Biological role of bacterial inclusion bodies: a model for amyloid aggregation

Authors

  • Elena García-Fruitós,

    1.  Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Spain
    2.  Department of Genetics and Microbiology, Universitat Autònoma de Barcelona, Spain
    3.  CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Barcelona, Spain
    Search for more papers by this author
    • These authors contributed equally to this work

  • Raimon Sabate,

    1.  Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Spain
    2.  Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Spain
    Search for more papers by this author
    • These authors contributed equally to this work

  • Natalia S. de Groot,

    1.  Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Spain
    2.  Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Spain
    Search for more papers by this author
  • Antonio Villaverde,

    1.  Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Spain
    2.  Department of Genetics and Microbiology, Universitat Autònoma de Barcelona, Spain
    3.  CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Barcelona, Spain
    Search for more papers by this author
  • Salvador Ventura

    1.  Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Spain
    2.  Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Spain
    Search for more papers by this author

A. Villaverde, Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Bellaterra, 08193 Barcelona, Spain
Fax: +34 93 581 2011
Tel: +34 93 581 2148
E-mail: Antoni.Villaverde@uab.cat

S. Ventura, Institute for Biotechnology and Biomedicine, Universitat Autònoma de Barcelona, Bellaterra, 08193 Barcelona, Spain
Fax: +34 93 581 2011
Tel: +34 93 586 8956
E-mail: salvador.ventura@uab.es

Abstract

Inclusion bodies are insoluble protein aggregates usually found in recombinant bacteria when they are forced to produce heterologous protein species. These particles are formed by polypeptides that cross-interact through sterospecific contacts and that are steadily deposited in either the cell’s cytoplasm or the periplasm. An important fraction of eukaryotic proteins form inclusion bodies in bacteria, which has posed major problems in the development of the biotechnology industry. Over the last decade, the fine dissection of the quality control system in bacteria and the recognition of the amyloid-like architecture of inclusion bodies have provided dramatic insights on the dynamic biology of these aggregates. We discuss here the relevant aspects, in the interface between cell physiology and structural biology, which make inclusion bodies unique models for the study of protein aggregation, amyloid formation and prion biology in a physiologically relevant background.

Abbreviations
IB

inclusion body

PFD

prion forming domain

Biotechnology of bacterial inclusion bodies; a historical view

Production of recombinant proteins in microorganisms, powered in the late 1970s by the identification of restriction enzymes, has provided much fewer products than initially expected [1]. The ready-to-use concept of recombinant DNA technologies has proved to be unrealistic and has faced severe obstacles associated with the physiology of the host microorganism. This is because the cellular protein factories are usually forced to produce heterologous polypeptides, encoded in a multicopy expression plasmid, over physiological rates – a combination of facts that tend to saturate the protein synthesis machinery and activate the quality control system. Essentially, protein production processes in bacteria (as well as in other microorganisms) suffer from protein degradation and lack of solubility and, to a minor extent, toxicity exerted by the product on the cells and consequent genetic instability (including plasmid loss). These events occur in the context of several cell stress responses, which depending on the nature of the host microorganism include triggering of oxidative stress, the unfolded protein response, the heat shock and the stringent response and the activation of the DNA repair SOS system [2].

Traditionally, lack of solubility and the formation of inclusion bodies (IBs), large insoluble clusters enriched by misfolded versions of the recombinant protein species [3], have been the main obstacle for the smooth consecution of production processes, aimed at high yields of soluble, biologically active species. Believed to be irreversibly formed and containing inactive proteins, how to minimize IB formation in midstream has been a matter of extensive discussion. Essentially, reducing the growth temperature, lowering the transcription rate and co-producing folding modulators selected from the quality control system have been thoroughly explored strategies [4]. Also, at the downstream level, refolding of IB proteins has also been approached [5]. Both midstream- and downstream-focused approaches have been successful for an important number of specific proteins but they do not offer generic solutions to the lack of solubility in protein production.

Despite the economical relevance of IB formation for both catalysis and biotech industries, IBs have been in general poorly characterized. Consequently, the discovery of the reversibility of IB formation [6], the general acceptance of IBs being formed by functional proteins [7] and the recognition of the amyloid-like architecture of IB proteins [8] have represented dramatic insights in the biology of these structures that has favoured important advances in the comprehension of their physiological and structural nature. For instance, the conceptual unlinking between solubility and functional quality [9], and the fact that enhanced protein yields result in lower quality protein species [10,11], has permitted IBs to be explored as powerful biocatalysts (the embedded proteins acting as immobilized enzymes) [12,13]. On the other hand, the fine and timely analysis of the amyloid architecture of IB proteins [14,15] has led to the use of these underestimated bacterial aggregates as intriguing models for the analysis of protein–protein interactions in the context of amyloid and prion diseases.

