capability of the charged domains in denaturing conditions. While subjected to 8 M Urea, the engineered domains were eluted below a conductivity of 10 mS/cm. Thus, the superior adsorption characteristics in native conditions seem to be structure dependent.

58

5.3. THE USE OF ZBASIC AND ZACID AS GENERAL PURIFICATION TAGS (I, III)

To evaluate the usefulness of the engineered domains as purification handles, it was investigated if their chromatographic properties were preserved also after fusion to a target protein.

5.3.1 Zbasic Among the basic variants, the most positively charged Zbasic2 (from now on denoted Zbasic) was chosen for further studies as it provides a stronger interaction to the cation exchange resin and still seems to have retained stability. As a model target protein, the Klenow fragment, an exonuclease deficient variant of E. coli DNA-polymerase I, was fused on the genetic level to the Zbasic moiety. The ZbasicKlenow fusion protein was successfully produced intracellulary in E. coli. Previous studies had revealed that the Zbasic domain was adsorbed to a cation exchanger at pH values up to 11. However, already at lower pH values the adsorption could be selective, since Zbasic was eluted at significantly higher conductivity than the majority of other proteins. Because Klenow, and a lot of other proteins as well, has maximally retained bioactivity at physiological pH, the purification process was chosen to be performed at pH 7.5. The Klenow fragment is a large protein (86 kDa) with a low pI of 5.65. Hence, the calculated pI value of the ZbasicKlenow fusion protein was determined to be as low as 5.8. Despite this, the fusion protein was tightly adsorbed to the ion exchange resin at pH values beyond 5.8. An explanation to this could be that the extremely charged moiety is able to confer a local positive charge concentration in the fusion protein. Probably the tag does not interfere with the target partner to any great extent. This observation is further supported by the fact that the same fusion protein also can be adsorbed to an anion exchanger at the same pH values and that the Klenow fragment retained its polymerase activity in the fusion protein. To further optimize the purification conditions, the influence of conductivity in the loading and washing steps was investigated by the addition of NaCl in the buffers and E. coli lysate. It was found that also in an environment with rather high conductivity, up to 20 mS/cm, an efficient adsorption of the protein to the cation exchange column was still accomplished. This allowed for effective washing to reduce the adsorption of endogenous E. coli proteins, without decreasing the purification yield. The ZbasicKlenow fusion protein was eluted at a relatively high NaCl concentration, corresponding to a conductivity of 40 mS/cm (Fig. 14). Next step was to evaluate the generality of Zbasic as purification tag and to investigate if the optimized process conditions could be universal. In this purpose, Zbasic was genetically fused to two other target proteins, the viral protease 3C (calculated pI=9.7) and the fungal lipase cutinase in fusion to a Z dimer (calculated pI=7.2). While fused to protease 3C, the Zbasic domain was positioned at the N-terminal and the resulting protein, Zbasic3C, was produced intracellulary. Together with ZZcutinase, the Zbasic domain was positioned in the C-terminus and the accumulation of the fusion protein was directed to the periplasmic space of the E. coli host. Both constructs were quantitatively recovered employing the previously optimized, stringent but gentle, parameters for cation exchange chromatography. These two fusion proteins were found to be eluted at about the same conductivity (~40 mS/cm). Interestingly, the purity obtained by this optimized single unit operation was analogous to what is normally achieved using affinity chromatography (Fig. 14). The activity of the target proteins were investigated and they all showed retained activity. This was ensured by the possibility to perform the entire cationexchange chromatographic purification procedure at a physiological pH of 7.5, in contrast to the extremes of pH often used for elution in affinity purification. Taken together these results show the power of using Zbasic as a purification tag for facilitated purification of a wide range of target proteins with different isoelectric points, by a general purification scheme.

59

Figure 14. SDS-PAGE analysis of Zbasic fusion proteins Samples collected at different stages during the purification of ZbasicKlenow, Zbasic3C and ZZCutinaseZbasic respectively. M: Low molecular weight marker. L: Loaded sample. FT: Flow through. E: Eluted sample.

5.3.2 Zacid In certain cases, cation exchange chromatography is not suitable and anion exchange chromatography is preferred. The purification tags engineered for selective anion exchange recovery showed low alpha helicity in low salt solution, but still adsorbed efficiently to an anion matrix and was quantitatively eluted by elevating the conductivity. Among these acidic variants, Zacid2 (from now on denoted Zacid) was chosen for further evaluation since it showed strongest adsorption to the AIEX matrix. In order to obtain a high selectivity but still use mild conditions, a pH value of 6 was chosen for the purification process. Gene fusions between Zacid and three different target proteins, Green Fluorescent Protein (GFP, calculated pI=5.59), Maltose Binding Protein (MBP, calculated pI=5.52), and Firefly Luciferase (Luc, calculated pI=6.0), were made. All the corresponding gene products could be efficiently purified from whole cell lysates at pH 6 as can be seen on the corresponding SDS-page analysis in figure 15.

Figure 15. SDS-PAGE analysis of Zacid fusion proteins Samples collected at different stages during the purification of ZacidMBP, ZacidGFP, ZacidLuc, ZacidZacidGFP and ZacidZacidLuc respectively. L: Loaded sample. E: Eluted sample.

60

ZacidLuc and ZacidMBP were eluted at about the same conductivity, 18.7 and 18.5 mS/cm respectively, although their pI differs by 0.4 units. This suggests that the N-terminal Zacid domain is mainly responsible for the adsorption. ZacidGFP however, has a pI only 0.1 lower than ZacidMBP, but was adsorbed remarkably stronger and eluted at a conductivity of 20.5 mS/cm. A likely explanation is that the compact structure of GFP is more shielded from the charged Zacid domain. The elution profiles of the Zacid fusion proteins led to the conclusion that a running buffer with a conductivity of 18 mS/cm could be used to omit binding of undesired proteins to the matrix, without loss of wanted material. Using these optimized parameters the Zacid fusion proteins were eluted with a purity of about 90% (Fig. 15). In order to reveal if dimerization of the Zacid domain would improve the adsorption to the matrix, genetic constructs with two Zacid domains followed by the target molecule were constructed. The same purification parameters as for the monomeric tag were used when purifying the dimeric tag constructs. As expected, these proteins were eluted with high purity at a remarkably higher conductivity, ZacidZacidGFP at 28 mS/cm and ZacidZacidLuc at 25.7 mS/cm (Fig. 15). As an anion exchanger captures other negatively charged biomolecules as well, the eluted material was analyzed by agarose gel electrophoresis to detect co-purified nucleic acids. The nucleic acids were eluted at a higher conductivity (above 40 mS/cm) and contamination of the target protein could therefore be avoided. The acidic purification tag, Zacid, both as a monomer and as a dimer, adsorb to an anion exchanger in a moderate way, strongly enough to be selective versus other proteins but still possible to elute separately from nucleic acids.

5.4. INTEGRATED PROCESS FOR SELECTIVE PROTEIN RECOVERY USING ZBASIC (II)

The promising results achieved with Zbasic encouraged the development of an efficient, integrated pilot plant scale purification strategy using this cation exchange purification tag. An optimal process demands the use of as few steps as possible, operates under physiological conditions and provides high amounts of a pure and native product. A strategy to reach these goals with the use of Zbasic, was outlined as depicted in figure 16. The Klenow fragment of DNA-polymerase I was again chosen as a model protein for the development of the process.

Figure16. Process setup of an integrated strategy for selective EBA-purification and site specific cleavage of the fusion protein After cultivation, purification of target protein and protease, both produced as Zbasic fusions, is accomplished using cation exchange chromatography in expanded bed mode. The purification tag is then proteolytically removed. Both the released Zbasic moiety as well as the fusion protease is then removed in a second cation exchange purification step. A highly pure target protein is collected in the flow through.

61

5.4.1 Cultivation In order to obtain a high production level, the expression was performed in a high density E. coli cell culture controlled by glycerol feeding in a fed-batch fermentor. Thorough investigations were performed in shake flaks, to set some important cultivation parameters. It was found that 0.6 mM IPTG was sufficient for full activation of the T7 promoter, in order to give maximal expression. This amount is less than what is normally used. Since IPTG is an expensive chemical the thereby obtained economical benefit gets noticeable for production in large-scale. Induction time was also investigated. It was shown that the percentage of target protein in the cells increased until three hours after induction and was maintained without any visible product degradation until at least six hours after induction. Hence, after 6 hours of induction, the cells were harvested and the cells were disrupted by passage through a high-pressure homogenizer.

5.4.2 Purification Removal of suspended solids after cell disruption is a cumbersome process, especially in large-scale production. The use of chromatographic purification in an expanded bed adsorption (EBA) format enables capture and purification of biomolecules directly from the unclarified feed-stock (Chase, 1994) (section 3.1.1.2). This is due to the expanded distance between the adsorbent particles, which offers unhindered passage of whole cells, cell debris and their suspended solids. The use of the positively charged Zbasic tag enabled the purification to be performed on a cation exchange resin. Ion exchangers are well suited for purification of unclarified samples, since they tolerate the harsh conditions needed for sanitization that is obligate in these cases. Among the two ion exchange options, a cation exchanger is considered a better choice for use in EBA implementations since the positive charge of this adsorbent confers less interactions with cell surfaces, contaminating DNA and endotoxins (Feuser et al., 1999, Garke et al., 2000) . The amount of contaminating nucleic acids is rather high in an unclarified homogenate and therefore 20 mM MgCl2 was added to the disintegrated cells to minimize the interaction between nucleic acids and the positively charged Zbasic moiety (Muise et al., 1985). Moreover, the conductivity of the disintegrated cell samples was increased to 20 mS/cm, to avoid contaminants to bind, as optimized for the purification in packed bed mode. This action has dual benefits, as high ionic concentrations generally also inhibit electrostatic interactions between the particles in the feed-stream and the chromatographic matrix, leading to a more stable expanded column. The raw feed stock was then loaded in upward flow and a stable expanded bed was accomplished, where the sample advanced in a plug-flow. Adsorption of Zbasic was well established, even at flow rates high enough to keep the bed in an expanded mode. After washing of the expanded bed, the resin was packed by reversing the flow. Elution from the packed bed with 1 M NaCl resulted in a release of the adsorbed protein. The single EBA chromatography step resulted in 1.8 g eluted protein of about 90% purity per litre crude fermentor broth.

5.4.3 Cleavage Some proteins do not retain activity when fused to a purification handle and thus requires subsequent tag removal. To make this process more general, a specific cleavage site, recognized by protease 3C, was incorporated between the target protein and the fusion tag. The required protease, 3C, was also produced as a fusion to Zbasic and could be efficiently produced and purified using the same conditions as for ZbasicKlenow. Zbasic3C protease was active at pH 7.5, which allowed for a streamlined process where eluted material from the EBA-column could be directly cleaved by the fusion protease without the need for a subsequent buffer exchange step.

62

The eluted fractions from the EBA operation that contained ZbasicKlenow were simply diluted to a conductivity of 20 mS/cm and cleaved directly by adding Zbasic3C in a 50:1 molar ratio. The cleavage mixture was supplemented with 5 mM β-mercaptoethanol, as required for protease 3C activity. The cleavage reaction was completed after 90 minutes incubation at room temperature. The dilution of the eluted EBA fractions, originally containing 1 M NaCl, to a lower conductivity probably increased the activity of the protease. Furthermore, it allowed the fusion protease and the Zbasic tag to regain their strong adsorption ability to a cation exchanger. This provided the opportunity to subsequently capture the protease and the released tag simultaneously on a second cation exchanger. The target protein, Klenow, on the other hand, does not bind to the cation exchanger by itself, and was obtained in the flow through. Moreover, since the target protein this time was recovered in the flow through, the second cation exchange chromatographic step was able to remove other contaminating molecules that were co-purified in the first cation exchange chromatography step. The total integrated procedure allowed for collection of 1.4 g of an essentially pure protein (99%) per litre crude fermentor broth. In comparison with a commercially available Klenow polymerase, the purified Klenow protein showed similair activity in analyzes measuring the incorporation of dTTP in a poly-dA. The entire process, from inoculation of the fermentor to the recovery of the site-specifically cleaved fusion protein, was performed within 40 hours.

5.5. IMMOBILISED PROTEASE FOR CLEAVAGE OF FUSION PROTEINS (VI)

During the development of the integrated purification process it became apparent that tag removal could be conveniently accomplished using a Zbasic tagged protease 3C. The Zbasic tag facilitated manufacturing of the protease and also its removal after completed cleavage. However, using this strategy a new batch of protease has to be produced each time a substrate is to be cleaved. In order to be able to reuse the protease, it was immobilised onto a solid support in a column. A process-setup as illustrated in figure 17A was used. Zbasic tagged fusion proteins, with a site specific cleavage site in between the domains, were circulated in the protease column and complete cleavage was obtained within one hour (Fig. 17B). The protease column was repeatedly used without declining activity. Actually, the immobilised protease was stable for at least two months. Protease provided from the same production batch but stored in solution lost usefulness due to precipitation after only a couple of days. This is in accordance with other studies, where immobilised enzymes have been characterized by higher stability than their soluble form (Suh et al., 2003). A possible explanation is that while immobilised onto a solid support, the protease molecules are separated from each other which prevents destructive intermolecular interactions. This reduces the risk of aggregation (Katchalski-Katzir, 1993, Calleri et al., 2004) and allows thus a higher concentration of the protease. The high amount of protease that was immobilised in the column (10 mg/ml) provided that one hour was sufficient for complete cleavage of 30 nmol fusion protein. Moreover, the use of a protease that has retained activity at lower temperatures allowed the reaction to be performed at 4°C, which increases the stability of both the target protein and the protease column. The localisation of the cleavage reaction into a column facilitates process integration. In a protease column the substrate can be recycled until sufficiently cleaved. By simply directing the flow from recirculation to passage the reaction can easily be stopped at any time point. If the target protein is sensitive for reducing environments, the protease column offers the possibility to reduce the protease on the column prior to application of the substrate. After completed reaction, fast removal of the target is easily accomplished.

