Microbiol Monogr (1)
`J.M. Shively: Inclusions in Prokaryotes
`DOI 10.1007/7171_009/Published online: 12 April 2006
`© Springer-Verlag Berlin Heidelberg 2006
`Protein Inclusion Bodies in Recombinant Bacteria
`
`Peter Neubauer1 (u) · Beatrix Fahnert2 · Hauke Lilie3 · Antonio Villaverde4
`1Bioprocess Engineering Laboratory,
`Department of Process and Environmental Engineering, University of Oulu,
`PO Box 4300, 90014 Oulu, Finland
`peter.neubauer@oulu.fi
`2Cardiff School of Biosciences, Biomedical Sciences Building (BIOSI2),
`Cardiff University, PO Box 911, Cardiff CF10 3US, UK
`3Institute of Biotechnology, Martin Luther University Halle-Wittenberg,
`06099 Halle (Saale), Germany
`4Institut de Biotecnologia i de Biomedicina and Departament de Genètica
`i de Microbiologia, Universitat Aut`onoma de Barcelona, Bellaterra, 08193 Barcelona,
`Spain
`
`1
`
`Introduction .
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`Protein Structure Determinants for Aggregation .
`2
`Protein Folding and Sequence Features .
`.
`.
`.
`.
`.
`2.1
`2.2 Disulfide Bonds .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`2.3 Membrane Proteins .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`2.4
`Glycosylation .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`238
`
`240
`240
`242
`243
`244
`
`244
`245
`247
`
`Structure and Composition of IBs .
`3
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`3.1 Morphology, Structure and Molecular Organization .
`3.2
`Compositional Analysis .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`
`248
`250
`252
`254
`258
`258
`
`259
`259
`259
`260
`261
`262
`264
`
`265
`265
`267
`268
`
`4
`4.1
`4.2
`4.3
`4.4
`4.5
`
`5
`
`5.1
`5.2
`5.3
`5.4
`5.5
`5.6
`
`6
`6.1
`6.2
`6.3
`
`Dynamics of IB Formation and Disintegration .
`Protein Synthesis and Aggregation .
`.
`.
`.
`.
`.
`.
`.
`Stress Responses to IBs .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Chaperone Action .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Periplasmic Response to Misfolded Protein .
`.
`.
`Response to Aggregation in Other Bacteria .
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`Impact of the Production Process,
`the Host and the Target Protein on Aggregation .
`Rate of Synthesis .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Cultivation Conditions
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Components of the Cultivation Medium .
`.
`.
`.
`.
`.
`Chaperones and Foldases .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Target Proteins with Disulfide Bonds
`.
`.
`.
`.
`.
`.
`.
`Fusion Proteins
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`
`.
`.
`.
`.
`.
`.
`.
`
`Production of IBs and Downstream Functionalization .
`Fermentation .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`IB Isolation and Purification .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Solubilization of IBs .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`
`Page 1
`
`KASHIV EXHIBIT 1010
`IPR2019-00791
`
`