Dynamics of IB formation and biological activity

Intracellular electrodense proteinaceous granules had been observed in classical experiments when bacteria were cultured in the presence of non-natural amino acids. This observation, which indicated the transient nature of protein aggregates formed by conformationally aberrant proteins, was more recently repeated with bacterial IBs [6], so far believed to be irreversible protein clusters averse to in vivo protein refolding [16]. This is indicative of cellular activities acting on these protein aggregates, including release to the soluble cell fraction but also proteolytic events [6,17,18] that might promote degradation of IB proteins in situ [19]. In addition, for protein species that are found in both soluble and insoluble cell fractions, the conformational quality and biological activity of IB embedded proteins evolve in parallel with those of the soluble counterparts, under different environmental conditions affecting folding, such as temperature and chaperone availability [20,21]. Therefore, IB proteins appear not to be excluded from quality control [22], in which a complex network of chaperones and proteases survey the folding status of cellular proteins [23], soluble but also insoluble.

In agreement with this concept, the main Escherichia coli chaperone DnaK (a holding agent and a foldase and disaggregase), is almost exclusively found, in IB-producing bacteria, attached at the IB surface, while the foldase GroEL is present within the IB core [24]. DnaK, which participates in the in vivo refolding of bacterial thermal aggregates [25,26], appears to be highly active on bacterial IBs [20,27,28]. In fact, we have recently shown that the chaperone DnaK promotes protein extraction from bacterial IBs but that this event is intimately associated with proteolysis [10,11]. This explains the reduction of protein yield eventually observed during co-production of this chaperone and others [10], as a side effect of this strategy [29] addressed to improve the solubility of recombinant proteins. Interestingly, the specific dependence of the DnaK-mediated stimulation on bacterial chaperones makes this chaperone very useful for co-production in eukaryotic systems [30].

The simultaneous surveillance of soluble and IB protein species by bacterial chaperones and proteases indicates the occurrence of similar targets in both protein versions and strongly suggests a highly dynamic transition between the two forms. In fact, aggregation and disaggregation seem to be simultaneous events in actively producing recombinant bacteria [16], while disaggregation will be highly favoured in the absence of protein synthesis [6]. Such a bidirectional protein transit between the cells’ virtual fractions (soluble and insoluble [22]) accounts for the unexpected and recently determined abundance of soluble aggregates in recombinant cells [31]. These particles, either globular or fibrillar, might be intermediates in the in/out IB protein transition, or just members of the conformational spectrum that recombinant proteins can adopt in host bacteria, irrespective of whether they are found in soluble or insoluble cell fractions. Interestingly, increasing evidence supports the presence of biologically active proteins embedded in IBs, indicating that both folded and misfolded polypeptides coexist in these proteinaceous aggregates [32]. Regarding the presence of functional protein in such aggregates, different enzyme-based IBs have been successfully tested as catalysts of different bioprocesses [33]. Galactosidases [7,34], reductases [7], oxidases [35], kinases [36], phosphorylases [37] and aldolases [38] are just some examples of the enzymes used in aggregated form to catalyse specific reactions, opening a promising market in the biotechnological industry [33]. In this context, other authors have also described the use of IBs for the intracellular capture of a co-synthesized target enzyme, obtaining IB particles with the enzyme of interest immobilized in their surface [39].