63

Apart from offering convenient manufacturing, the production of the protease fused to a tag can also be advantageous for subsequent immobilisation of the protease. Having the Zbasic moiety fused to the N-terminus probably direct the covalent linkage away from the protease. Moreover, it can serve as a spacer that permits the protease to rotate more freely and to reach the substrate molecules that passes through the column. With the protease covalently attached to a solid support it is successfully restricted from contamination of the final product. However, after completed cleavage the released tag still has to be separated from the target product. This is preferably done by capture on a cation exchanger in a similar way as described in the integrated purification process. However, using this strategy, the proteins to be separated are of course limited to mainly two different species, the target protein and the Zbasic tag. Since the Zbasic moiety of approximately 7 kDa is smaller than most other proteins, the separation can also be performed according to size by for example SEC, in cases where ion exchange chromatography is not suitable.

Figure 17. Cleavage of fusion proteins using an immobilised protease A. The strategy for cleavage of fusion proteins is outlined. Zbasic-tagged fusion proteins including a proteolytic cleavage sequence (1) are circulated (2) in a column with immobilised protease Zbasic3C. After completed cleavage (3) the released Zbasic tag is captured by a cation exchanger while the highly pure target protein (4) is collected in the flow through. B. SDS-PAGE analysis of samples taken at different stages in the process of the cleavage of Zbasic-QG-Klenow. (1) start sample of noncleaved Zbasic-QG-Klenow; (2) partly cleaved Zbasic-QG-Klenow after 30 min recirculation; (3) completely cleaved Zbasic-QG-Klenow after 1h recirculation; (4) Klenow after removal of released Zbasic.

64

6. NEW METHODS FOR ASSESSMENT, PURIFICATION AND REFOLDING OF NON-SOLUBLE PROTEINS 6.1. A FLOW CYTOMETRY BASED METHOD FOR ANALYSIS OF

EXPRESSION LEVELS (IV) When developing a process for the production of recombinant proteins it is of interest to determine the expression levels. Several efforts can be done to optimise the conditions and thereby improve the expression pattern. To assess the influences of such actions it is of importance to get direct feedback about the amount target protein expressed in the soluble fraction as well as destined to inclusion bodies. In the herein developed screening method, the test proteins were fused to the N-terminus of an optimised mutant of GFP, denoted enhanced Green Fluorescent Protein (eGFP) (Cormack et al., 1996). This mutant fluoresces more intensely than the wild type when excited at 488 nm and is thus optimal for flow cytometry analysis. The inherent fluorescence of GFP provides a convenient assay to explore intracellular concentration of soluble fusion proteins. Due to its structure dependence, GFP is fluorescent only when expressed in fusion to a soluble partner. If the counterpart aggregates, the following GFP module will co-precipitate and is then unable to fluoresce. Using samples directly from the cultivation, the whole cell fluorescence from eGFP fused to the target protein was measured by flow cytometry. To analyse the correlation between fluorescence and amount soluble protein, the cells were subsequently disrupted, physically separated into soluble and non-soluble fractions and analysed by SDS-PAGE. Afterwards, the median fluorescence of the cells was plotted versus the relative amount soluble target protein. A clear linear correlation (R2=0.96) could be seen between these two parameters; whole cell fluorescence and the amount expressed soluble protein (Fig. 18).

Figure 18. The correlation between the measured whole cell fluorescence and the amount expressed soluble protein The linear regression between whole cell fluorescence and expressed soluble protein for 60 cell samples (30 proteins produced at 25°C and 30°C), as determined from SDS-PAGE. R2= 0.96.

65

There has been some conflicting data in the literature about the reliability of GFP as a folding reporter. Tsumoto et al. claimed that insoluble fluorescent particles can contribute to the whole cell fluorescence (Tsumoto et al., 2003). Therefore, in order to further prove that the fluorescence originates from the soluble part of the protein, the cells were analysed by confocal microscopy (Fig. 19).

Figure 19. Confocal micrographs of cells producing eGFP fusion proteins.

A. The GFP fluorescence of a soluble protein, green areas indicate soluble eGFP fusion protein. B. The DIC signal of the same cell as in (A). C. The GFP fluorescence of a non-soluble protein. D. The DIC signal of the same cell as in (C).

This revealed that intracellular areas with increased density showed no fluorescence while the surrounding areas were green (Fig. 19). A further proof was found after physical separation of the protein content in the cell, since the soluble fraction was green while the non-soluble fraction was not. For post-production purposes, e.g. in purification processes and subsequent storage, the in vitro solubility of a protein is also interesting. However, this might not necessarily be the same as in vivo. To elucidate if the solubility data achieved from the flow cytometric analyses correlate with data achieved when analysing the same proteins in vitro, the His6 tagged proteins, without eGFP, were purified. A simple dilution assay was used to analyse the in vitro solubility of the purified proteins (Stenvall et al., 2005). Interestingly, a correlation between fluorescence within the cell and solubility data in vitro could be detected, although some proteins behaved slightly different. The differences indicate that in vivo folding of recombinant proteins in E. coli is somewhat translational dependent. However, the differences might also be due to that the eGFP reporter portion of the fusion protein measured in vivo alter the solubility of the target protein. Although whole cell fluorescence seems to reliably probe the intracellular concentration of soluble protein, it became apparent from the SDS-PAGE analysis that the fluorescence is independent of the amount non-soluble protein expressed. Bright clones producing the same amount of soluble protein contained varying amount of insoluble protein. Some cells with a large amount of soluble target protein also had a large amount of non-soluble target protein while other showed almost no precipitated protein. Yet other cells showed low fluorescence since they lacked soluble target protein, but still contained a high amount of non-soluble target protein. Since the aim of this project was to develop an integrated system, to analyse inclusion body forming target protein within the cell as well, other available data from the flow cytometric analyses were also investigated. Apart from fluorescence, a flow cytometer simultaneously measures light scattering, which gives indications about physical cell characteristics like internal structure, shape and relative size. Light scattering data of the different cell samples showed great variance in distribution within the cell population. Some of the cell samples were a tight population with low forward scattered light (FSC) and mostly low side scattered light (SSC), resembling cells

66

producing eGFP without target protein. Other cell samples had populations that were more differentiated, including a high fraction of cells with high FSC values. Untransformed E. coli cells generate a unimodal distribution with respect to FSC and a bimodal distribution with respect to SSC (Lewis et al., 2004). Accordingly, all cell samples showed a rather diffuse cluster shape with respect to SSC, while the cluster with respect to FSC ranged from narrow to broad. Interestingly, those cell samples that showed a broad distribution of the cells regarding FSC also had a larger fraction of aggregated proteins as determined by SDS-PAGE. When plotting the standard deviation (SD) in FSC versus the relative amount aggregated target protein a rather linear correlation (R2=0.85) was found (Fig. 20).

Figure 20. The correlation between the spread in forward scattered light and the amount expressed soluble protein The linear regression between S.D. in FSC and expressed non-soluble protein for 60 cell samples (30 proteins produced at 25°C and 30°C), as determined from SDS-PAGE. R2= 0.85.

To track an explanation for this finding, some further analysis were performed. Microscopic analysis of the cells revealed that the fusion proteins aggregated in distinctive inclusion bodies that not were homogeneously distributed throughout the cell but seemed to accumulate at the end of a distended cell. This affects internal structure, shape as well as relative size of the cell and explains the increase in light scattering (Shapiro, 1995). After flow cytometry mediated isolation of cells with high FSC, it was revealed that those cells had a significantly decreased viability. This is probably due to that those cells that have densely packed inclusion bodies are less disposed to divide. The microscopic analysis also showed that one or a couple of inclusion bodies of different sizes could be present within one single cell. In a living cell culture the cells are dividing constantly, giving a mixture of cells with varying characteristics such as amount of expressed protein and consequently varying sizes of inclusion bodies. Cell populations that produce proteins with high tendency to aggregate consist of cells in the whole range from newly divided to old cells containing large inclusion bodies. This is probable the reason for an increased spread of FSC values in cell populations that express an aggregation prone protein. Accordingly, forward scattered light can visualise inclusion body formation. However, the amount of soluble protein did not seem to affect the light scattering. To be able to visualise the complete expression pattern, both measurements of GFP fluorescence and light scattering have to be done. The newly developed flow cytometry based method has the dual benefit of revealing the amount of expressed soluble protein and indicating if there also is non-soluble protein expressed as well. Moreover, the analysis is performed on living cells directly from the cultivation and provides knowledge about the expression pattern right away. This information gives the opportunity to evaluate and perform process optimisation in real time. In addition, the rapid analysis enables fast screening of solubility using many different promoters as well as protein constructs, which may give valuable information about the rules of protein expression and solubility. The developed method can also be used as a high throughput library-screening tool, since fluorescence activated cell sorting provides a powerful means to quantitatively identify and isolate cells on the basis of well-defined multi-parameter properties; fluorescence and light scattering.

67

6.2. THE IMPACT OF ZBASIC ON PROTEIN SOLUBILITY (IV) A widely ongoing topic of discussion is the effect of various tags on target protein solubility. One commonly used strategy to increase solubility, with varying success, is to express the target protein with a protein that has high solubility e.g. thioredoxin and MBP (Sachdev et al., 1998). Also, Z, the parental domain to Zbasic, is highly soluble and has been shown to increase the solubility of fusion proteins that otherwise are insoluble (Samuelsson et al., 1996). In order to analyse the effect Zbasic has on the solubility and inclusion body formation, the newly developed screening method was used. As test proteins, a set of 20 different protein fragments were chosen. For comparison, the same target proteins were expressed and analysed using the newly developed method, while fused to the His6 tag as well. The expression pattern of ABPeGFP produced as either Zbasic or His6 fusions are shown in figure 21. Moreover, the pattern is reflected in the confocal micrographs of cells from the same samples (Fig. 22).

Figure 21. The expression pattern of Zbasic and His6 fusion proteins SDS-PAGE analysis of the separated soluble (S) and non-soluble fractions (P) of ABPeGFP produced at 25°C and 30°C in fusion with Zbasic and His6 respectively.

Figure 22. Confocal micrographs of cells producing ABPeGFP in fusion with Zbasic and His6 respectively A. The GFP fluorescence of ZbasicABPeGFP. B. The DIC signal of the same cell as in (A). C. The GFP fluorescence of His6ABPeGFP. D. The DIC signal of the same cell as in (C).

68

Due to its specificity and small size (6 aa), the His6 tag is a commonly used purification tag. However, it has been found to not be optimal for production of soluble proteins (Pryor et al., 1997, Hammarstrom et al., 2002, Routzahn et al., 2002, Kohli et al., 2003, Woestenenk et al., 2004). The vectors used were equivalent except from the N-terminal tag, which starts on the fifth codon. All constructs were expressed at two different temperatures; 25°C and 30°C. Whole cell fluorescence and the spread in forward light scattering of cells was measured to probe soluble and non-soluble protein levels (Fig. 23).

Figure 23. Comparison of expression levels when fused to Zbasic or His6 respectively A. The whole cell fluorescence of cells producing the same target when fused to Zbasic (black) and His6 (grey). B. The spread in FSC of cells producing the same target when fused to Zbasic (black) and His6 (grey).

69

The amount soluble target protein was clearly affected by the purification tag chosen, as the fluorescence differs significantly for most protein fragments when expressed with different tags. The different cultivation temperatures were also reflected in the fluorescence data. Highest amount of soluble target protein was obtained when fused to Zbasic in 15 of 21 cases. The remaining target proteins (6) gave more soluble product when fused to His6, if produced at 25°C. The temperature effect was most evident for the His6 tagged proteins. Among the His6 tagged proteins, all except one gave more soluble protein at 25°C compared to 30°C. When the target protein was fused to the Zbasic moiety on the other hand, 6 of 21 constructs gave more soluble target protein at 30°C compared to 25°C. Thus, for production at 30°C, Zbasic is clearly the better choice since the amount of soluble protein was higher when fused to Zbasic instead of His6 in 18 of 21 cases. At 25°C this advantage was less dominant, as the amount of soluble protein then was higher in marginally 11 of 21 cases when fused to Zbasic instead of His6. However, if the intention is to produce high amounts of protein, no matter of solubility, the His6 tag must be considered a better choice. Most His6 tagged protein where produced in large amounts although mostly as non-soluble protein. While giving rise to the same amount, or more, of soluble protein, the Zbasic tagged variant formed none, or only little, inclusion bodies. The His6 tagged proteins are even more prone to form inclusion bodies when produced at 30°C. However, there is only a little increase in inclusion body formation for the Zbasic tagged proteins at elevated temperatures. The results from this study imply that Zbasic has a more solubilising effect than the His6 tag on its fusion partner. However, the mechanism behind this phenomenon is unclear. The observed effect might be purely due to the solubility of the fusion partner. It might also be affected by the translation rate. The distribution between soluble and non-soluble protein depends, apart from folding of the amino acid sequences, also on translation rate and total protein concentration. Consequently, it is affected both by differences in the N-terminal sequence and growth temperature, as also has been seen in this study. In most cases the His6 tag gave higher total protein levels, indicating a higher translation rate that may be due to the smaller size of the His6 tag. Secondary structures of RNA in the region around the start codon can interfere with ribosome binding and is thus also important for translation initiation efficiency (Dyson et al., 2004). The number of CA repeats has been suggested to increase the expression level (Martin-Farmer et al., 1999). This might be a factor contributing to increased protein concentrations, since the codons used for the His6 tag were six repetitive CAT. The high translation rate of His6 tagged proteins might thereby exceed the capacity of the cell to fold the newly synthesized protein. A lower translation rate which probably is obtained while using the Zbasic tag on the other hand, allows critical folding events to occur and contribute to the higher yield of soluble protein, despite an aggregation prone peptide. Accordingly with the herein obtained results, this effect is more pronounced at higher temperatures where the overall transcription and translation rates are higher. Another hypothesis suggests that a highly soluble fusion tag actually decreases aggregation of the downstream polypeptide (Georgiou et al., 1996). During translation, the expressed polypeptide that emerges from the ribosome first is thought to be the one that determines the folding pathway (Sachdev et al., 1998). This is concurrent with findings that protein domains prone to aggregation can arrest folding of the following module, as exemplified by GFP. On the contrary, when the easily folded Zbasic tag forms a stable helical configuration and nucleate to a folded domain, it may accelerate or stabilise the folding of the rest of the molecule.