`

`238
`
`P. Neubauer et al.
`
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Refolding of Proteins from IBs .
`6.4
`.
`.
`.
`.
`.
`.
`.
`.
`6.5 Disulfide Bond Formation During Protein Renaturation .
`.
`.
`.
`.
`.
`.
`.
`.
`6.6
`Improvement of Renaturation .
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`.
`Industrial Processes Based on Refolding of IB Proteins—Human t-PA .
`6.7
`
`7
`
`Outlook .
`
`References
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`.
`.
`.
`
`.
`
`.
`
`.
`.
`.
`.
`
`.
`
`.
`
`269
`270
`271
`273
`
`274
`
`274
`
`Abstract Fast and high-level expression of heterologous proteins in bacterial hosts often
`results in the accumulation of almost pure aggregates (inclusion bodies, IBs) of the target
`protein. Although knowledge of the pathways and influential factors of protein folding in
`vivo has increased for many years, the complexity of the cellular networks does not allow
`easily the prediction of favourable conditions for production of correctly folded proteins.
`Therefore, IB-based production is still a potential and straightforward strategy for the
`production of complex recombinant proteins. IB-based processes combine the advan-
`tages of a high concentration of the target protein produced in well-characterized bacteria
`such as Escherichia coli, efficient protocols for IB isolation, purification and in vitro pro-
`tein refolding without the need of elaborate coexpression systems and time-consuming
`trial-and-error expression optimization. Recent advances in understanding the molecu-
`lar physiology of IB formation and in resolubilization enable a streamlined development
`of fermentation processes to obtain a high-quality product. In addition, simple strategies
`have been established to improve the purification and renaturation of disulfide bond con-
`taining proteins allowing for a fast transfer of those processes to industrial production
`scale.
`
`1I
`
`ntroduction
`
`Thousands of DNA sequences from various sources have been expressed in
`bacteria such as in Escherichia coli and many of the target proteins have been
`obtained in their correct three-dimensional structure. However, in as many as
`40% of the cases the product accumulates in the form of insoluble aggregates
`(Fig. 1; Mayer and Buchner 2004).
`Mostly, investigators then start to optimize the conditions to obtain the
`soluble and correctly folded form of the product. For this production strat-
`egy a large number of new host strains, vectors and experiences have been
`described in the literature and it also has resulted in efficient processes. How-
`ever, despite the amount of accumulated knowledge this way is still based on
`long-lasting trial-and-error approaches and even experienced scientists may
`not predict successful and feasible process strategies.
`Especially a low yield is mostly obtained for eukaryotic secreted proteins
`when produced in E. coli, even though such proteins are often a target for
`drug research. Recent progress in understanding the protein folding in eu-
`karyotes, especially in the endoplasmic reticulum, has revealed the high com-
`plexity of this compartment and the significant difference to the periplasm
`
`Page 2
`
`

`

`Protein Inclusion Bodies in Recombinant Bacteria
`
`239
`
`Fig. 1 Electron micrograph of negatively stained Escherichia coli cells displaying inclusion
`bodies (IBs)
`
`of bacteria. In particular, it becomes better understood that a simple coex-
`pression of eukaryotic foldases in bacterial hosts will not guarantee that these
`proteins will work in the same way. Their action is not limited to the re-
`dox situation which has been considered for a long time, but extends to their
`interplay with other proteins and the availability of ATP in the eukaryotic
`endoplasmic reticulum.
`Therefore, inclusion body (IB) based processes belong today to the usual
`repertoire of producing target proteins, especially if the product is struc-
`turally complicated, for example it has a number of disulfide bonds or a mul-
`tidomain structure. IB-based processes are also a feasible alternative if the
`product is a domain fragment of a protein, for example from a receptor, or if it
`is either very large or a small peptide. An IB-based process demands the sol-
`ubilization and in vitro refolding of the target protein during the downstream
`processing, which might be considered to be difficult. However, generally the
`production of IBs is comparably easy and high yields are obtained, mostly on
`the order of 10 to more than 50% of the total cell protein. Also the initial iso-
`lation of the IBs, which contain the target protein in almost pure form, is easy
`by simple cell disruption and centrifugation. Consequently, in vitro optimiza-
`tion of the folding in a defined environment seems more straightforward for
`proteins which are not easily obtained in a soluble form in vivo, and is per-
`formed more easily than in a complex cellular system with a large number of
`interacting parameters and reaction sequences.
`This review will mainly focus on the characteristics of IBs which are pro-
`duced in the bacterium E. coli, although principally IBs can be obtained
`also in other organisms, even including eukaryotic hosts (Lefebvre-Legendre
`et al. 2005; Outeiro and Lindquist 2003). Although the formation of IBs has
`been reported extensively in many recent reviews (e.g. Baneyx and Mujacic
`2004; Fahnert et al. 2004; Hoffmann and Rinas 2004; Sorensen and Mortensen
`2005a), new results with respect to the composition and formation process
`
`Page 3
`
`