Stereospecific interactions in protein aggregation

Chiti and coworkers pointed out that the intrinsic physicochemical properties of an amino acid sequence, such as hydrophobicity, secondary structure propensity and charge, can determine the aggregation behaviour of a given polypeptide [40,41]. Many examples support the correlation between protein aggregation tendency and amino acid sequence, and it is also possible to identify the aggregation-prone regions of polypeptides using software such as aggrescan [42] or tango [43]. Protein aggregation can be understood as an anomalous type of protein–protein interaction. As for native interactions, the attainment of ordered aggregated structures requires the establishment of stereospecific intermolecular contacts. Accordingly, it has been observed that both bacterial [8] and mammalian protein aggregates are formed through a conserved, selective and sequence-specific process. Specificity during protein aggregation is best exemplified by the nucleation-driven polymerization of proteins into amyloid aggregates [44], a mechanism reminiscent of that occurring in crystallization processes [45]. Mature amyloid fibrils possess the faculty to accelerate the formation of new fibrils by acting as a nucleus that seeds the growth of fibrillar structures [46]. However, molecular recognition between aggregated and soluble proteins only occurs when they share a high sequence similarity. The requirement for stereospecific interactions during protein aggregation would explain why disease-linked amyloid deposits are composed almost exclusively of the pathogenic protein [47] and bacterial IBs are highly enriched in the target recombinant protein [22]. The distribution of side chains in the sequence, such as occurs in protein folding, plays a pivotal role in determining the conformational properties of the aggregated state and the way in which this supramolecular ensemble is reached from the initial soluble state. This control is so exquisite that a protein and its backward version (a protein with exactly the same succession of side chains but with a reverted backbone) do not cross seed each other and form aggregates displaying different conformational and functional properties [48]. However, apart from the primary sequence, the particular structural and thermodynamic properties of proteins modulate their deposition in physiologically relevant conditions, making it difficult to predict the effective aggregation propensities of polypeptides in cellular environments.

Protein aggregation into amyloid structures

Protein aggregation can occur from multiple structural conformations such as intrinsically disordered polypeptides, oligomeric species or globular proteins [47,49]. The macromolecular assemblies formed by these proteins are all sustained by intermolecular interactions but their arrangement and specificity define the degree of order in the structure of the final aggregate. The energy landscape of protein aggregation is rough and complex, comprising both highly energetic amorphous deposits and well-ordered amyloid fibrils of lower energy than the native structure of the protein [50,51]. Amorphous aggregates can be formed rapidly by simple precipitation of the protein, whereas ordered fibrillation requires specific intermolecular contacts, the formation of which is strongly influenced by the protein local environment [47,52].

The number of identified amyloid-forming proteins increases each year. These fibrillar structures were initially discovered in human tissues of patients suffering from amyloidoses such as Alzheimer’s or Parkinson’s diseases. The study of these deposits has shown that mature fibrils can be less cytotoxic than the intermediary forms in the aggregation pathway suggesting that the amyloid structure might play in fact a protective function [47,53]. Importantly, amyloid conformations are not only associated with pathological conditions but are also exploited by Nature to execute important regulatory, structural and genetic functions [54,55]. In fact, the ability to form amyloid assemblies has been suggested to be a generic protein property [47,56] and, as we shall see in the next sections, a conformation accessible to structurally and sequentially unrelated proteins upon recombinant expression [51].

Despite their diverse origin, all amyloid structures share common morphological characteristics: straight unbranched fibrils 7–12 nm in diameter made up of two to six protofilaments 2–5 nm in diameter with a cross-β-sheet spine [47,57] in which each polypeptide chain is structured into β-strands and each β-strand is arranged perpendicular to the long axis of the fibril. This arrangement allows a tightly packed quaternary structure sustained mainly by generic hydrogen bonds and hydrophobic contacts [58], explaining why, in spite of the high sequential specificity driving amyloid formation pathways, any sequence able to be accommodated in a β-sheet conformation can, potentially, reach the amyloid state [51,56].

Amyloid-like properties of bacterial IBs

The architecture and mechanisms of IB formation in bacteria have remained unexplored for years. However, important insights in this field have lately emerged. Although IBs were conventionally described as disordered aggregates being formed by non-specific interactions of exposed hydrophobic surfaces, an increasing amount of evidence is showing that in fact IBs are highly ordered protein deposits formed through a process similar to that observed during amyloid deposition [8,14]. Just as occurs for amyloids, IB formation is driven by intermolecular interactions occurring through homologous protein patches in a nucleation-dependent manner (Figure 1) [8]. On the one hand, a study published by Carrió and coworkers demonstrates that target recombinant protein aggregation in vitro is a tightly regulated phenomenon, and recombinant proteins preferentially associate with themselves rather than with other proteins in the environment in a dose-dependent way [8]. On the other hand, an in vivo study performed using fluorescence resonance energy transfer shows that, when co-producing two different recombinant proteins in the complex bacterial cytoplasmic environment, the distribution of the two proteins in the formed IBs is also tightly regulated through specific contacts, each protein being specifically localized in a different region of the aggregate depending on its sequence. Therefore, it is not surprising that, in spite of the IBs’ amorphous macroscopic appearance, recently different groups have converged to demonstrate unequivocally the effective existence of amyloid-like structures inside bacterial aggregates [14,59]. Accordingly, relative to the native conformation, proteins embedded in IBs appear to be enriched in β-sheet secondary structure elements displaying the minimum at 217 nm characteristic of this conformation in the far-UV circular dichroism spectra (which can be displaced slightly to higher wavelengths due to the stacking of aromatic residues) as well as a band at 1620–1630 cm−1 in the infrared spectra, typical of the tightly bound intermolecular β-strands in amyloid structures [8,15,59–61], and X-ray diffraction patterns with meridional (4.8 Å) and equatorial (10–11 Å) reflections compatible with the presence of a cross-β structure [59]. In addition, amyloid-specific dyes like Congo Red or thioflavin-T and S bind to bacterial IBs with similar affinity to the affinity they exhibit for amyloid structures [8,15,59–61], confirming a high degree of conformational similarity between the two types of aggregates. As in amyloid fibrils, IBs display regions with high resistance against proteolytic attack, probably corresponding to a preferentially protected β-sheet core. The presence of fibrillar structures with amyloid-like morphology in IBs has been observed directly or after controlled proteolytic digestion by transmission electronic microscopy, cryo-electron microscopy [59,62] and atomic force microscopy [15,60,61]. In addition, IBs formed by amyloid proteins display the capacity to seed and accelerate in a highly specific manner the formation of amyloid structures by their soluble and monomeric forms [15,60–62] (see Figure 2).