70

6.3. PURIFICATION UNDER DENATURING CONDITIONS (V) The robustness of CIEX resins and the gentle conditions that were found to be suitable for Zbasic mediated purification under native conditions, encouraged an investigation of the possibility to use Zbasic for purification of inclusion body proteins. First the capability of the unfused Zbasic moiety to bind the cation exchange matrix under denaturing conditions was investigated. The non-ionic denaturant urea was used in order to minimize disturbance of the electrostatic interactions needed for successful IEXC purification. In 8 M urea at pH 7.5, the Zbasic moiety still was completely adsorbed to a cation exchanger. However, the interaction was less strong, as Zbasic was eluted at a conductivity of 17 mS/cm, compared to 88 mS/cm for the natively folded domain (Fig. 24). Interestingly, the Zbasic moiety could rapidly regain its binding capability, while captured on the column, if the urea concentration was slightly decreased. In 6 M of urea the Zbasic was absorbed up to a conductivity of 27 mS/cm and in 4 M urea it remained bound up to 45 mS/cm. This capability would be enough to provide selective purification.

Figure 24. Elution profiles of Zbasic at different urea concentrations Purified Zbasic was loaded in 8 M urea onto a cation exchanger. Prior to elution the urea concentration was gradually reduced to the labelled concentration respectively. In order to evaluate the general use of Zbasic, it was investigated if its selective capability was preserved also after fusion to a target protein. Three different target proteins of varying size and composition, protease 3C, Albumin binding Protein (ABP) and ABP fused to Green Fluorescent Protein (ABPeGFP), were produced in fusion to the Zbasic tag. Cells that had over-expressed the different fusion proteins were solubilised in 8 M urea directly after harvest, without prior removal of endogenous proteins. This process is cost-effective but renders a lot of contaminating proteins, which provides a challenge for the following purification process. All three Zbasic fusion proteins were successfully adsorbed to the cation exchanger although denatured in 8 M urea. To offer variability, two different setups were developed for the purification process, as depicted in figure 25. In the first setup (Fig. 25A), the Zbasic fusion protein was loaded onto a column directly in the lysis buffer containing 8 M urea. The column had been equilibrated with a buffer with the slightly lower urea concentration of 6 M and this buffer was used in the rest of the process. After the loading step unbound material was washed out and a pre-elution washing step containing 120 mM NaCl was incorporated to elute weakly bound material. Afterwards, a second elution step with 160 mM NaCl was implemented to elute the target protein in a narrow peak (Fig. 25A). This strategy revealed an essentially pure product for all three constructs (Fig. 25C, 6M). The adsorption to the matrix was disrupted at the same conductivity in all cases, which implies that the interaction was mainly due to the charges on the Zbasic moiety. In such denaturing conditions as provided by 6 M urea, almost no other proteins were adsorbed to the CIEX resin, which makes Zbasic very selective.

71

In most cases solubilisation in 8 M urea leads to total denaturation of the protein. If the protein has substantial amount of native conformation in the inclusion bodies, it is possible to choose a weaker solubilising buffer that does not disrupt the already formed structure. Accordingly, a second setup (Fig. 25B), where the Zbasic fusion protein is purified in a less denaturing environment, was developed. This time the samples were loaded on to a column in 4 M urea and this buffer was used throughout the process. Using this setup, weakly bound proteins could be eluted more stringently, with a stepwise raise of the salt content to 200 mM. When the NaCl content was further raised to 300 mM, Zbasic tagged fusion proteins were eluted in a narrow peak (Fig. 25B). Also this setup could be used for all three different target proteins without need for further optimisation (Fig. 25C, 4M). However, one protein, Zbasic3C, was less pure when purified in 4 M urea compared to 6 M urea, since proteins of smaller size were co-purified (Fig. 25C). These protein bands were not seen when using the first setup. A possible explanation is that some E. coli proteins were able to bind to the target protein in a structural dependent way, and were therefore not co-purified in 6 M urea. Alternatively, the contaminants are adsorbed due to unsuccessful folding provoked by the rapid change from 8 to 4 M urea prior to loading. Less pure target proteins are also experienced when purifying His6 tagged proteins on IMAC matrices under less denaturing conditions.

Figure 25. Purification of Zbasic fusion proteins under denaturing conditions A. The chromatogram showing purification of ZbasicABP according the first process setup using 6 M urea. B. Chromatogram of the purification of ZbasicABP according to the second process setup using 4 M urea. C. SDS-PAGE analysis of loaded sample (L) and eluted peaks (4 M and 6 M) from the purification of Zbasic fusion proteins.

72

Since His6 is the most commonly used tag for purification of inclusion bodies, the newly developed strategy was compared with the His6 tag strategy, to explore if it offers a competitive alternative. The same target proteins were thus also produced as His6 tagged fusion proteins. The subsequent purification was performed using a standard protocol with pH mediated elution (Fig. 26).

Figure 26. Purification of His6 fusion proteins under denaturing conditions A. Chromatogram of the purification of His6ABP. B. SDS-PAGE analysis of loaded sample (L) and eluted peaks (6 M) from the purification of His6 fusion proteins. As concluded in the previous study (IV), the expression of target protein is generally higher for His6 tagged proteins (Fig. 26B) than for the same target fused to Zbasic (Fig. 25B). However, the relatively small protein, ABP, was found to be an exception, since it is produced at equal levels despite fusion tag. The higher production level obtained with the His6 tag in the other two cases provided a low relative contamination background. Despite the significantly higher contamination level of the loaded sample when using Zbasic, the final product was equally pure (Fig. 25B, 26B). In conclusion, both the Zbasic mediated purification strategies generated pure products for all three target proteins without need for optimisation. Although partly unstructured in denaturing conditions, the ability of Zbasic to bind the cation exchanger seems not to be influenced by the target protein. The obtained results strongly imply that Zbasic can be generally used as a purification tag, even under denaturing conditions. Moreover, Zbasic offers a promising alternative to the His6 tag for purification of inclusion body proteins.

73

6.4. SOLID PHASE REFOLDING OF ZBASIC TAGGED FUSION PROTEINS (V)

For most purposes the native structure of a protein is obligate. Accordingly, purification of inclusion body proteins has to be followed by in vitro refolding. As previously described, denatured Zbasic tagged fusion proteins could be captured on a cation exchanger in a way that seemed to be mostly due to the charges on the Zbasic surface. Moreover, the adsorption capacity Zbasic increases if the urea concentration is decreased. The elution profiles of Zbasic at different urea concentrations are shown in figure 24. This implies that the Zbasic moiety can regain its highly charged surface area while captured on the column. Upon complete removal of denaturant, Zbasic regained the identical adsorption characteristics as the native protein. This reveals that the Zbasic moiety is able to completely refold while reversibly interacting with the matrix. The promising results encouraged a further investigation of the use of Zbasic as a spacer for solid phase refolding. With a physical separation of the target protein from the matrix, which can be provided by the Zbasic tag, promiscuous interactions leading to aggregation of the denatured target protein on the solid support are minimised. Using the solid phase refolding strategy, the restriction of the denatured protein onto a solid support provides spatial separation of the individual polypeptide chains that otherwise would aggregate with each other upon removal of the denaturant. While the protein is adsorbed on the column, the buffer composition can be varied regarding denaturant concentration, pH, reducing agents etc. to ensure optimal conditions for refolding. The rate of denaturant removal can easily be adjusted depending on the degree of solubility of the folding intermediates and the folding rate. However, on the other hand interactions with the matrix have to be considered while using this strategy. Except for the aimed ionic interaction between the polycationic Zbasic tag and the resin, electrostatic as well as hydrophobic interactions between the solid support and the rest of the protein may affect adsorption and folding. The electrostatic binding of the unfolded target protein to the resin is not expected to hinder folding to any great extent, since globular proteins generally have their ionic groups on the surface. However, in order to gain flexibility, the environment should allow only highly reversible electrostatic interactions. The hydrophobic interactions will be minimised with the use of an agarose based solid-support that has a very hydrophilic surface. However, since an elevated ionic strength promotes hydrophobic interactions, this effect should still be taken into consideration. A process for solid phase refolding after capture by the Zbasic tag was developed using enhanced Green Fluorescent Protein (eGFP) from the jellyfish Aequorea Victoria as a model target protein. The mature chromophore of GFP remains chemically intact in the denatured state, but loses fluorescence when not shielded from quenching (Waldo et al., 1999, Fukuda et al., 2000). However, the chromophore starts to emit fluorescent light again when the full tertiary structure of the protein forms around the chromophore, which signals correct fold. Purified eGFP was thus completely denatured in urea and heat until no fluorescence could be detected. Thereafter, the protein solution was loaded onto a cation exchanger. From the initial refolding attempts of the Zbasic moiety, it became apparent that a 20 minutes gradient for removal of denaturant during ensured sufficient time for Zbasic to adapt its native three dimensional structure. These data was used as guidelines when developing the refolding process. Recent results have shown that a gentle removal of denaturants often enhances the yield of refolded proteins (Li et al., 2002). While adsorbed on the resin, a gradient removal of denaturant is easily accomplished. Moreover, the conductivity of the environment is of importance for successful renaturation of a protein while adsorbed on a cation exchange column. At the initial

74

high denaturant concentration the Zbasic tag is largely unstructured and has only scattered ionic interactions with the matrix. The rate for folding is decreased by the denaturant (Li et al., 2002) and early folding intermediates are slowly formed as the denaturant concentration is lowered. Addition of salt at a high denaturant level (3M urea) led to that a large proportion of the protein was forced into a pathway leading to a wrong conformation. Hence, the first folding step seems to be obligated to a low conductivity environment. One of the first steps in folding is thought to be the formation of a buried hydrophobic core (Dill, 1990). Probably the elevated ionic strength disturbed the arrangement of internal hydrophobic residues. When the folding intermediates are formed, the protein has most of its charged residues exposed on the surface. If the conductivity was not raised at this point, a high degree of the protein retained aggregated on the matrix, probably due to irreversible electrostatic interactions. After these observations, it was concluded that the refolding process should be performed in a stepwise manner. The refolding program is outlined in figure 27.

Figure 27. An outline of the program used for solid phase refolding of ZbasiceGFP and Zbasic3C The denatured protein was loaded onto a cation exchanger previously equilibrated with 6 M urea and non-bound protein was washed away (LW). The urea concentration was gradually decreased using a linear gradient from 6 to 2 M during 20 min (D1). Then the conductivity was increased using a linear gradient from 0 to 200 mM NaCl during 30 min while the urea concentration was kept at 2 M (S). In the last step of the renaturation the urea concentration was decreased from 2 to 0 M, and simultaniously DTT was removed, using a linear gradient of 30 min (D2). Elution of the renatured protein was achieved using a linear gradient from 200 mM to 2 M NaCl during 20 min (E).

As eGFP contains reduced cysteines in its native state, a reducing agent was added in the first steps of the refolding process, in order to prevent the formation of unwanted disulphide bridges. After accomplished refolding, a reducing agent was no longer required and thus omitted from the buffer prior to elution. After capture of the denatured ZbasiceGFP on the solid support, the folding intermediates were allowed to form in a low conductivity environment during a decrease of denaturant concentration to 2 M urea (Fig 27, D1). This step seemed to be time dependent, as a too steep gradient resulted in an inactive product. Moreover, a steep gradient gave a product that easily was released from the cation exchanger (~25mS/cm), which implies a partly unstructured Zbasic moiety. In the next step, the conductivity was increased by a salt gradient up to 200 mM NaCl (Fig. 27, S). The rest of the denaturant was gradually removed in the elevated ionic strength that would prevent non-specific electrostatic interactions (Fig. 27, D2). Elution of the refolded protein was then performed with an appropriate salt gradient (Fig. 27, E) and the elution profile was comparable with that of the native protein.

75

For ZbasiceGFP, approximately 60% of the loaded amount of protein was obtained as fluorescent protein in this fraction, which is a very good renaturation yield. At a higher conductivity, a second peak containing non-active protein was eluted. This indicates that the process provides a way of separating different conformational states. During optimisation of the refolding process the proportion protein eluted in this second peak was decreased to only 10-15% of the loaded sample. The major loss of protein (30%) was due to intermediates that were irreversibly aggregated on the column and only could be removed after denaturant solubilisation and subsequent salt elution with 2 M NaCl. To determine the generality of the newly developed refolding process, the same conditions were used for refolding of another target protein, protease 3C. By overexpressed of Zbasic3C to provoke inclusion body formation, the system could be challenge with real life conditions. Inclusion bodies solubilised with denaturant were directly captured on a cation exchanger. In situ purification was achieved by washing out unbound protein before the refolding process was initiated. Contaminating proteins from the pellet of precipitated material did not bind to the cation exchanger under the denaturing conditions but was detected only in the flow through. Successful refolding was achieved this time as well, without further optimisation of the refolding program. Despite the very high protein concentration (5.3 mg/ml matrix), the refolding yield of Zbasic3C was even slightly higher (62%). The protease 3C activity was confirmed by cleavage of a fusion protein with a protease 3C cleavage site inserted between the two moieties. Approximately 34% of the loaded protein was eluted in a second peak at higher conductivity and this fraction showed no protease activity. On the other hand, no irreversibly aggregated protein on to the column was detected this time. Refolding on a solid matrix provides high flexibility in buffer composition and length of the different steps. Buffer exchange is facilitated while the protein is attached to the column which enables, in a very simple way, the imposition of an optimised environment for each step in the folding process. To minimise carry-over of denaturants, additives and reducing agents the final concentration of buffer components can be controlled before elution. The procedure has also the inherited advantage of combining purification and refolding in one step and still enabling elution of a concentrated product in a suitable buffer. Moreover, since Zbasic enables solid phase refolding on a cation exchanger at a physiological pH, the whole process can be performed at very mild conditions.