`

`240
`
`P. Neubauer et al.
`
`of IBs have improved our understanding of their nature and will be reported
`here specifically.
`
`rotein Structure Determinants for Aggregation
`
`2P
`
`Expression of recombinant proteins at a high rate often results in the ac-
`cumulation of IBs, independent of whether the protein is of homologous
`or heterologous origin. This formation of insoluble aggregates owing to the
`high, mostly non-natural expression of proteins has been considered for
`a long time to be an unspecific process driven by contacts between partially
`folded or unfolded peptides. However, studies in recent years have put this
`dogma into question. The methodological progress has been greatly influ-
`enced by the investigation of protein aggregation in human disorders, such
`as Alzheimer’s and Parkinson’s diseases. Aside from clear differences also
`homologies were found between the underlying principles of amyloid fibril
`formation in these disorders and the IBs in bacterial systems. These results
`have entirely changed the picture. Although more studies are needed to un-
`derstand this aggregation at the molecular, structural and physiological level,
`the studies performed so far have revealed that also protein aggregation in
`bacteria is a more specific process than previously expected driven by distinct
`characteristics of the target protein.
`
`2.1
`Protein Folding and Sequence Features
`
`Generally the function of most proteins is based on their solubility in connec-
`tion with a specific three-dimensional structure, whereas insoluble aggregates
`are usually toxic (Bucciantini et al. 2002), with exception of proteins devoted
`to specifically function in well-ordered aggregated structures, such as silk and
`spider web proteins. Therefore, it seems that both amino acid sequences of
`the proteins and the components of the cellular protein synthesis machin-
`ery have evolved to prevent uncontrolled protein aggregation (Dobson 2003).
`Thereby the soluble state is connected to one specific configuration, although
`a protein principally can reach different intermediate states prone to aggre-
`gation during the protein folding process. Factors which either destabilize
`the native structure or decelerate the folding pathway contribute to a higher
`aggregation.
`Protein aggregation is a specific process which starts with a slow nu-
`cleation phase, possibly through self-assembly of protein monomers via
`a nucleation-dependent pathway (Ventura 2005).
`Several observations indicate that the formation of IBs is likely to arise
`from specific and selective mechanisms, similar to amyloid fibril polymeriza-
`
`Page 4
`
`