Figure 1.

 A nucleation/polymerization self-assembly process drives the formation of IBs in bacteria. (A) In vivo formation of IBs in recombinant bacteria. Aggregation-prone versions of the recombinant protein (green) slowly form seeding nuclei by cross-molecular stereospecific interactions. These proto-aggregates recruit further copies of the target protein, in a process compatible with first-order kinetics. This promotes a fast mass growth of IBs. Non-homologous cellular or recombinant proteins (red, black) are excluded from these seeding events. (B) Kinetics of aggregation monitored by time-dependent increase of turbidity at 350 nm using soluble VP1LAC and LACZ incubated with different IBs (figure modified from [8]). (C) Kinetics of Aβ42 peptide seeding monitored through thioflavin-T fluorescence emission (figure taken from [58]). All figures have been reproduced with permission.

Figure 2.

 Presence of amyloid-like structures in the IBs formed by the prion protein HET-s from P. anserina. (A) HET-s PFD IBs from E. coli observed by cryo-electron microscopy in intact E. coli cells. (B) Transmission electron micrograph of negatively stained purified HET-s PFD IBs. (C), (D) HET-s IB structure before (C) and after (D) 30 min of proteinase K digestion monitored by transmission electron microscopy, showing the apparition of fibrillar structures. (E)–(H) Congo Red (CR) binding to different HET-s IBs monitored by UV/vis spectroscopy and staining and birefringence under cross-polarized light using an optical microscope: (E), (F) CR spectral changes in the presence of different HET-s IBs; (E) changes in λmax and intensity in CR spectra in the presence of HET-s IBs; (F) difference absorbance spectra of CR in the presence and absence of IBs showing in all cases the characteristic amyloid band at ∼ 540 nm; (G) HET-s PFD IBs stained with CR and observed at 40× magnification and (H) the same field observed between crossed polarizers displaying the green birefringence characteristic of amyloid structures. (I) 13C–13C solid-state NMR correlation spectrum (proton-driven spin-diffusion with a mixing time of 50 ms) of purified HET-s PFD IBs (blue) compared with a spectrum of in vitro HET-s PFD amyloid fibrils (red) recorded under identical conditions. All the signals assigned for the purified fibrils were also observed in the spectrum of the IBs. The insets demonstrate that no significant changes in the chemical shifts appear and that the linewidths of the two samples are virtually identical. The individual spectra were recorded at a 1H frequency of 600 MHz (static field B0 = 14.9 T), 10 kHz magic angle spinning. (J)–(L) Secondary structure of HET-s PFD IBs: (J) CD spectra, and (K), (L) FTIR absorbance and second derivative spectra in the amide I region of HET-s PFD spectra showing the characteristic spectral bands of β-sheet conformations. (M) Seeding-dependent maturation of HET-s PFD amyloid growth. The aggregation reaction was seeded with HET-s full length, HET-s (157–289), HET-s PFD, Aβ40 or Aβ42 IBs. The fibrillar fraction of HET-s PFD is represented as a function of time. The formation of HET-s PFD amyloid fibrils is accelerated only in the presence of HET-s IBs. (A), (B) and (I) adapted, with permission, from [60]; (C)–(H) and (J)–(M) adapted, with permission, from [15].