76

CONCLUDING REMARKS Recombinant expression of heterologous genes in bacteria is by far the most inexpensive means available for obtaining large amounts of many targeted polypeptides. In this thesis several methods that aim to facilitate product recovery of target proteins after recombinant production in E. coli have been presented. High throughput purification of recombinant proteins desires a general approach that uses a minimum of unit operations during the purification procedure but still yields a predictable result. The use of purification tags has become a major breakthrough since it contributes with a selective property to the target protein which generalises and facilitates the recovery process. However, most affinity fusion systems require harsh conditions to release the fusion protein from the affinity

matrix, which can affect the protein of interest. An attractive strategy is to fuse a charged moiety to the recombinant protein, thus allowing efficient adsorption on an ion exchange resin. Here, two protein domains engineered to have a highly charged surface, Zbasic and Zacid, were shown to be suitable for general use as fusion partners to different target proteins. It was shown that all fusion proteins investigated could be selectively captured by ion exchange chromatography at conditions excluding the majority of E. coli host proteins. A single cation exchange chromatographic operation at physiological pH was sufficient to yield Zbasic fusion proteins of purity close to homogeneity. Moreover, efficient isolation from unclarified E. coli homogenates could also be accomplished in a single step using cation exchange chromatography in an expanded bed mode. All Zacid fusion proteins investigated could be efficiently purified at pH 6 from whole cell lysates using anion exchange chromatography. With the use of either Zbasic or Zacid, selective purification could be performed using the preferred ion exchange chromatographic operation suitable for the specific target protein. Due to the advantages of being more structured and suitable for a process operating at physiological pH, Zbasic was investigated for further usage. The development of an integrated strategy for purification, efficient cleavage of fusion proteins and subsequent removal of the purification tag was considerably helped by the positively charged fusion tag. Zbasic was shown to facilitate manufacturing of both target protein and protease. Furthermore, the released tag was easily captured by adsorption on a cation exchanger, while the different target proteins were recovered in the flow through. Directed immobilisation of the Zbasic fusion protease on a solid support further improved the process. The obtained protease columns were shown to be reusable without any measurable decrease in activity. Efficient synthesis of polypeptides in E. coli can reach quantities up to 50% of the total protein content in the cell. However, in many cases, the recombinant polypeptides can not adapt their correct three-dimensional conformation and are found to be sequestered in inclusion bodies. A high throughput method for screening of protein expression levels was developed. By the use of a flow cytometer, levels of both soluble and precipitated protein could simultaneously be assessed in vivo. The presented approach was shown to be a convenient, highly sensitive and straight forward strategy to evaluate and optimise protein expression. Future usage of the developed method is promising. Using a high throughput instrument, this method provides information that allows protein production protocols to be optimised for growth conditions in real time. In addition, the method enables fast screening of solubility of many different proteins, and also the use of many different vectors, which may give valuable information about the rules of protein expression and solubility. Flow cytometry is also a high throughput library-screening tool, since

77

fluorescence activated cell sorting provides powerful means to quantitatively identify and isolate cells on the basis of well-defined multi-parameters properties; fluorescence and light scattering. The newly developed method was used to evaluate the effect that Zbasic has on the expression pattern. The obtained data revealed that Zbasic tagged proteins are less prone to form inclusion bodies than the same proteins fused to the His6 tag, especially when produced at 30°C. However, in cases when the target protein is produced as inclusion bodies, the property used for separation must be selective under denaturing conditions. The positively charged domain Zbasic was shown to be selective also for purification of inclusion body proteins solubilised in urea. Moreover, the ability of Zbasic to bind the cation exchanger was shown not to be influenced by the target protein even under denaturing conditions. Hence, Zbasic offers a promising alternative for purification of inclusion body proteins. If the activity and correct structure is needed, inclusion body proteins have to be refolded. Therefore, a new flexible process for solid phase refolding was developed, using Zbasic for reversible capture to a cation exchanger resin. This procedure was shown to have the inherited advantage of combining purification and refolding in one step and still yield a concentrated product in a suitable buffer. The newly developed method would be suitable for functional and structural proteomic projects since it provides a high throughput strategy for purification and refolding of proteins from inclusion bodies. Taken together, the positively charged Zbasic has been shown to be highly suitable as a purification tag. It combines the selectivity of affinity purification with the convenience and cost-effectiveness of ion exchange chromatography. Most Zbasic tagged proteins produced are highly soluble and the mild conditions used during the purification process increase the possibility to render active target proteins. However, for proteins produced as inclusion bodies, the opportunity of purification as well as solid phase refolding exists. The negatively charged purification tag, Zacid, has not yet been investigated for these purposes. However, the success of Zbasic implies a promising future for Zacid as well.

78

ABBREVIATIONS ABD Albumin Binding Domain ABP Albumin Binding Protein AC Affinity Chromatography AIEXC Anion Exchange Chromatography CD Circular Dichroism CIEXC Cation Exchange Chromatography DTT 1,4-dithiothreitol E. coli Escherichia coli EBA Expanded Bed Adsorption eGFP enhanced Green Fluorescent Protein FACS Fluorescent Activated Cell Sorter Fc constant region of immunoglobulin FSC forward scattered light GFP Green Fluorescent Protein GST Glutathione S-transferase HIC Hydrophobic Interaction Chromatography His6 hexahistidine tag HPLC High Performance Liquid Chromatography IEXC Ion Exchange Chromatography IgG Immunoglobulin IMAC Immobilised Metall ion Affinity Chromatography IPTG isopropyl-P-D-thiogalactopyranoside KD dissociation equilibrium constant Klenow exonuclease deficient variant of E. coli DNA-polymerase I Luc Luciferase MBP Maltose Binding Protein pI Isoelectric Point RPC Reverse Phase Chromatography SA Serum Albumin SD Standard Deviation SEC Size Exclusion Chromatography SPA Staphylococcal Protein A SPG Staphylococcal Protein G SSC side scattered light

79

ACKNOWLEDGMENTS ...och så vill jag hälsa alla nytillkomna läsare välkomna:=) Jag vill tacka alla som gjort det möjligt för mig att få avsluta den här perioden av mitt liv med ett bohemiskt författarliv i en träkoja på en öde ö. Lite arbete till därefter, så blev det en lååång bok i alla fall! Stort tack till alla nuvarande och gamla kollegor och vänner på DNA-corner. Förutom den otroliga expertpanel man har i er, så har julfester, DNAcorners, peköl, musikkvällar mm förgyllt tiden här. Ni är så många underbara människor (och jag e så trött nu) så jag kan inte nämna er alla här, men ni vet nog att ni är bäst! Speciellt vill jag i alla fall ge ett stort Tack! till: Sophia Hober för vetenskaplig handledning såväl som ledsagande i Florida. Dessutom, tack för att du är så förstående och har så skön livsinställning! Torbjörn Gräslund för ett roligt exjobb och att jag fick fortsätta Zbasic-spåret Mina fd X-jobbare: Tove, Rosa, Hye Ryung och Olle för massa bra labbande Jenny O för att du orkade läsa alla sidor och hjälpte mig att få flyt i the språket Mattias Uhlén för din smittande entusiasm Övriga PIs för att ni fått ihop en så trevlig avdelning MariaW för all koll på vad man ska göra för att få disputera Christin för hjälp med fermentorns alla delar, trots att du samtidigt skrev avhandling (hade ingen aning om vad det innebar då...) Hjalmar Brismar för att jag fick en ”extreme close-up” på mina celler Olof och John för körlektioner på FACSen. Nu har jag ett sånt körkort i alla fall! Cecilia Heyman för alla artiklar du fixat fram på nolltid Johanna för folding input Maria Stenvall för PrESTar och löslighetsdata Bahram för sekvensning av otal baser Adnane för hjälp att få ihop mina sista poäng, samt en övningsgrillning Gamla Proteinnissar: Martin, Jenny, Susanne, MalinE, MariaE, Henrik, Amelie, Sussie, Stisse m.fl för allt jag fått fråga Monica och Inger för att ni ordnar massa åt oss och fixar specialmackor till mig Fredrik, Rebecka och Emma för hjälp med all disk jag producerat Labbassarna: Lotta, MariaW, MariaE, Amelie, Martin, MalinA, MalinE, Torun, Tove, Mikaela, Björn, Ronny, Mats, Ason m.fl för att ni är så himla trevliga och hjälpsamma Proteinfabriken för utrustning och SOPar Sörgården för trevligaste labbet Luciakören: Susanne, MariaS, Lotta, Maria St, Hanna, Cilla, ToveO, Åsa, Caroline, Tobbe m.fl; för allt sjungande PerU för alla sånger du förgyllt med gitarrer, peruker mm Springkompanjonerna: Torun, Malin, Emilie, Mikaela för alla flåsturer, snart ska jag ut igen... ”Tjejgänget”: för trevliga middagar och fester, snart ska jag bjuda igen! Tack även till alla goa vänner från andra delar av livet (barndomen, kemi-tiden, Kårspexet mm) som har gjort livet glatt även när arbetsplatsen har varit ett kloningsträsk. Om jag ska hålla mig till det här arbetet så har en del av er även bidragit till bokens innehåll; Colin ”språkkonsulten” Ray, Jenny ”matrisexperten” Lundberg, Johan ”EBA-gurun” Englund m.fl.

80

Ett extra tack till familjen (Ulla, Åke, Erika och Mira) och ö-släkten (Niklas, Olof, Ann och Sten) för att ni alltid hjälper mig. Speciellt tack för all service (mat, fika, frukt, promenader, mm) under skrivartiden på ön. Magen har nog aldrig mått bättre! Tack! Jocke för att du är världens bäste på att ta hand om (natta, data, mata, bita, trösta mm) din apa!

81

REFERENCES Abbott, A. Protein structures hint at the shape of things to come. Nature (2005). 435, 547. Abrahmsen, L., T. Moks, et al. Secretion of heterologous gene products to the culture medium of Escherichia coli. Nucleic

Acids Res (1986). 14, 7487-500. Achari, A., S. P. Hale, et al. 1.67-A X-ray structure of the B2 immunoglobulin-binding domain of streptococcal protein G

and comparison to the NMR structure of the B1 domain. Biochemistry (1992). 31, 10449-57. Agashe, V. R., S. Guha, et al. Function of trigger factor and DnaK in multidomain protein folding: increase in yield at the

expense of folding speed. Cell (2004). 117, 199-209. Ahlborg, N., F. Sterky, et al. Predominance of H-2d- and H-2k-restricted T-cell epitopes in the highly repetitive

Plasmodium falciparum antigen Pf332. Mol Immunol (1997). 34, 379-89. Ahn, J. H., Y. P. Lee, et al. Investigation of refolding condition for Pseudomonas fluorescens lipase by response surface

methodology. J Biotechnol (1997). 54, 151-60. Alexander, P., S. Fahnestock, et al. Thermodynamic analysis of the folding of the streptococcal protein G IgG-binding

domains B1 and B2: why small proteins tend to have high denaturation temperatures. Biochemistry (1992). 31, 3597-603.

Altamirano, M. M., C. Garcia, et al. Oxidative refolding chromatography: folding of the scorpion toxin Cn5. Nat Biotechnol (1999). 17, 187-91.

Altamirano, M. M., R. Golbik, et al. Refolding chromatography with immobilized mini-chaperones. Proc Natl Acad Sci U S A (1997). 94, 3576-8.

Andrews, S. R., E. J. Taylor, et al. The use of forced protein evolution to investigate and improve stability of family 10 xylanases. The production of Ca2+-independent stable xylanases. J Biol Chem (2004). 279, 54369-79.

Anfinsen, C. B. Principles that govern the folding of protein chains. Science (1973). 181, 223-30. Asplund, M., Ramberg, M., Johansson, B. Development of a cleaning in place protocol and repetitive application of

Escherichia coli homogenate on STREAMLINE QXL. Process Biochemistry (2000). 35, 1111-1118. Babbitt, P. C., B. L. West, et al. Removal of a proteolytic activity associated with aggregates formed from expression of

creatine kinase in Escherichia coli leads to improved recovery of active enzyme. Biotechnology (N Y) (1990). 8, 945-9.

Baneyx, F. and M. Mujacic. Recombinant protein folding and misfolding in Escherichia coli. Nat Biotechnol (2004). 22, 1399-408.

Bardwell, J. C., K. McGovern, et al. Identification of a protein required for disulfide bond formation in vivo. Cell (1991). 67, 581-9.

Barker, D. F. and A. M. Campbell. The birA gene of Escherichia coli encodes a biotin holoenzyme synthetase. J Mol Biol (1981). 146, 451-67.

Berdichevsky, Y., R. Lamed, et al. Matrix-assisted refolding of single-chain Fv- cellulose binding domain fusion proteins. Protein Expr Purif (1999). 17, 249-59.

Bjorck, L., W. Kastern, et al. Streptococcal protein G, expressed by streptococci or by Escherichia coli, has separate binding sites for human albumin and IgG. Mol Immunol (1987). 24, 1113-22.

Blattner, F. R., G. Plunkett, 3rd, et al. The complete genome sequence of Escherichia coli K-12. Science (1997). 277, 1453-74.

Bollag, D. M. Gel-filtration chromatography. Methods Mol Biol (1994). 36, 1-9. Brake, A. J. Alpha-factor leader-directed secretion of heterologous proteins from yeast. Methods Enzymol (1990). 185,

408-21. Brizzard, B. L., R. G. Chubet, et al. Immunoaffinity purification of FLAG epitope-tagged bacterial alkaline phosphatase

using a novel monoclonal antibody and peptide elution. Biotechniques (1994). 16, 730-5. Bryson, J. W., S. F. Betz, et al. Protein design: a hierarchic approach. Science (1995). 270, 935-41. Bucher, M. H., A. G. Evdokimov, et al. Differential effects of short affinity tags on the crystallization of Pyrococcus

furiosus maltodextrin-binding protein. Acta Crystallogr D Biol Crystallogr (2002). 58, 392-7. Buchner, J., U. Brinkmann, et al. Renaturation of a single-chain immunotoxin facilitated by chaperones and protein

disulfide isomerase. Biotechnology (N Y) (1992). 10, 682-5. Buckholz, R. G. and M. A. Gleeson. Yeast systems for the commercial production of heterologous proteins. Biotechnology

(N Y) (1991). 9, 1067-72. Bukau, B. and A. L. Horwich. The Hsp70 and Hsp60 chaperone machines. Cell (1998). 92, 351-66. Burgess, R. R. Purification of overproduced Escherichia coli RNA polymerase sigma factors by solubilizing inclusion

bodies and refolding from Sarkosyl. Methods Enzymol (1996). 273, 145-9. Busso, D., R. Kim, et al. Using an Escherichia coli cell-free extract to screen for soluble expression of recombinant

proteins. J Struct Funct Genomics (2004). 5, 69-74.

82

Butt, T. R., S. C. Edavettal, et al. SUMO fusion technology for difficult-to-express proteins. Protein Expr Purif (2005). 43, 1-9.