`

`Protein Inclusion Bodies in Recombinant Bacteria
`
`241
`
`tion. Generally the IBs formed are relatively pure in respect to the aggregated
`protein. That is why they are very useful for protein production if a process ex-
`ists to refold them in vitro into their native structure, as discussed later. Other
`cellular proteins found in IBs often result from coisolation during the enrich-
`ment process more than from integration into the original IBs (Georgiou and
`Valax 1999), especially of proteins connected to the membrane fraction.
`A large number of results allow the conclusion that IB formation de-
`pends on the specific folding behaviour rather than on general characteristics
`of a protein such as size, fusion partners, subunit structure and relative
`hydrophobicity (Rudolph 1996). Especially folding-rate-limiting structural
`characteristics such as disulfide bonds result in proteins which easily aggre-
`gate. Also in normally fast folding proteins, mutations which cause a slower
`folding increase aggregation (Rinas et al. 1992; Schulze et al. 1994). Addition-
`ally, also mutations which destabilize the folded state of a protein increase
`aggregation by producing partially unfolded stretches which result in inter-
`molecular interactions (Idicula-Thomas and Balaji 2005). This effect has been
`documented for a number of proteins, such as P22 tailspike protein, single-
`chain antibodies, interferon-γ, colicin A, CheY, immunoglobulin domains
`and interleukin-1β.
`Interestingly, with respect to proteins which form amyloids, mutations
`which suppress amyloid formation also show a lower aggregation potential
`in E. coli and accumulate more as soluble protein (Idicula-Thomas and Bal-
`aji 2005; Wigley et al. 2001). Also vice versa, if a soluble protein is genetically
`modified so that it yields increased amyloid propensity, it accumulates in IBs
`during recombinant expression (Sirangelo et al. 2002; Wigley et al. 2001). De
`novo designed amyloid proteins which displayed amyloid properties in vitro
`also formed IBs in E. coli (West et al. 1999; reviewed by Ventura 2005).
`No sequence or structural similarity is apparent between any of the pro-
`teins which accumulate extracellularly in amyloidoses. However, despite their
`heterogeneity and regardless of their origin, all polypeptides involved in
`amyloid fibrils display similar features, namely (1) binding to various dye
`molecules such as Congo red and thioflavin-T, (2) similar fibrillar morpholo-
`gies and (3) aggregated proteins organized in a cross-β-sheet architecture
`(Carrio et al. 2005; Nilsson 2004) For all of these proteins it is characteris-
`tic that the organized structures formed are highly pure with respect to the
`protein species involved. Interestingly, also other proteins which are unre-
`lated to any known human disease have been found to organize in vitro into
`higher-ordered amyloid-like structures (Fandrich et al. 2001; Guijarro et al.
`1998; Pallares et al. 2004; Ventura et al. 2004); therefore, the ability to form
`these structures was concluded to be a generic property of the polypeptide
`backbone. The specific amino acid sequence determines the propensity of
`aggregation for a specific protein species (Dobson 1999).
`As a consequence of such results, computational tools have been de-
`veloped which calculate the propensity to aggregate for a protein (Fernandez-
`
`Page 5
`
`

`

`242
`
`P. Neubauer et al.
`
`Escamilla et al. 2004; Lopez et al. 2005; Lopez and Serrano 2004; Ventura et al.
`2004) and might be valuable future tools to assess the behaviour of proteins
`also in E. coli.
`Beyond the sequence-based folding characteristics in the cell also the vec-
`torial nature of protein synthesis, the high protein concentration (molecular
`crowding), the presence and action of partner proteins, such as molecular
`chaperones, foldases and proteases, the availability of metal ions and other
`cofactors as well as the potential for transport and posttranslational modifi-
`cation are all factors likely to influence the propensity for protein aggregation.
`
`2.2
`Disulfide Bonds
`
`The formation of the correct disulfide bonds is usually the rate-limiting re-
`action for the formation of the native structure of disulfide bond containing
`proteins under otherwise optimum conditions. Especially in the reducing re-
`dox environment of the bacterial cytosol disulfide bonds form very slowly or
`not at all and therefore disulfide bond containing proteins will easily aggre-
`gate. Mutations which lead to a more oxidizing intracellular milieu have been
`described, such as the knockouts of thioredoxin reductase (trxB) and glu-
`tathione reductase (gor). These mutants promote the formation of disulfide
`bonds and thereby decrease aggregation (Bessette et al. 1999; Derman et al.
`1993; Proba et al. 1995). Although the gor and trxB gene products are central
`in the known major reductive pathways in the E. coli cell the mutations are
`not lethal and suppressor mutants with an oxidized cytosol can be easily se-
`lected. The results from Bessette et al. (1999) with a truncated form of tissue
`plasminogen activator (t-PA) with nine disulfide bonds and coexpression of
`oxidoreductases such as DsbC or other proteins with a thioredoxin fold were
`a breakthrough for the in vivo synthesis of proteins with a high number of
`disulfide bonds. The gor trxB double mutants have been applied in the past
`few years for the production of many other proteins in their native form, in-
`cluding proteins with a lower number of disulfide bonds such as Fab antibody
`fragments, human collagen prolyl 4-hydroxylase, and domains of membrane
`proteins (Cassland et al. 2004; Dutta et al. 2001, 2002; He et al. 2004; Jurado
`et al. 2002; Kersteen et al. 2004; Lauber et al. 2001; Lehmann et al. 2003; Levy
`et al. 2001; Li et al. 2005; Liu et al. 2002; Miertzschke and Greiner-Stoffele
`2003; Neubauer et al. 2005; Premkumar et al. 2003; Schuhmann et al. 2003;
`Shimizu et al. 2005; Venturi et al. 2002; Xiong et al. 2005; Zhao et al. 2003)
`Other disulfide-bond-rich proteins containing more complex structures
`(e.g. disulfide knots) have not been successfully produced in vivo in their
`correct fold in high amounts yet. In vivo production of BMP2 has been
`thoroughly investigated under different conditions and with various fusions.
`Although a soluble product was obtained in vivo as a fusion to maltose bind-
`ing protein by Fahnert (2001) a high yield of active protein was gained only
`
`Page 6
`
`