Aggregated structures are non-crystalline and insoluble and are therefore not amenable to X-ray crystallography and solution NMR, the classical tools of structural biology, making it difficult to characterize the fine structure of these assemblies at the residue level, even when they display a high degree of internal order [63]. Quenched hydrogen/deuterium exchange with solution NMR allows the identification of solvent-protected backbone amide protons involved in hydrogen bonds. Interestingly enough, three recent studies using this approach to study the IBs formed by different protein models convincingly demonstrate the presence of sequence-specific motifs displaying protection compatible with a cross-β conformation [59,61,62]. High resolution information on the conformation of proteins in the aggregated state can be obtained by solid-state NMR, a technique that has allowed amyloid fibrils to be modelled at atomic resolution [64]. Two recent works have exploited solid-state NMR to address the fine structure of the IBs formed by two amyloidogenic proteins, the HET-s prion forming domain (PFD) of the fungus Podospora anserina and the Alzheimer’s amyloid β peptide (Aβ). The comparison between the signals of the in vitro formed amyloid fibrils and the corresponding IBs indicates the existence of regions with highly similar structural disposition in these aggregates, in particular in the case of HET-s PFD where the NMR signals of the two types of aggregates overlap significantly [61,62,65]. Overall, it appears that the formation of amyloid-like assemblies is an omnipresent process in both eukaryotic and prokaryotic cells.

Infectious conformations in bacterial IBs

Prions represent a particular subclass of amyloids in which the aggregation process becomes self-perpetuating in vivo and thus infectious [14]. The possibility that the bacterial IBs formed by recombinant prion proteins could display infectious properties has important implications. On the one hand, bacteria might become a simple and tunable in vivo system to study the determinants of prion formation. On the other hand, bacterial IBs would be an ideal system for the production of significant amounts of infectious proteins ready to use for cell biology studies, without the requirement of the highly inefficient in vitro unfolding/refolding and controlled aggregation procedures necessary to obtain proteins in transmissible conformations. Therefore, the infectious capacity of prion proteins deposited in bacteria during recombinant production is receiving increasing attention. Meier and co-workers have tested the ability of HET-s PFD IBs purified from E. coli to infect strains of its natural host, P. anserina, using different protein transfection methods [62]. Strains transfected with HET-s PFD IBs acquired the [Het-s] prion phenotype at a frequency comparable with that obtained with HET-s PFD infectious fibrils assembled in vitro, confirming that bacterial HET-s PFD IBs display a high prion infectivity [62]. In contrast, the IBs of a heterologous amyloid protein were not infectious. The yeast prion protein Sup35 has also been shown recently to access an infectious structure when produced in E. coli cells [66]. These two independent observations confirm that the content of the bacterial cytoplasm can support the formation of infectious conformations and suggest that bacterial aggregation might become a generic model system to understand prion biology.

Bacteria as model systems to study protein aggregation

In addition to being the default protein production cellular factories, bacterial cells are valuable systems to understand the integration of metabolic, regulatory and structural features in living cells. The similarities between bacterial aggregates and the deposits formed in higher organisms in pathological processes like amyloid fibrils, nuclear inclusions and aggresomes [67,68] provide a unique opportunity to dissect the molecular pathways triggering these disorders in a simple, yet physiologically relevant, organism. Accordingly, E. coli has been used to study the link between protein aggregation and ageing [69], the role of the highly conserved protein quality machinery on the conformational properties of aggregated states [20,67], the effect of the protein sequence on in vivo aggregation kinetics [41], the influence of extrinsic factors like temperature on protein aggregation properties [21,70] or the control of polypeptide solubility in biological environments by the thermodynamic [71] and kinetic stability of proteins [72]. In addition, the possibility of labelling aggregation-prone proteins with natural [41] or artificial fluorophores [73] allows in vivo deposition pathways to be tracked in real time and compounds able to block the self-assembly process to be identified [74]. Finally, bacteria provide a means to trap and study the highly toxic, unstable and transient intermediates in the fibrillation reaction, illuminating one of the more obscure but crucial steps in amyloid fibril formation [61].

Acknowledgements

We appreciate the financial support from MICINN (BFU2010-17450 and BFU2010-14901), AGAUR (2009SGR-00108 and 2009SGR-00760) and CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN, Spain), an initiative funded by the VI National R&D&i Plan 2008–2011, Iniciativa Ingenio 2010, Consolider Program, CIBER Actions and financed by the Instituto de Salud Carlos III with assistance from the European Regional Development Fund. A.V. and S.V. have been distinguished with an ICREA Academia award.

Ancillary