Cabrita, L. D. and S. P. Bottomley. Protein expression and refolding - A practical guide to getting the most out of inclusion bodies. Biotechnol Annu Rev (2004). 10, 31-50.

Cahill, D. J. and E. Nordhoff. Protein arrays and their role in proteomics. Adv Biochem Eng Biotechnol (2003). 83, 177-87. Calleri, E., C. Temporini, et al. Development of a bioreactor based on trypsin immobilized on monolithic support for the

on-line digestion and identification of proteins. J Chromatogr A (2004). 1045, 99-109. Carbonell, L. F., M. J. Klowden, et al. Baculovirus-mediated expression of bacterial genes in dipteran and mammalian

cells. J Virol (1985). 56, 153-60. Carlson, J. D. and M. L. Yarmush. Antibody assisted protein refolding. Biotechnology (N Y) (1992). 10, 86-91. Carrio, M., N. Gonzalez-Montalban, et al. Amyloid-like properties of bacterial inclusion bodies. J Mol Biol (2005). 347,

1025-37. Carrio, M. M. and A. Villaverde. Protein aggregation as bacterial inclusion bodies is reversible. FEBS Lett (2001). 489,

29-33. Carter, P. Site-specific proteolysis of fusion proteins. Protein Purification: from molecular mechanism to large-scale

processes. (1990). M. R. Ladish, Willson, R.C., Painton, C.-D.C., Builder, S.E. American Chemical Society. Cedergren, L., R. Andersson, et al. Mutational analysis of the interaction between staphylococcal protein A and human

IgG1. Protein Eng (1993). 6, 441-8. Cereghino, J. L. and J. M. Cregg. Heterologous protein expression in the methylotrophic yeast Pichia pastoris. FEMS

Microbiol Rev (2000). 24, 45-66. Chaga, G., D. E. Bochkariov, et al. Natural poly-histidine affinity tag for purification of recombinant proteins on

cobalt(II)-carboxymethylaspartate crosslinked agarose. J Chromatogr A (1999). 864, 247-56. Chalfie, M., Y. Tu, et al. Green fluorescent protein as a marker for gene expression. Science (1994). 263, 802-5. Chalmers, J. J., E. Kim, et al. Effects of temperature on Escherichia coli overproducing beta-lactamase or human

epidermal growth factor. Appl Environ Microbiol (1990). 56, 104-11. Chao, H., D. L. Bautista, et al. Use of a heterodimeric coiled-coil system for biosensor application and affinity

purification. J Chromatogr B Biomed Sci Appl (1998). 715, 307-29. Chase, H. A. Purification of proteins by adsorption chromatography in expanded beds. Trends Biotechnol (1994). 12, 296-

303. Chase, H. A. and N. M. Draeger. Affinity purification of proteins using expanded beds. J Chromatogr (1992). 597, 129-

45. Chen, H., L. Pomeroy-Cloney, et al. The influence of adenine-rich motifs in the 3' portion of the ribosome binding site on

human IFN-gamma gene expression in Escherichia coli. J Mol Biol (1994). 240, 20-7. Chicz, R. M. and F. E. Regnier. Surface-mediated retention effects of subtilisin site-specific variants in cation-exchange

chromatography. J Chromatogr (1988). 443, 193-203. Chong, S., F. B. Mersha, et al. Single-column purification of free recombinant proteins using a self-cleavable affinity tag

derived from a protein splicing element. Gene (1997). 192, 271-81. Clark, E. D. Protein refolding for industrial processes. Curr Opin Biotechnol (2001). 12, 202-7. Cleland, J. L. Impact of Protein Folding on Biotechnology. (1993). J. Comstock. Cody, C. W., D. C. Prasher, et al. Chemical structure of the hexapeptide chromophore of the Aequorea green-fluorescent

protein. Biochemistry (1993). 32, 1212-8. Cohen, S. N., A. C. Chang, et al. Construction of biologically functional bacterial plasmids in vitro. Proc Natl Acad Sci U

S A (1973). 70, 3240-4. Colman, A. Production of therapeutic proteins in the milk of transgenic livestock. Biochem Soc Symp (1998). 63, 141-7. Conrad, U. and U. Fiedler. Compartment-specific accumulation of recombinant immunoglobulins in plant cells: an

essential tool for antibody production and immunomodulation of physiological functions and pathogen activity. Plant Mol Biol (1998). 38, 101-9.

Corchero, J. L., R. Cubarsi, et al. Limited in vivo proteolysis of aggregated proteins. Biochem Biophys Res Commun (1997). 237, 325-30.

Cormack, B. P., R. H. Valdivia, et al. FACS-optimized mutants of the green fluorescent protein (GFP). Gene (1996). 173, 33-8.

Cornvik, T., S. L. Dahlroth, et al. Colony filtration blot: a new screening method for soluble protein expression in Escherichia coli. Nat Methods (2005). 2, 507-9.

Crameri, A., E. A. Whitehorn, et al. Improved green fluorescent protein by molecular evolution using DNA shuffling. Nat Biotechnol (1996). 14, 315-9.

Creighton, T. Folding of proteins adsorbed reversibly to ion-exchange resins. Protein Structure Folding and Design. (1985). D. Oxender. Alan R Liss Inc: 249-251.

Creighton, T. The protein folding problem. Mechanism of protein folding. (1994). R. Pain. IRL Press. Creighton, T. E. Proteins: structure and molecular properties, (1993). W.H. Freeman and Company. Cronan, J. E., Jr. Biotination of proteins in vivo. A post-translational modification to label, purify, and study proteins. J

Biol Chem (1990). 265, 10327-33. Cubarsi, R., M. M. Carrio, et al. In situ proteolytic digestion of inclusion body polypeptides occurs as a cascade process.

Biochem Biophys Res Commun (2001). 282, 436-41.

83

Dalboge, H., H.-H. M. Dahl, et al. A novel Enzymatic method for production of autentic hGH from an Escherichia Coli produced hGH-precursor. Bio/Technology (1987). 5, 161-164.

Das, A. Overproduction of proteins in Escherichia coli: vectors, hosts, and strategies. Methods Enzymol (1990). 182, 93-112.

Davis, G. D., C. Elisee, et al. New fusion protein systems designed to give soluble expression in Escherichia coli. Biotechnol Bioeng (1999). 65, 382-8.

De Bernardez Clark, E., D. Hevehan, et al. Oxidative renaturation of hen egg-white lysozyme. Folding vs aggregation. Biotechnol Prog (1998). 14, 47-54.

De Marco, V., G. Stier, et al. The solubility and stability of recombinant proteins are increased by their fusion to NusA. Biochem Biophys Res Commun (2004). 322, 766-71.

de Vos, W. M., M. Kleerebezem, et al. Expression systems for industrial Gram-positive bacteria with low guanine and cytosine content. Curr Opin Biotechnol (1997). 8, 547-53.

Deisenhofer, J. Crystallographic refinement and atomic models of a human Fc fragment and its complex with fragment B of protein Afrom staphylococcus aureus at 2.9- and 2.8-Å resolution. Biochemistry (1981). 20, 2361-2370.

Delahunty, C. and J. R. Yates, 3rd. Protein identification using 2D-LC-MS/MS. Methods (2005). 35, 248-55. Derman, A. I., W. A. Prinz, et al. Mutations that allow disulfide bond formation in the cytoplasm of Escherichia coli.

Science (1993). 262, 1744-7. di Guan, C., P. Li, et al. Vectors that facilitate the expression and purification of foreign peptides in Escherichia coli by

fusion to maltose-binding protein. Gene (1988). 67, 21-30. Dill, K. A. Dominant forces in protein folding. Biochemistry (1990). 29, 7133-55. Dobson, C. M. Protein folding and its links with human disease. Biochem Soc Symp (2001). 1-26. Dobson, C. M. Protein folding and misfolding. Nature (2003). 426, 884-90. Dobson, C. M. Experimental investigation of protein folding and misfolding. Methods (2004). 34, 4-14. Dodd, I., S. Jalalpour, et al. Large scale, rapid purification of recombinant tissue-type plasminogen activator. FEBS Lett

(1986). 209, 13-7. Dong, X. Y., Y. Huang, et al. Refolding kinetics of denatured-reduced lysozyme in the presence of folding aids. J

Biotechnol (2004). 114, 135-42. Duplay, P., H. Bedouelle, et al. Sequences of the malE gene and of its product, the maltose-binding protein of Escherichia

coli K12. J Biol Chem (1984). 259, 10606-13. Duplay, P. and M. Hofnung. Two regions of mature periplasmic maltose-binding protein of Escherichia coli involved in

secretion. J Bacteriol (1988). 170, 4445-50. Dyson, M. R., S. P. Shadbolt, et al. Production of soluble mammalian proteins in Escherichia coli: identification of protein

features that correlate with successful expression. BMC Biotechnol (2004). 4, 32. Echelard, Y. Recombinant protein production in transgenic animals. Curr Opin Biotechnol (1996). 7, 536-40. Egmond, M. R., W. P. Antheunisse, et al. Engineering surface charges in a subtilisin: the effects on electrophoretic and

ion-exchange behaviour. Protein Eng (1994). 7, 793-800. el Rassi, Z., A. L. Lee, et al. Reversed-phase and hydrophobic interaction chromatography of peptides and proteins.

Bioprocess Technol (1990). 9, 447-94. Enfors, S. O., Häggström, L. Bioprocess Technology, Fundamentals and Applications, (1994). Royal Institute of

Technology. Enfors, S. O., L. van der Wielen, et al. Process integration challenges in biotechnology yesterday, today and tomorrow.

Adv Biochem Eng Biotechnol (2003). 80, IX-XV. Ewalt, K. L., J. P. Hendrick, et al. In vivo observation of polypeptide flux through the bacterial chaperonin system. Cell

(1997). 90, 491-500. Fahnert, B., H. Lilie, et al. Inclusion bodies: formation and utilisation. Adv Biochem Eng Biotechnol (2004). 89, 93-142. Fall, R. R. and P. R. Vagelos. Acetyl coenzyme A carboxylase. Proteolytic modification of biotin carboxyl carrier protein.

J Biol Chem (1973). 248, 2078-88. Fandrich, M. and C. M. Dobson. The behaviour of polyamino acids reveals an inverse side chain effect in amyloid

structure formation. Embo J (2002). 21, 5682-90. Fedorov, A. N. and T. O. Baldwin. Cotranslational protein folding. J Biol Chem (1997). 272, 32715-8. Feuser, J., J. Walter, et al. Cell/adsorbent interactions in expanded bed adsorption of proteins. Bioseparation (1999). 8, 99-

109. Flaschel, E. and K. Friehs. Improvement of downstream processing of recombinant proteins by means of genetic

engineering methods. Biotechnol Adv (1993). 11, 31-77. Ford, C. F., I. Suominen, et al. Fusion tails for the recovery and purification of recombinant proteins. Protein Expr Purif

(1991). 2, 95-107. Forsyth, W. R., J. M. Antosiewicz, et al. Empirical relationships between protein structure and carboxyl pKa values in

proteins. Proteins (2002). 48, 388-403. Freedman, R. B. Protein disulfide isomerase: multiple roles in the modification of nascent secretory proteins. Cell (1989).

57, 1069-72. Fuh, G., M. G. Mulkerrin, et al. The human growth hormone receptor. Secretion from Escherichia coli and disulfide

bonding pattern of the extracellular binding domain. J Biol Chem (1990). 265, 3111-5. Fukuda, H., M. Arai, et al. Folding of green fluorescent protein and the cycle3 mutant. Biochemistry (2000). 39, 12025-32.

84

Gaberc-Porekar, V. and V. Menart. Perspectives of immobilized-metal affinity chromatography. J Biochem Biophys Methods (2001). 49, 335-60.

Galloway, C. A., M. P. Sowden, et al. Increasing the yield of soluble recombinant protein expressed in E. coli by induction during late log phase. Biotechniques (2003). 34, 524-6, 528, 530.

Garke, G., W. D. Deckwer, et al. Preparative two-step purification of recombinant human basic fibroblast growth factor from high-cell-density cultivation of Escherichia coli. J Chromatogr B Biomed Sci Appl (2000). 737, 25-38.

Gavin, A. C., M. Bosche, et al. Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature (2002). 415, 141-7.

Geisse, S., H. Gram, et al. Eukaryotic expression systems: a comparison. Protein Expr Purif (1996). 8, 271-82. Gellissen, G., K. Melber, et al. Heterologous protein production in yeast. Antonie Van Leeuwenhoek (1992). 62, 79-93. Geng, X., Chang, X. High-performance hydrophobic interaction chromatography as a tool for protein refolding. J

Chromatogr A (1992). 599, 185-194. Georgiou, G., J. N. Telford, et al. Localization of inclusion bodies in Escherichia coli overproducing beta-lactamase or

alkaline phosphatase. Appl Environ Microbiol (1986). 52, 1157-61. Georgiou, G. and P. Valax. Expression of correctly folded proteins in Escherichia coli. Curr Opin Biotechnol (1996). 7,

190-7. Germino, J. and D. Bastia. Rapid purification of a cloned gene product by genetic fusion and site-specific proteolysis.