`

`Protein Inclusion Bodies in Recombinant Bacteria
`
`243
`
`by in vitro renaturation. In the case of another disulfide-knot-containing pro-
`tein, human nerve growth factor (hNGF), the natural prosequence was shown
`to support the folding of the protein and an efficient in vitro refolding process
`from E. coli IBs was established (Rattenholl et al. 2001a, b).
`Proteins consisting of different subunits, such as antibody Fab fragments
`– containing both intermolecular and intramolecular disulfide bonds – have
`been traditionally produced as IBs. Both separate expression of the heavy and
`light chains and coexpression systems have been used and in vitro refolding
`protocols have been developed (Buchner and Rudolph 1991a, b; De Bernardez
`et al. 1999). However, also the production of single-chain antibody fragments
`and Fabs in the E. coli periplasm became a standard strategy (Hayhurst 2000;
`Hayhurst and Georgiou 2001; Hayhurst and Harris 1999; Horn et al. 1996;
`Humphreys et al. 1996, 1997, 2002, 2004; Kujau et al. 1998; Kujau and Riesen-
`berg 1999; Plückthun et al. 1996; Raffai et al. 1999) although the optimization
`process which includes the expression of cofactors is still time-consuming
`and the yields vary depending on the antibody.
`Small proteins with disulfide bonds such as human proinsulin are mainly
`produced as IBs as preprotein, a fusion protein (Nilsson et al. 1996) or by sep-
`arate expression of the different chains. Although the formation of disulfide
`bonds is limited in bacteria, the problems to produce proinsulin do not arise
`from insufficient disulfide bond formation in general, but mainly arise from
`incorrectly formed ones and the low stability of the product (Winter et al.
`2000).
`
`2.3
`Membrane Proteins
`
`Overexpression of membrane proteins has always been a complicated chal-
`lenge. Such proteins aggregate easily, can be even toxic and difficult to express
`owing to their hydrophobic nature. Several E. coli membrane proteins such
`as OmpF, OmpC, PhoE and LamB have been successfully produced in suf-
`ficient amounts (about 80% of the total cellular membrane protein) (Ghosh
`et al. 1998).
`However, the production of heterologous membrane proteins seems to be
`more problematic. A breakthrough came in 1996 when Miroux and Walker
`(1996) succeeded in isolating specific E. coli mutants, allowing the produc-
`tion of membrane proteins. If membrane proteins are overproduced in these
`mutants, new intracellular membranes proliferate containing the recombi-
`nant product in the correct conformation (Arechaga et al. 2000). The authors
`reported the soluble production of the β subunit of F1F0 ATP synthase, con-
`taining one transmembrane span, in the E. coli BL21(DE3) mutants C41 and
`C43. Also Shanklin (2000) expressed an E. coli membrane protein, the acyl–
`acyl carrier protein synthase (80.6 kDa) efficiently in the C41 mutant. In this
`case a smaller amount of this protein (one third) was also correctly integrated
`
`Page 7
`
`

`

`244
`
`P. Neubauer et al.
`
`in membranes of the BL21(DE3) strain. Promising results have also been re-
`ported by others for the expression of active membrane-bound cytochromes
`in these mutants (Mulrooney and Waskell 2000; Saribas et al. 2001) and a few
`other eukaryotic membrane proteins (Chapot et al. 1990; Grisshammer et al.
`1993; Quick and Wright 2002).
`Also IB production with subsequent renaturation of urea-dissolved
`IBs in phosphate buffer containing n-dodecyl-N,N-dimethyl-1-ammonio-3-
`propanesulfonate (SB12) has been successfully applied in a few cases. Jansen
`et al. (2000) produced Neisseria meningitidis PorA using the E. coli BL21 pET
`◦
`C and established an effi-
`system in high amounts as very pure IBs at 37
`cient in vitro refolding protocol for this outer-membrane protein. A few other
`membrane proteins from prokaryotes have been refolded (Charbonnier et al.
`2001; Kumar and Krishnaswamy 2005), but the number of successful exam-
`ples is still low. From larger membrane proteins such as G-protein coupled
`membrane receptors only extracellular binding domains have been expressed
`in E. coli as IBs and refolded (Chauhan et al. 2005; Grauschopf et al. 2000).
`
`2.4
`Glycosylation
`
`Many therapeutically interesting proteins from eukaryotes are glycosylated.
`Glycosylation is known to affect the folding behaviour and especially the
`solubility (Idicula-Thomas and Balaji 2005; Zhang et al. 1998). Therefore, gly-
`cosylated proteins may be prone to aggregation if produced non-glycosylated
`in E. coli.
`Often glycosylation is not a prerequisite for function of the protein but
`it influences its activity and degradation characteristics, such as thermosta-
`bility (Solovicova et al. 1996); therefore, many proteins can be produced in
`bacterial hosts lacking the eukaryotic glycosylation system for therapeutic
`applications. Recombinant proteins produced in E. coli may even have bene-
`ficial properties compared with the glycosylated forms. For example, a non-
`glycosylated recombinant variant of human t-PA obtained by refolding from
`E. coli IBs showed a longer half-life and a lower clearance rate in rats (Martin
`et al. 1992; Mattes 2001).
`
`3S
`
`tructure and Composition of IBs
`
`Being extremely common during the bacterial production of recombinant
`proteins, IBs have been in general poorly studied regarding their structure
`and the mechanics of aggregation of the protein composing them. The poor
`solubility of recombinant proteins and specifically the formation of these de-
`posits of insoluble protein often represent a serious obstacle for the produc-
`
`Page 8
`
`