Proc Natl Acad Sci U S A (1984). 81, 4692-6. Gething, M. J. and J. Sambrook. Protein folding in the cell. Nature (1992). 355, 33-45. Glatz, C. E. Separation processes in biotechnology. Precipitation. Bioprocess Technol (1990). 9, 329-56. Gluzman, Y. SV40-transformed simian cells support the replication of early SV40 mutants. Cell (1981). 23, 175-82. Goeddel, D. V., H. L. Heyneker, et al. Direct expression in Escherichia coli of a DNA sequence coding for human growth

hormone. Nature (1979). 281, 544-8. Goeddel, D. V., D. G. Kleid, et al. Expression in Escherichia coli of chemically synthesized genes for human insulin. Proc

Natl Acad Sci U S A (1979). 76, 106-10. Goldberg, M. E., R. Rudolph, et al. A kinetic study of the competition between renaturation and aggregation during the

refolding of denatured-reduced egg white lysozyme. Biochemistry (1991). 30, 2790-7. Gouda, H., M. Shiraishi, et al. NMR Study of the Interaction between the B Domain of Staphylococcal Protein A and the

Fc Portion of Immunoglobulin G. Biochemistry (1997). 37, 129-136. Graslund, T., G. Lundin, et al. Charge engineering of a protein domain to allow efficient ion-exchange recovery. Protein

Eng (2000). 13, 703-9. Green, N. M. Avidin. Adv Protein Chem (1975). 29, 85-133. Gribskov, M. and R. R. Burgess. Overexpression and purification of the sigma subunit of Escherichia coli RNA

polymerase. Gene (1983). 26, 109-18. Griffin, B. A., S. R. Adams, et al. Specific covalent labeling of recombinant protein molecules inside live cells. Science

(1998). 281, 269-72. Gronenborn, B. Overproduction of phage lambda repressor under control of the lac promotor of Escherichia coli. Mol Gen

Genet (1976). 148, 243-50. Guan, K. L. and J. E. Dixon. Eukaryotic proteins expressed in Escherichia coli: an improved thrombin cleavage and

purification procedure of fusion proteins with glutathione S-transferase. Anal Biochem (1991). 192, 262-7. Gulich, S., M. Uhlen, et al. Protein engineering of an IgG-binding domain allows milder elution conditions during affinity

chromatography. J Biotechnol (2000). 76, 233-44. Gunneriusson, E., P. Samuelson, et al. Staphylococcal surface display of immunoglobulin A (IgA)- and IgE-specific in

vitro-selected binding proteins (affibodies) based on Staphylococcus aureus protein A. Appl Environ Microbiol (1999). 65, 4134-40.

Guss, B., M. Eliasson, et al. Structure of the IgG-binding regions of streptococcal protein G. Embo J (1986). 5, 1567-75. Hagel, P., J. J. Gerding, et al. Cyanate formation in solutions of urea. I. Calculation of cyanate concentrations at different

temperature and pH. Biochim Biophys Acta (1971). 243, 366-73. Hale, G., A. Drumm, et al. Repeated cleaning of protein A affinity column with sodium hydroxide. J Immunol Methods

(1994). 171, 15-21. Hall, M. N., J. Gabay, et al. A role for mRNA secondary structure in the control of translation initiation. Nature (1982).

295, 616-8. Hammarberg, B., T. Moks, et al. Differential stability of recombinant human insulin-like growth factor II in Escherichia

coli and Staphylococcus aureus. J Biotechnol (1990). 14, 423-37. Hammarstrom, M., N. Hellgren, et al. Rapid screening for improved solubility of small human proteins produced as fusion

proteins in Escherichia coli. Protein Sci (2002). 11, 313-21. Hammond, C. and A. Helenius. Quality control in the secretory pathway. Curr Opin Cell Biol (1995). 7, 523-9. Hannig, G. and S. C. Makrides. Strategies for optimizing heterologous protein expression in Escherichia coli. Trends

Biotechnol (1998). 16, 54-60. Hansson, M., S. Stahl, et al. Single-step recovery of a secreted recombinant protein by expanded bed adsorption.

Biotechnology (N Y) (1994). 12, 285-8. Hansson, M., S. Stahl, et al. Expression of recombinant proteins on the surface of the coagulase-negative bacterium

Staphylococcus xylosus. J Bacteriol (1992). 174, 4239-45.

85

Harrison, S. C. and R. Durbin. Is there a single pathway for the folding of a polypeptide chain? Proc Natl Acad Sci U S A (1985). 82, 4028-30.

Hartl, F. U. and M. Hayer-Hartl. Molecular chaperones in the cytosol: from nascent chain to folded protein. Science (2002). 295, 1852-8.

Head, J. F. A better grip on calmodulin. Curr Biol (1992). 2, 609-11. Hearn, M. T. and D. Acosta. Applications of novel affinity cassette methods: use of peptide fusion handles for the

purification of recombinant proteins. J Mol Recognit (2001). 14, 323-69. Hedhammar, M., Graslund, T., Hober, S. Protein engineering strategies for selective protein purification. Chemical

Engineering & Technology (2005). In Press. Hefti, M. H., C. J. Van Vugt-Van der Toorn, et al. A novel purification method for histidine-tagged proteins containing a

thrombin cleavage site. Anal Biochem (2001). 295, 180-5. Henner, D. J. Expression of heterologous genes in Bacillus subtilis. Methods Enzymol (1990). 185, 199-201. Hewinson, R. G. and W. P. Russell. Processing and secretion by Escherichia coli of a recombinant form of the

immunogenic protein MPB70 of Mycobacterium bovis. J Gen Microbiol (1993). 139 Pt 6, 1253-9. Hjorth, R. Expanded-bed adsorption in industrial bioprocessing: recent developments. Trends Biotechnol (1997). 15, 230-

5. Ho, I. C., M. R. Hodge, et al. The proto-oncogene c-maf is responsible for tissue-specific expression of interleukin-4. Cell

(1996). 85, 973-83. Hochuli, E. Large-scale chromatography of recombinant proteins. J Chromatogr (1988). 444, 293-302. Hochuli, E., H. Dobeli, et al. New metal chelate adsorbent selective for proteins and peptides containing neighbouring

histidine residues. J Chromatogr (1987). 411, 177-84. Hockney, R. C. Recent developments in heterologous protein production in Escherichia coli. Trends Biotechnol (1994).

12, 456-63. Hodgson, J. Expression systems: a user's guide. Emphasis has shifted from the vector construct to the host organism.

Biotechnology (N Y) (1993). 11, 887-93. Hopp, T. P., B. Gallis, et al. A short polypeptide marker sequence useful for recombinant protein identification and

purification. Bio/Technology (1988). 6, 1204-1210. Hu, S. Y., J. L. Wu, et al. Production of tilapia insulin-like growth factor-2 in high cell density cultures of recombinant

Escherichia coli. J Biotechnol (2004). 107, 161-71. Hu, X., Q. Shi, et al. Specific replacement of consecutive AGG codons results in high-level expression of human cardiac

troponin T in Escherichia coli. Protein Expr Purif (1996). 7, 289-93. Idicula-Thomas, S. and P. V. Balaji. Understanding the relationship between the primary structure of proteins and its

propensity to be soluble on overexpression in Escherichia coli. Protein Sci (2005). 14, 582-92. Imamura, H., B. Jeon, et al. High level expression of Thermococcus litoralis 4-alpha-glucanotransferase in a soluble form

in Escherichia coli with a novel expression system involving minor arginine tRNAs and GroELS. FEBS Lett (1999). 457, 393-6.

Iordanescu, S. Host controlled restriction mutants of Staphylococcus aureus. Arch Roum Pathol Exp Microbiol (1975). 34, 55-8.

Itakura, K., T. Hirose, et al. Expression in Escherichia coli of a chemically synthesized gene for the hormone somatostatin. Science (1977). 198, 1056-63.

Ito, T., T. Chiba, et al. A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc Natl Acad Sci U S A (2001). 98, 4569-74.

Ito, Y., T. Kawama, et al. Evolution of an arbitrary sequence in solubility. J Mol Evol (2004). 58, 196-202. Jackson, S. E. How do small single-domain proteins fold? Fold Des (1998). 3, R81-91. Janson, J. C., Rydén, L. Protein purification, Principles, high resolution methods and applications, (1998). Wiley-Liss. Jenkins, N., R. B. Parekh, et al. Getting the glycosylation right: implications for the biotechnology industry. Nat

Biotechnol (1996). 14, 975-81. Jenny, R. J., K. G. Mann, et al. A critical review of the methods for cleavage of fusion proteins with thrombin and factor

Xa. Protein Expr Purif (2003). 31, 1-11. Johansson, H. J., C. Jagersten, et al. Large scale recovery and purification of periplasmic recombinant protein from E. coli

using expanded bed adsorption chromatography followed by new ion exchange media. J Biotechnol (1996). 48, 9-14.

Johansson, M. U., I. M. Frick, et al. Structure, specificity, and mode of interaction for bacterial albumin-binding modules. J Biol Chem (2002). 277, 8114-20.

Jungbauer, A., W. Kaar, et al. Folding and refolding of proteins in chromatographic beds. Curr Opin Biotechnol (2004). 15, 487-94.

Kapust, R. B. and D. S. Waugh. Escherichia coli maltose-binding protein is uncommonly effective at promoting the solubility of polypeptides to which it is fused. Protein Sci (1999). 8, 1668-74.

Karlsson, E., Ryden, L. Ion-Exchange Chromatography. Protein Purification: Principles, High-Resolution Methods and Applications. (1998). J. C. Janson, Ryden, L. Wiley-VCH.

Karpeisky, M., V. N. Senchenko, et al. Formation and properties of S-protein complex with S-peptide-containing fusion protein. FEBS Lett (1994). 339, 209-12.

Karplus, M. and D. L. Weaver. Protein-folding dynamics. Nature (1976). 260, 404-6.

86

Katchalski-Katzir, E. Immobilized enzymes--learning from past successes and failures. Trends Biotechnol (1993). 11, 471-8.

Katoh, S., Katoh, Y. Continuous refolding of lysosyme with fed-batch addition of denatured protein solution. Process Biochemistry (2000). 35, 1119-1124.

Kauzmann, W. Some factors in the interpretation of protein denaturation. Adv Protein Chem (1959). 14, 1-63. Keefe, A. D., D. S. Wilson, et al. One-step purification of recombinant proteins using a nanomolar-affinity streptavidin-

binding peptide, the SBP-Tag. Protein Expr Purif (2001). 23, 440-6. Keranen, S. and M. Penttila. Production of recombinant proteins in the filamentous fungus Trichoderma reesei. Curr Opin

Biotechnol (1995). 6, 534-7. Khushoo, A., Y. Pal, et al. Extracellular expression and single step purification of recombinant Escherichia coli L-

asparaginase II. Protein Expr Purif (2004). 38, 29-36. Kim, C. S., Lee, EK. Effects of operating parameters in in vitro renaturation of a fusion protein of human growth hormone

and glutathione S transferase from inclusion body. Process Biochemistry (2000). 36, 111-117. Kim, J. S. and R. T. Raines. Ribonuclease S-peptide as a carrier in fusion proteins. Protein Sci (1993). 2, 348-56. Kimata, Y., M. Iwaki, et al. A novel mutation which enhances the fluorescence of green fluorescent protein at high

temperatures. Biochem Biophys Res Commun (1997). 232, 69-73. King, J., C. Haase-Pettingell, et al. Thermolabile folding intermediates: inclusion body precursors and chaperonin

substrates. Faseb J (1996). 10, 57-66. Kleber-Janke, T. and W. M. Becker. Use of modified BL21(DE3) Escherichia coli cells for high-level expression of

recombinant peanut allergens affected by poor codon usage. Protein Expr Purif (2000). 19, 419-24. Knappik, A. and A. Pluckthun. An improved affinity tag based on the FLAG peptide for the detection and purification of

recombinant antibody fragments. Biotechniques (1994). 17, 754-61. Knaust, R. K. and P. Nordlund. Screening for soluble expression of recombinant proteins in a 96-well format. Anal

Biochem (2001). 297, 79-85. Knorre, W. A., W. D. Deckwer, et al. High cell density fermentation of recombinant Escherichia coli with computer-

controlled optimal growth rate. Ann N Y Acad Sci (1991). 646, 300-6. Kohli, B. M. and C. Ostermeier. A Rubredoxin based system for screening of protein expression conditions and on-line

monitoring of the purification process. Protein Expr Purif (2003). 28, 362-7. Krivi, G. G., M. L. Bittner, et al. Purification of recA-based fusion proteins by immunoadsorbent chromatography.

Characterization of a major antigenic determinant of Escherichia coli recA protein. J Biol Chem (1985). 260, 10263-7.

Kronina, V. V., H. J. Wirth, et al. Characterisation by immobilised metal ion affinity chromatographic procedures of the binding behaviour of several synthetic peptides designed to have high affinity for Cu(II) ions. J Chromatogr A (1999). 852, 261-72.

Kukuruzinska, M. A. and K. Lennon. Protein N-glycosylation: molecular genetics and functional significance. Crit Rev Oral Biol Med (1998). 9, 415-48.

Kweon, D. H., D. H. Lee, et al. Characterization of polycationic amino acids fusion systems for ion-exchange purification of cyclodextrin glycosyltransferase from recombinant Escherichia coli. Biotechnol Prog (2002). 18, 303-8.

Kweon, D. H., D. H. Lee, et al. Solid-phase refolding of cyclodextrin glycosyltransferase adsorbed on cation-exchange resin. Biotechnol Prog (2004). 20, 277-83.

Labrou, N. E. Design and selection of ligands for affinity chromatography. J Chromatogr B Analyt Technol Biomed Life Sci (2003). 790, 67-78.

Lamla, T., W. Stiege, et al. An improved protein bioreactor: efficient product isolation during in vitro protein biosynthesis via affinity tag. Mol Cell Proteomics (2002). 1, 466-71.

Langone, J. J. Applications of immobilized protein A in immunochemical techniques. J Immunol Methods (1982). 55, 277-96.

Larrick, J. W. and D. W. Thomas. Producing proteins in transgenic plants and animals. Curr Opin Biotechnol (2001). 12, 411-8.

Larsson, M., E. Brundell, et al. A general bacterial expression system for functional analysis of cDNA-encoded proteins. Protein Expr Purif (1996). 7, 447-57.

Lesley, S. A. High-throughput proteomics: protein expression and purification in the postgenomic world. Protein Expr Purif (2001). 22, 159-64.

Lesley, S. A., J. Graziano, et al. Gene expression response to misfolded protein as a screen for soluble recombinant protein. Protein Eng (2002). 15, 153-60.

Levinthal, C. How to fold graciously. Mossbauer Spectroscopy in Biological Systems. (1969). P. Debrunner, Tibris, J.C.M., Münck, E. University of Illionois Press: 22-24.

Lewis, G., I. W. Taylor, et al. The application of multi-parameter flow cytometry to the study of recombinant Escherichia coli batch fermentation processes. J Ind Microbiol Biotechnol (2004). 31, 311-22.

Li, M., Z. G. Su, et al. In vitro protein refolding by chromatographic procedures. Protein Expr Purif (2004). 33, 1-10. Li, M., G. Zhang, et al. Dual gradient ion-exchange chromatography improved refolding yield of lysozyme. J Chromatogr

A (2002). 959, 113-20. Lilie, H., E. Schwarz, et al. Advances in refolding of proteins produced in E. coli. Curr Opin Biotechnol (1998). 9, 497-

501.

87

Lim, C. R., Y. Kimata, et al. Thermosensitivity of green fluorescent protein fluorescence utilized to reveal novel nuclear-like compartments in a mutant nucleoporin NSP1. J Biochem (Tokyo) (1995). 118, 13-7.

Lin, T. Y. and S. N. Timasheff. On the role of surface tension in the stabilization of globular proteins. Protein Sci (1996). 5, 372-81.

Linhult, M., S. Gulich, et al. Affinity ligands for industrial protein purification. Protein Pept Lett (2005). 12, 305-10. Liu, H. S., Chang, C.K. Chaperon solvent plug to enhance refolding in size exclusion chromatography. Enzyme and

Microbial Technology (2003). 33, 424-429. Lofdahl, S., B. Guss, et al. Gene for staphylococcal protein A. Proc Natl Acad Sci U S A (1983). 80, 697-701. Lorimer, G. H. A quantitative assessment of the role of the chaperonin proteins in protein folding in vivo. Faseb J (1996).

10, 5-9. Luckow, V. A. and M. D. Summers. Signals important for high-level expression of foreign genes in Autographa

californica nuclear polyhedrosis virus expression vectors. Virology (1988). 167, 56-71. Maga, E. A. and J. D. Murray. Mammary gland expression of transgenes and the potential for altering the properties of

milk. Biotechnology (N Y) (1995). 13, 1452-7. Maier, T., N. Drapal, et al. Strep-tag II affinity purification: an approach to study intermediates of metalloenzyme

biosynthesis. Anal Biochem (1998). 259, 68-73. Maier, T., L. Ferbitz, et al. A cradle for new proteins: trigger factor at the ribosome. Curr Opin Struct Biol (2005). 15,

204-12. Majander, K., L. Anton, et al. Extracellular secretion of polypeptides using a modified Escherichia coli flagellar secretion

apparatus. Nat Biotechnol (2005). 23, 475-81. Makrides, S. C. Components of vectors for gene transfer and expression in mammalian cells. Protein Expr Purif (1999).

17, 183-202. Maksimow, M., K. Hakkila, et al. Simultaneous detection of bacteria expressing GFP and DsRed genes with a flow

cytometer. Cytometry (2002). 47, 243-7. Malakhov, M. P., M. R. Mattern, et al. SUMO fusions and SUMO-specific protease for efficient expression and

purification of proteins. J Struct Funct Genomics (2004). 5, 75-86. Mann, M., R. C. Hendrickson, et al. Analysis of proteins and proteomes by mass spectrometry. Annu Rev Biochem (2001).

70, 437-73. Marston, F. A. The purification of eukaryotic polypeptides synthesized in Escherichia coli. Biochem J (1986). 240, 1-12. Martin-Farmer, J. and G. R. Janssen. A downstream CA repeat sequence increases translation from leadered and

unleadered mRNA in Escherichia coli. Mol Microbiol (1999). 31, 1025-38. Matthew, J. B. and F. R. Gurd. Stabilization and destabilization of protein structure by charge interactions. Methods

Enzymol (1986). 130, 437-53. Matthew, J. B., F. R. Gurd, et al. pH-dependent processes in proteins. CRC Crit Rev Biochem (1985). 18, 91-197. Maxwell, K. L., A. K. Mittermaier, et al. A simple in vivo assay for increased protein solubility. Protein Sci (1999). 8,

1908-11. Mayer, M. and J. Buchner. Refolding of inclusion body proteins. Methods Mol Med (2004). 94, 239-54. McDonald, J. K., B. B. Zeitman, et al. New observations on the substrate specificity of cathepsin C (dipeptidyl

aminopeptidase I). Including the degradation of beta-corticotropin and other peptide hormones. J Biol Chem (1969). 244, 2693-709.

Mehler, A. Glutathione S-transferase function in detoxification reactions. Textbook of Biochemistry. (1993). T. Devlin. Wiley-Liss: 523-524.

Misawa, S. and I. Kumagai. Refolding of therapeutic proteins produced in Escherichia coli as inclusion bodies. Biopolymers (1999). 51, 297-307.

Mitraki, A. and J. King. Amino acid substitutions influencing intracellular protein folding pathways. FEBS Lett (1992). 307, 20-5.

Miyashita, K., M. Kusumi, et al. Expression and purification of recombinant 3C proteinase of Coxsackievirus B3. Biosci Biotechnol Biochem (1992). 56, 746-50.

Moks, T., L. Abrahmsen, et al. Staphylococcal protein A consists of five IgG-binding domains. Eur J Biochem (1986). 156, 637-43.

Mosavi, L. K. and Z. Y. Peng. Structure-based substitutions for increased solubility of a designed protein. Protein Eng (2003). 16, 739-45.

Muise, O. and E. Holler. Interaction of DNA polymerase I of Escherichia coli with nucleotides. Antagonistic effects of single-stranded polynucleotide homopolymers. Biochemistry (1985). 24, 3618-22.

Murby, M., E. Samuelsson, et al. Hydrophobicity engineering to increase solubility and stability of a recombinant protein from respiratory syncytial virus. Eur J Biochem (1995). 230, 38-44.

Muronetz, V. I., M. Sholukh, et al. Use of protein-protein interactions in affinity chromatography. J Biochem Biophys Methods (2001). 49, 29-47.

Nagai, K. and H. C. Thogersen. Generation of beta-globin by sequence-specific proteolysis of a hybrid protein produced in Escherichia coli. Nature (1984). 309, 810-2.

Nakano, H., T. Tanaka, et al. An increased rate of cell-free protein synthesis by condensing wheat-germ extract with ultrafiltration membranes. Biosci Biotechnol Biochem (1994). 58, 631-4.

Neri, D., C. de Lalla, et al. Calmodulin as a versatile tag for antibody fragments. Biotechnology (N Y) (1995). 13, 373-7.

88

Nicola, A. V., W. Chen, et al. Co-translational folding of an alphavirus capsid protein in the cytosol of living cells. Nat Cell Biol (1999). 1, 341-5.

Niederauer, M. Q., I. Suominen, et al. Characterization and polyelectrolyte precipitation of beta-galactosidase containing genetic fusions of charged polypeptides. Biotechnol Prog (1994). 10, 237-45.

Nilsson, B., T. Moks, et al. A synthetic IgG-binding domain based on staphylococcal protein A. Protein Eng (1987). 1, 107-13.

Nilsson, J., M. Larsson, et al. Multiple affinity domains for the detection, purification and immobilization of recombinant proteins. J Mol Recognit (1996). 9, 585-94.

Nilsson, J., P. Nilsson, et al. Competitive elution of protein A fusion proteins allows specific recovery under mild conditions. Eur J Biochem (1994). 224, 103-8.

Nilsson, J., S. Stahl, et al. Affinity fusion strategies for detection, purification, and immobilization of recombinant proteins. Protein Expr Purif (1997). 11, 1-16.

Nishihara, K., M. Kanemori, et al. Overexpression of trigger factor prevents aggregation of recombinant proteins in Escherichia coli. Appl Environ Microbiol (2000). 66, 884-9.

Noiva, R. and W. J. Lennarz. Protein disulfide isomerase. A multifunctional protein resident in the lumen of the endoplasmic reticulum. J Biol Chem (1992). 267, 3553-6.

Nolan, J. P., S. Lauer, et al. Flow cytometry: a versatile tool for all phases of drug discovery. Drug Discov Today (1999). 4, 173-180.

Nord, K., E. Gunneriusson, et al. Binding proteins selected from combinatorial libraries of an alpha-helical bacterial receptor domain. Nat Biotechnol (1997). 15, 772-7.

Nord, K., E. Gunneriusson, et al. Ligands selected from combinatorial libraries of protein A for use in affinity capture of apolipoprotein A-1M and taq DNA polymerase. J Biotechnol (2000). 80, 45-54.

Nord, K., J. Nilsson, et al. A combinatorial library of an alpha-helical bacterial receptor domain. Protein Eng (1995). 8, 601-8.

Noronha, S., J. Kaufman, et al. Use of streamline chelating for capture and purification of poly-His-tagged recombinant proteins. Bioseparation (1999). 8, 145-51.

Nossal, N. G. and L. A. Heppel. The release of enzymes by osmotic shock from Escherichia coli in exponential phase. J Biol Chem (1966). 241, 3055-62.

Nygren, P. A., M. Eliasson, et al. Analysis and use of the serum albumin binding domains of streptococcal protein G. J Mol Recognit (1988). 1, 69-74.

Nygren, P. A., C. Ljungquist, et al. Species-dependent binding of serum albumins to the streptococcal receptor protein G. Eur J Biochem (1990). 193, 143-8.

Nygren, P. A., S. Stahl, et al. Engineering proteins to facilitate bioprocessing. Trends Biotechnol (1994). 12, 184-8. Ohgushi, M. and A. Wada. 'Molten-globule state': a compact form of globular proteins with mobile side-chains. FEBS Lett

(1983). 164, 21-4. Olejnik, J., S. Sonar, et al. Photocleavable biotin derivatives: a versatile approach for the isolation of biomolecules. Proc

Natl Acad Sci U S A (1995). 92, 7590-4. Olszewski, K. A., A. Kolinski, et al. Folding simulations and computer redesign of protein A three-helix bundle motifs.

Proteins (1996). 25, 286-99. O'Neil, K. T. and W. F. DeGrado. A thermodynamic scale for the helix-forming tendencies of the commonly occurring

amino acids. Science (1990). 250, 646-51. Panda, A. K. Bioprocessing of therapeutic proteins from the inclusion bodies of Escherichia coli. Adv Biochem Eng

Biotechnol (2003). 85, 43-93. Parente, E. S. and D. B. Wetlaufer. Influence of urea on the high-performance cation-exchange chromatography of hen

egg white lysozyme. J Chromatogr (1984). 288, 389-98. Pauling, L. and R. B. Corey. The pleated sheet, a new layer configuration of polypeptide chains. Proc Natl Acad Sci U S A

(1951). 37, 251-6. Pauling, L., R. B. Corey, et al. The structure of proteins; two hydrogen-bonded helical configurations of the polypeptide

chain. Proc Natl Acad Sci U S A (1951). 37, 205-11. Pelham, H. R. and R. J. Jackson. An efficient mRNA-dependent translation system from reticulocyte lysates. Eur J

Biochem (1976). 67, 247-56. Pelletier, J. N., F. X. Campbell-Valois, et al. Oligomerization domain-directed reassembly of active dihydrofolate

reductase from rationally designed fragments. Proc Natl Acad Sci U S A (1998). 95, 12141-6. Phizicky, E., P. I. Bastiaens, et al. Protein analysis on a proteomic scale. Nature (2003). 422, 208-15. Porath, J., J. Carlsson, et al. Metal chelate affinity chromatography, a new approach to protein fractionation. Nature

(1975). 258, 598-9. Prickett, K. S., D. C. Amberg, et al. A calcium-dependent antibody for identification and purification of recombinant

proteins. Biotechniques (1989). 7, 580-9. Pryor, K. D. and B. Leiting. High-level expression of soluble protein in Escherichia coli using a His6-tag and maltose-

binding-protein double-affinity fusion system. Protein Expr Purif (1997). 10, 309-19. Przybycien, T. M., J. P. Dunn, et al. Secondary structure characterization of beta-lactamase inclusion bodies. Protein Eng

(1994). 7, 131-6.

89

Ptitsyn, O. B. [Stages in the mechanism of self-organization of protein molecules]. Dokl Akad Nauk SSSR (1973). 210, 1213-5.

Rees, A. R., D. Staunton, et al. Antibody design: beyond the natural limits. Trends Biotechnol (1994). 12, 199-206. Reid, B. G. and G. C. Flynn. Chromophore formation in green fluorescent protein. Biochemistry (1997). 36, 6786-91. Richards, F. M. Titration of amino groups released during the digestion of ribonuclease by subtilisin. C R Trav Lab

Carlsberg [Chim] (1955). 29, 322-8. Righetti, P. G. and T. Caravaggio. Isoelectric points and molecular weights of proteins. J Chromatogr (1976). 127, 1-28. Rogl, H., K. Kosemund, et al. Refolding of Escherichia coli produced membrane protein inclusion bodies immobilised by

nickel chelating chromatography. FEBS Lett (1998). 432, 21-6. Romanos, M. A., F. J. Hughes, et al. Production of a phosphorylated GST::HPV-6 E7 fusion protein using a yeast

expression vector and glutathione S-transferase fusions. Gene (1995). 152, 137-8. Routzahn, K. M. and D. S. Waugh. Differential effects of supplementary affinity tags on the solubility of MBP fusion

proteins. J Struct Funct Genomics (2002). 2, 83-92. Ruddon, R. W. and E. Bedows. Assisted protein folding. J Biol Chem (1997). 272, 3125-8. Rudolph, R. and H. Lilie. In vitro folding of inclusion body proteins. Faseb J (1996). 10, 49-56. Sachdev, D. and J. M. Chirgwin. Order of fusions between bacterial and mammalian proteins can determine solubility in

Escherichia coli. Biochem Biophys Res Commun (1998). 244, 933-7. Sachdev, D. and J. M. Chirgwin. Solubility of proteins isolated from inclusion bodies is enhanced by fusion to maltose-

binding protein or thioredoxin. Protein Expr Purif (1998). 12, 122-32. Sambrook, J., Fritch,FE., Maniatis,T. Molecular cloning. A Laboratory manual, (1989). Cold Spring Harbor Laboratory

Press. Samols, D., C. G. Thornton, et al. Evolutionary conservation among biotin enzymes. J Biol Chem (1988). 263, 6461-4. Samuelsson, E., P. Jonasson, et al. Affinity-assisted in vivo folding of a secreted human peptide hormone in Escherichia

coli. Nat Biotechnol (1996). 14, 751-5. Samuelsson, E., T. Moks, et al. Enhanced in vitro refolding of insulin-like growth factor I using a solubilizing fusion

partner. Biochemistry (1994). 33, 4207-11. Samuelsson, E. and M. Uhlen. Chaperone-like effect during in vitro refolding of insulin-like growth factor I using a

solubilizing fusion partner. Ann N Y Acad Sci (1996). 782, 486-94. Sandkvist, M. and M. Bagdasarian. Secretion of recombinant proteins by Gram-negative bacteria. Curr Opin Biotechnol

(1996). 7, 505-11. Sassenfeld, H. M. Engineering proteins for purification. Trends Biotechnol (1990). 8, 88-93. Sassenfeld, H. M. and S. J. Brewer. A polypeptide fusion designed for the purification of recombinant proteins.