`

`Protein Inclusion Bodies in Recombinant Bacteria
`
`245
`
`tion of readily usable protein products (Baneyx and Mujacic 2004; Sorensen
`and Mortensen 2005a). Since, in addition, minimization of IB formation is
`not a straightforward task, IB formation has been usually seen as an unavoid-
`able complication associated with the prokaryotic nature of the production
`system itself. Interestingly, IBs are also the source of relatively pure poly-
`peptides, since they are easily separable from cellular debris by low-speed
`centrifugation (Georgiou and Valax 1999) and can be isolated with a reason-
`ably purity by simple detergent-washing procedures (Carrio et al. 2000). This
`has also been exploited for large-scale-addressed production purposes, pro-
`vided a convenient, in vitro refolding procedure had been developed. In this
`case, the enhancement of IB formation can be easily achieved by conditions
`favouring protein aggregation, such as high temperature, high rate of recom-
`binant gene expression and the absence of relevant proteases in the producing
`strain. Again, these strategies have been empirically embarked on by very lit-
`tle, if any, physiological investigation of the biology of the aggregation event
`itself.
`In the context of arising concern of conformational diseases, such as those
`caused by prions or involving amylodgenesis, IBs have been turned into excit-
`ing models for the in vivo study of protein aggregation. Recent insights about
`their structural and physiological traits have transformed the classic picture
`of IBs and provided intriguing clues to understand the mechanics of their
`formation.
`
`3.1
`Morphology, Structure and Molecular Organization
`
`Under optical microscopy, cytoplasmic IBs are usually seen as refractile struc-
`tures, present in low number (generally one or two) per cell, usually with
`a polar distribution, and with a volume often comparable to that of the cell
`cytoplasm itself (Bowden et al. 1991; Carrio et al. 1998). Electron microscopy
`usually pictures IBs as electrodense particles (Carrio and Villaverde 2005),
`generally amorphous, but in some cases showing icosahedra-like shapes.
`Scanning microscopy of purified IBs shows cylinders with a surface tending to
`be rough (Bowden et al. 1991; Carrio et al. 2000), a fact that could be indica-
`tive of a porous structure. On the other hand, the morphological evolution
`of IBs during in vitro trypsin digestion shows that the proteolytic attack is
`not surface-restricted (Carrio et al. 2000). Instead, the enzyme penetrates the
`bodies and in the first phases of the disintegration they are fragmented into
`smaller pieces of protease-resistant protein (Carrio et al. 2000). Thus, IBs ap-
`pear as being formed by multiple cores of protease-resistant polypeptides,
`each of them surrounded by protein species more sensitive to proteases, that
`are initially lost during proteolytic attack, revealing the particulate nature of
`IBs. IB subunits could have an independent origin in the cell, although this
`extent has not been proved experimentally. However, protein synthesis is not
`
`Page 9
`
`

`

`246
`
`P. Neubauer et al.
`
`punctually localized in the bacterial cytoplasm and IBs are expected to result
`from aggregation of protein species produced in distant ribosomal sets. Such
`IB subunits could be aggregation intermediates or independent nucleation
`cores that finally cluster. In this respect, in GroEl140 mutants lacking a func-
`◦
`C, IBs produced at this temperature are smaller
`tional GroEL protein at 42
`and much more abundant than in the wild type (Carrio and Villaverde 2003).
`Although the molecular basis of this observation is not completely solved,
`GroEL could act, either directly or indirectly, by condensing small aggregates
`formed dispersedly in the cell. In this context, GroEL can stimulate in vitro
`protein aggregation as revealed by both E. coli β-galactosidase (Ayling and
`Baneyx 1996) and the prion protein models (De Burman et al. 1997; Stockel
`and Hartl 2001).
`The in vitro digestion analysis of IBs renders two additional observations
`relevant to the molecular organization of IB proteins. Firstly, the relative com-
`position of protease-resistant and protease-sensitive species changes during
`IB formation, indicating inner molecular reorganization and therefore both
`structural and evolving flexibility of these aggregates (Carrio et al. 2000). Sec-
`ondly, protein digestion occurs as a cascade process, in which protease target
`sites are sequentially exposed for cleavage (Cubarsi et al. 2001; Cubarsi et al.
`2005). Interestingly, the sequential cleavage takes place on protein that re-
`mains associated with IBs during their progressive fragmentation. This fact
`indicates a packaging scheme loose enough to allow for aggregated polypep-
`tides to undergo conformational modifications while being IB components.
`Fourier transform IR (FTIR) analysis and other structural approaches have
`revealed an increase in the β-sheet structure of IB protein relative to that of
`the soluble version (Ami et al. 2003; Przybycien et al. 1994), which adopts
`an antiparallel, intermolecular organization (Carrio et al. 2005). In addition,
`many amyloid-like traits observed in IBs such as binding of amyloid-tropic
`dyes or a sequence-dependent seeding process (Carrio et al. 2005) suggest
`that IB proteins are not completely disorganized but are packaged in a rather
`regular way. The recent finding of recombinant proteins forming amyloid-like
`fibres in E. coli and occurring as loose aggregates in the soluble cell fraction
`(de Marco and Schroedel 2005) prompts us to speculate that such structures
`could be among the IB precursors.
`Interestingly, a certain extent of native-like secondary structure has been
`also observed by FTIR analysis in IBs formed by different proteins (Ami
`et al. 2005; Oberg et al. 1994; Umetsu et al. 2005), coexisting with the dom-
`inating characteristic β-sheet. In this context, enzymatic activity associated
`with enzyme-based IBs that could be compatible with the occurrence of at
`least a fraction of native protein has also been reported (Garcia-Fruitos et al.
`2005a; Kuczynska-Wisnik et al. 2004; Tokatlidis et al. 1991; Worrall and Goss
`1989).
`Such active protein species, which are probably trapped by unfolded do-
`mains not affecting the active site, are able to confer at least some IB-cata-
`
`Page 10
`
`