Bio/Technology (1984). 2, 76-80. Sauer-Eriksson, A. E., G. J. Kleywegt, et al. Crystal structure of the C2 fragment of streptococcal protein G in complex

with the Fc domain of human IgG. Structure (1995). 3, 265-78. Sawasaki, T., T. Ogasawara, et al. A cell-free protein synthesis system for high-throughput proteomics. Proc Natl Acad

Sci U S A (2002). 99, 14652-7. Schatz, P. J. Use of peptide libraries to map the substrate specificity of a peptide-modifying enzyme: a 13 residue

consensus peptide specifies biotinylation in Escherichia coli. Biotechnology (N Y) (1993). 11, 1138-43. Schein, C. H., Noteborn, M.H.M. Formation of Soluble Recombinant Proteins in Escherichia Coli is Favored by Lower

Growth Temperature. Bio/Technology (1988). 6, 291-294. Schernthaner, G. Immunogenicity and allergenic potential of animal and human insulins. Diabetes Care (1993). 16 Suppl

3, 155-65. Schmidt, F. R. Recombinant expression systems in the pharmaceutical industry. Appl Microbiol Biotechnol (2004). 65,

363-72. Schmidt, T. G., J. Koepke, et al. Molecular interaction between the Strep-tag affinity peptide and its cognate target,

streptavidin. J Mol Biol (1996). 255, 753-66. Scopes, R. Protein Purification, Principle and Practice, (1994). Springer-Verlag. Scopes, R. K. and E. Algar. Yeast phosphoglycerate kinase -- evidence from affinity elution studies for conformational

changes on binding of substrates. FEBS Lett (1979). 106, 239-42. Seckler, R. and R. Jaenicke. Protein folding and protein refolding. Faseb J (1992). 6, 2545-52. Shapiro, H. M. Practical flow cytometry, (1995). Wiley-Liss. Sharp, P. M. and M. Bulmer. Selective differences among translation termination codons. Gene (1988). 63, 141-5. Sherwood, R. Protein fusions: bioseparation and application. Trends Biotechnol (1991). 9, 1-3. Shine, J., I. Fettes, et al. Expression of cloned beta-endorphin gene sequences by Escherichia coli. Nature (1980). 285,

456-63. Sieber, V., A. Plückthun, et al. Selecting proteins with improved stability by a phage-based method. Nature Biotechnology

(1998). 16, 955-960. Siemering, K. R., R. Golbik, et al. Mutations that suppress the thermosensitivity of green fluorescent protein. Curr Biol

(1996). 6, 1653-63. Simpson, J. C., R. Wellenreuther, et al. Systematic subcellular localization of novel proteins identified by large-scale

cDNA sequencing. EMBO Rep (2000). 1, 287-92. Singh, M., R. Sharma, et al. Biotechnological applications of cyclodextrins. Biotechnol Adv (2002). 20, 341-59.

90

Sjolander, A., Stahl, S., Perlmann, P. Bacterial expression systems based on protein A and protein G designed for the production of immunogens, applications to Plasmodium falciparum malaria antigens. ImmunoMethods2 (1993). 79-92.

Sluyterman, L. A. A., Elgersma, O. Chromatofocusing: Isoelectric focusing on ion-exchange columns I. J Chromatog (1978). 150, 17-30.

Smith, D. B. and K. S. Johnson. Single-step purification of polypeptides expressed in Escherichia coli as fusions with glutathione S-transferase. Gene (1988). 67, 31-40.

Smith, H. O. and K. W. Wilcox. A restriction enzyme from Hemophilus influenzae. I. Purification and general properties. J Mol Biol (1970). 51, 379-91.

Smith, J. C., R. B. Derbyshire, et al. Chemical synthesis and cloning of a poly(arginine)-coding gene fragment designed to aid polypeptide purification. Gene (1984). 32, 321-7.

Smith, M. C., T. C. Furman, et al. Chelating peptide-immobilized metal ion affinity chromatography. A new concept in affinity chromatography for recombinant proteins. J Biol Chem (1988). 263, 7211-5.

Smith, P. A., B. C. Tripp, et al. A plasmid expression system for quantitative in vivo biotinylation of thioredoxin fusion proteins in Escherichia coli. Nucleic Acids Res (1998). 26, 1414-20.

Sober, H. A. and E. A. Peterson. Protein chromatography on ion exchange cellulose. Fed Proc (1958). 17, 1116-26. Sofer, G. Preparative chromatographic separations in pharmaceutical, diagnostic, and biotechnology industries: current

and future trends. J Chromatogr A (1995). 707, 23-8. Sofer, G., Hagel, L. Handbook of Process Chromatography: A Guide to Optimization, Scale-up and Validation, (1997).

Academic Press. Spirin, A. S., V. I. Baranov, et al. A continuous cell-free translation system capable of producing polypeptides in high

yield. Science (1988). 242, 1162-4. Stahl, S., Hober, S., Nilsson, J., Uhlén, U., Nygren, P.A. Genetic approches to facilitate protein purification. Isolation and

purification of proteins. (2003). R. Hatti-Kaul, Mattiasson, B. Marcel Dekker, Inc. Stahl, S. and P. A. Nygren. The use of gene fusions to protein A and protein G in immunology and biotechnology. Pathol

Biol (Paris) (1997). 45, 66-76. Stahl, S., A. Sjolander, et al. A dual expression system for the generation, analysis and purification of antibodies to a

repeated sequence of the Plasmodium falciparum antigen Pf155/RESA. J Immunol Methods (1989). 124, 43-52.

Steers, E., Jr., P. Cuatrecasas, et al. The purification of beta-galactosidase from Escherichia coli by affinity chromatography. J Biol Chem (1971). 246, 196-200.

Stenvall, M., J. Steen, et al. High-throughput solubility assay for purified recombinant protein immunogens. Biochim Biophys Acta (2005). 1752, 6-10.

Stevens, R. C. Design of high-throughput methods of protein production for structural biology. Structure Fold Des (2000). 8, R177-85.

Stofko-Hahn, R. E., D. W. Carr, et al. A single step purification for recombinant proteins. Characterization of a microtubule associated protein (MAP 2) fragment which associates with the type II cAMP-dependent protein kinase. FEBS Lett (1992). 302, 274-8.

Strandberg, L. and S. O. Enfors. Factors influencing inclusion body formation in the production of a fused protein in Escherichia coli. Appl Environ Microbiol (1991). 57, 1669-74.

Stryer, L. Biochemistry, (1996). W. H. Freeman and Company. Stubenrauch, K., A. Bachmann, et al. Purification of a viral coat protein by an engineered polyionic sequence. J

Chromatogr B Biomed Sci Appl (2000). 737, 77-84. Studier, F. W. Use of bacteriophage T7 lysozyme to improve an inducible T7 expression system. J Mol Biol (1991). 219,

37-44. Studier, F. W. and B. A. Moffatt. Use of bacteriophage T7 RNA polymerase to direct selective high-level expression of

cloned genes. J Mol Biol (1986). 189, 113-30. Suh, C. W., G. S. Choi, et al. Enzymic cleavage of fusion protein using immobilized urokinase covalently conjugated to

glyoxyl-agarose. Biotechnol Appl Biochem (2003). 37, 149-55. Tanford, C. How protein chemists learned about the hydrophobic factor. Protein Sci (1997). 6, 1358-66. Tashiro, M., R. Tejero, et al. High-resolution solution NMR structure of the Z domain of staphylococcal protein A. J Mol

Biol (1997). 272, 573-90. Terpe, K. Overview of tag protein fusions: from molecular and biochemical fundamentals to commercial systems. Appl

Microbiol Biotechnol (2003). 60, 523-33. Thiele, C. and F. Fahrenholz. Photoaffinity labeling of central cholecystokinin receptors with high efficiency.

Biochemistry (1993). 32, 2741-6. Thomas, J. G., A. Ayling, et al. Molecular chaperones, folding catalysts, and the recovery of active recombinant proteins

from E. coli. To fold or to refold. Appl Biochem Biotechnol (1997). 66, 197-238. Thorn, K. S., N. Naber, et al. A novel method of affinity-purifying proteins using a bis-arsenical fluorescein. Protein Sci

(2000). 9, 213-7. Tripet, B., L. Yu, et al. Engineering a de novo-designed coiled-coil heterodimerization domain off the rapid detection,

purification and characterization of recombinantly expressed peptides and proteins. Protein Eng (1996). 9, 1029-42.

91

Tsumoto, K., D. Ejima, et al. Practical considerations in refolding proteins from inclusion bodies. Protein Expr Purif (2003). 28, 1-8.

Uhlen, M., E. Bjorling, et al. A human protein atlas for normal and cancer tissues based on antibody proteomics. Mol Cell Proteomics (2005).

Uhlén, M., B. Guss, et al. Complete sequence of the staphylococcal gene encoding protein A. J. Biol. Chem. (1984). 259, 1695-1702.

Urlaub, G. and L. A. Chasin. Isolation of Chinese hamster cell mutants deficient in dihydrofolate reductase activity. Proc Natl Acad Sci U S A (1980). 77, 4216-20.

Wacker, M., D. Linton, et al. N-linked glycosylation in Campylobacter jejuni and its functional transfer into E. coli. Science (2002). 298, 1790-3.

Waldo, G. S. Genetic screens and directed evolution for protein solubility. Curr Opin Chem Biol (2003). 7, 33-8. Waldo, G. S., B. M. Standish, et al. Rapid protein-folding assay using green fluorescent protein. Nat Biotechnol (1999).

17, 691-5. Walker, J. M. Protein structure, purification, charachterisation and function analysis. Principles and techniques of

Biochemistry and molecular biology. (2005). K. Wilson, Walker, J.M. Cambridge University Press. Walker, P. A., L. E. Leong, et al. Efficient and rapid affinity purification of proteins using recombinant fusion proteases.

Biotechnology (N Y) (1994). 12, 601-5. Wall, J. G. and A. Pluckthun. Effects of overexpressing folding modulators on the in vivo folding of heterologous proteins

in Escherichia coli. Curr Opin Biotechnol (1995). 6, 507-16. Vallejo, L. F. and U. Rinas. Strategies for the recovery of active proteins through refolding of bacterial inclusion body

proteins. Microb Cell Fact (2004). 3, 11. Weiss, E., J. Chatellier, et al. In vivo biotinylated recombinant antibodies: construction, characterization, and application

of a bifunctional Fab-BCCP fusion protein produced in Escherichia coli. Protein Expr Purif (1994). 5, 509-17. Werner, M. H., G. M. Clore, et al. Refolding proteins by gel filtration chromatography. FEBS Lett (1994). 345, 125-30. Westra, D. F., G. W. Welling, et al. Immobilised metal-ion affinity chromatography purification of histidine-tagged

recombinant proteins: a wash step with a low concentration of EDTA. J Chromatogr B Biomed Sci Appl (2001). 760, 129-36.

Wetlaufer, D. B. Nucleation, rapid folding, and globular intrachain regions in proteins. Proc Natl Acad Sci U S A (1973). 70, 697-701.

Wigley, W. C., R. D. Stidham, et al. Protein solubility and folding monitored in vivo by structural complementation of a genetic marker protein. Nat Biotechnol (2001). 19, 131-6.

Wilkinson, D. L. and R. G. Harrison. Predicting the solubility of recombinant proteins in Escherichia coli. Biotechnology (N Y) (1991). 9, 443-8.

Villaverde, A. and M. M. Carrio. Protein aggregation in recombinant bacteria: biological role of inclusion bodies. Biotechnol Lett (2003). 25, 1385-95.

Wilson, D. S., A. D. Keefe, et al. The use of mRNA display to select high-affinity protein-binding peptides. Proc Natl Acad Sci U S A (2001). 98, 3750-5.

Winson, M. K. and H. M. Davey. Flow cytometric analysis of microorganisms. Methods (2000). 21, 231-40. Woestenenk, E. A., M. Hammarstrom, et al. His tag effect on solubility of human proteins produced in Escherichia coli: a

comparison between four expression vectors. J Struct Funct Genomics (2004). 5, 217-29. Wolynes, P. G., J. N. Onuchic, et al. Navigating the folding routes. Science (1995). 267, 1619-20. Wong, S. L. Advances in the use of Bacillus subtilis for the expression and secretion of heterologous proteins. Curr Opin

Biotechnol (1995). 6, 517-22. Wood, W. I., D. J. Capon, et al. Expression of active human factor VIII from recombinant DNA clones. Nature (1984).

312, 330-7. Voss, S. and A. Skerra. Mutagenesis of a flexible loop in streptavidin leads to higher affinity for the Strep-tag II peptide

and improved performance in recombinant protein purification. Protein Eng (1997). 10, 975-82. Wulfing, C. and A. Pluckthun. Protein folding in the periplasm of Escherichia coli. Mol Microbiol (1994). 12, 685-92. Yip, T. T., Y. Nakagawa, et al. Evaluation of the interaction of peptides with Cu(II), Ni(II), and Zn(II) by high-

performance immobilized metal ion affinity chromatography. Anal Biochem (1989). 183, 159-71. Zardeneta, G. and P. M. Horowitz. Detergent, liposome, and micelle-assisted protein refolding. Anal Biochem (1994). 223,

1-6. Zardeneta, G. and P. M. Horowitz. Protein refolding at high concentrations using detergent/phospholipid mixtures. Anal

Biochem (1994). 218, 392-8. Zhang, C., R. T. Love, et al. Genetic engineering strategies for purification of recombinant proteins from canola by anion

exchange chromatography: an example of beta-glucuronidase. Biotechnol Prog (2001). 17, 161-7. Zhang, Z., J. Gildersleeve, et al. A new strategy for the synthesis of glycoproteins. Science (2004). 303, 371-3. Zhao, J. Y., C. F. Ford, et al. Polyelectrolyte precipitation of beta-galactosidase fusions containing poly-aspartic acid tails.

J Biotechnol (1990). 14, 273-83.