`

`Protein Inclusion Bodies in Recombinant Bacteria
`
`247
`
`lysing properties that could prompt us to consider the use of IBs as catalysers
`in enzymatic processes (Garcia-Fruitos et al. 2005b). The cultivation condi-
`tions can have an influence on the structure of the target protein which is
`entrapped in the IBs. A highly interesting strategy was published recently by
`Jevsevar et al. (2005), who developed a strategy for a bioprocess for human
`G-CSF on the basis of IBs which had a close to native state like structure. The
`accumulation of this kind of “nonclassical” IB was promoted by slow expres-
`sion and lower temperature. This new way of optimizing the folding state of
`the target protein in the IBs resulted in a simpler and very efficient down-
`stream purification procedure without the need to use high guanidinium
`chloride (GdmCl) or urea concentrations.
`The molecular basis for the coexistence of active, properly folded protein
`species and cross-β-sheet-organized polypeptides and the possible potential
`of IBs which contain the target product in a close-to-native conformation
`needs further investigation.
`
`3.2
`Compositional Analysis
`
`Early proteomic analyses revealed that the recombinant protein itself is the
`main component of IBs (Rinas et al. 1993; Rinas and Bailey 1992; Valax and
`Georgiou 1993), representing up to 95% of the total embedded protein (Car-
`rio and Villaverde 2002; Villaverde and Carrio 2003). Interestingly, in the case
`of proteolytically unstable IB-forming proteins, stable digestion fragments
`are also abundant in the aggregates (Corchero et al. 1996). An important part
`of those fragments could be generated from in situ proteolysis (Carbonell
`et al. 2002; Corchero et al. 1997; Vera et al. 2005), while others can result
`from site-limited protein digestion in the soluble fraction and further depo-
`sition (Carrio et al. 1999). Many cellular proteins, including some membrane
`proteins belonging to the Omp family, are also found associated with IBs (Jür-
`gen et al. 2000; Rinas and Bailey 1992). Most of these polypeptides are mere
`contaminants retained by unspecific attachment during the purification pro-
`cess (Georgiou and Valax 1999). Traces of cell material such as phospholipids
`and nucleic acids have also been identified (Valax and Georgiou 1993). The
`small heat shock proteins IbpA and IbpB seem to be a common compon-
`ent of protein aggregates in E. coli (Allen et al. 1992; Carrio and Villaverde
`2002; Hoffmann and Rinas 2000). These chaperones play an important and
`immediate role in mediating the release of IB proteins (Carrio and Villaverde
`2003), especially at high culture temperatures (Lethanh et al. 2005). Minor
`amounts of main chaperones such as DnaK and GroEL are also occasion-
`ally observed as IB components (Carrio and Villaverde 2002; Jürgen et al.
`2000), and they might also have a functional role in IB processing. In particu-
`lar, the surface-restricted localization of DnaK (Carrio and Villaverde 2005)
`indicates that this chaperone is not passively trapped by interaction with ag-
`
`Page 11
`
`