A DNA-promoted amyloid proteinopathy in Escherichia coli


E-mail rgiraldo@cib.csic.es; Tel. (+34) 91 8373112; Fax (+34) 91 5360432.


Protein amyloids arise from the conformational conversion and assembly of a soluble protein into fibrilar aggregates with a crossed β-sheet backbone. Amyloid aggregates are able to replicate by acting as a template for the structural transformation and accretion of further protein molecules. In physicochemical terms, amyloids arguably constitute the simplest self-replicative macromolecular assemblies. Similarly to the mammalian proteins PrP and α-synuclein, the winged-helix dimerization (WH1) domain of the bacterial, plasmid-encoded protein RepA can assemble into amyloid fibres upon binding to DNA in vitro. Here we report that a hyper-amyloidogenic functional variant (A31V) of RepA, fused to a red fluorescent protein, causes an amyloid proteinopathy in Escherichia coli with the following features: (i) in the presence of multiple copies of the specific DNA sequence opsp, WH1(A31V) accumulates as cytoplasmatic inclusions segregated from the nucleoid; (ii) such aggregates are amyloid in nature; (iii) bacteria carrying the amyloid inclusions age, exhibiting a fivefold expanded generation time; (iv) before cytokinesis, small inclusions are assembled de novo and transferred to the daughter cells, in which transmission failures cure amyloidosis; and (v) in the absence of inducer DNA, purified cellular WH1(A31V) inclusions seed amyloid fibre growth in vitro from the soluble protein. RepA-WH1 is a suitable bacterial model system for amyloid proteinopathies.


Protein amyloids, aggregated conformational variants of an otherwise soluble protein, are built as cross-β structures. These are fibrilar β-sheet assemblies in which strands donated by individual protein molecules are arranged perpendicular to the axis of the fibre (Chiti and Dobson, 2006; Sawaya et al., 2007). Protein amyloids are the causative agents of a large number of human neurodegenerative and systemic diseases, and are therefore the subject of intense biomedical research. Protein amyloidosis can be propagated by fibre fragmentation, followed by vertical spread of the generated seeds to progeny cells, where they act as a template for the conformational conversion of soluble protein molecules and nucleate their accretion into growing fibres (Knowles et al., 2009; Wiltzius et al., 2009). Prions are a subset of amyloid proteins with the distinctive ability to be horizontally spread by cell-to-cell infection (Krishnan and Lindquist, 2005; Cobb and Surewicz, 2009). Bacteria, while able to secrete functional extracellular protein amyloid aggregates (Otzen and Nielsen, 2008), do not naturally form intracellular amyloids. An exception to this rule is inclusion bodies, produced upon overexpression of aggregation-prone recombinant proteins, which exhibit some amyloid features (Morell et al., 2008; Wang et al., 2008; De Groot et al., 2009; Wasmer et al., 2009). However, an increasing number of proteins, many of bacterial origin, are being identified that can be forced to aggregate as amyloids in vitro through the induction of local unfolding under non-physiological conditions (Chiti and Dobson, 2006). Yeast prions such as Sup35p/[PSI+] and Ure2p/[URE3] have provided a wealth of valuable information on protein amyloidosis and its transmissibility (Wickner et al., 2007; Tessier and Lindquist, 2009). To date no endogenous prions have been identified in bacteria. However, bacteria are believed to posses the potential to support prion replication, since yeast [PSI+] can be heterologously propagated in Escherichia coli (Garrity et al., 2010).

RepA protein is encoded within the replicon module of the Pseudomonas plasmid pPS10 and homologues of this protein are found in many plasmids in Gram-negative bacteria (Giraldo and Fernández-Tresguerres, 2004). RepA binds as a dimer to an inversely repeated operator DNA sequence to repress its own transcription. Upon binding to directly repeated DNA sequences found at the plasmid replication origin, RepA dissociates into monomers to initiate DNA replication. Similarly to the mammalian prion protein PrP (Cordeiro et al., 2001; Nandi et al., 2002; Deleault et al., 2003; 2007; Geoghegan et al., 2007; Silva et al., 2008; Wang et al., 2010) and α-synuclein (Cherny et al., 2004; Hegde et al., 2010), the RepA winged-helix dimerization domain (WH1) undergoes a large α-helix to β-sheet conformational transition upon sequence-specific binding to double-stranded (ds) DNA (Díaz-López et al., 2003; 2006; Giraldo et al., 2003). In combination with a single point mutation in RepA (A31V), which expanded the replication host-range of pPS10 and duplicated the plasmid copy-number in vivo (Fernández-Tresguerres et al., 1995; Gasset-Rosa et al., 2008a), low-affinity and transient binding of short dsDNA sequences in vitro enhanced the conformational conversion of WH1, promoting protein amyloidosis (Giraldo, 2007). Thus, DNA is not a major structural component of the resulting amyloid fibres but acts as a real inducer of WH1(A31V) amyloidosis under quasi-physiological conditions (Giraldo, 2007; Gasset-Rosa et al., 2008b).

To challenge our understanding of the molecular basis of RepA-WH1 amyloidosis in vitro, we successfully engineered amyloid assemblies in E. coli using this protein as a chassis. We found that a fusion of RepA-WH1(A31V) to a monomeric red fluorescent protein forms intracellular aggregated amyloid inclusions that are transmitted to the bacterial progeny by seeding and self-templating, causing a marked decrease in cell fitness. The RepA-WH1 aggregated inclusions appear to be the causative agent of a synthetic amyloid proteinopathy in bacteria.


DNA-promoted amyloidosis of RepA-WH1 in E. coli

The DNA sequences coding for either the wild-type (WT) RepA-WH1 or its enhanced amyloidogenic, functional mutant variant A31V (Giraldo, 2007) were cloned as 5′ fusions to mCherry (mRFP), a monomeric DsRed fluorescent protein (Shanner et al., 2004), in a multi-copy, IPTG-inducible expression vector. A parallel vector series was constructed carrying 18 tandem repeats of the 11 bp DNA sequence opsp, which specifically promotes WH1 amyloidosis in vitro (Giraldo, 2007; Gasset-Rosa et al., 2008b), cloned immediately downstream of the WH1–mRFP fusions. The aim of this experimental design was to facilitate the coupling between protein biosynthesis and binding to adjacent target DNA sequences in the diffusion-limited, crowded, bacterial cytoplasm (Morris and Jensen, 2008; Montero-Llopis et al., 2010) (Fig. 1A). When these plasmids were transformed into E. coli MC4100 cells (a commonly used K-12 strain of this bacterium), twofold higher transformation efficiencies were observed for the WT than for the A31V constructs. Moreover, all of the WH1(WT) colonies exhibited an intense red-purple colour, whereas 10% of those carrying the WH1(A31V) variant were white and larger in size, while the remaining 90% were of a similar size to the WT colonies but were pale red in color (Fig. 1B). Similar results were obtained with other E. coli K-12 strains. When the plasmid contents of several colonies of each type were analysed (Fig. 1C), it became evident that the purified expression vectors from the red colonies were essentially the same, apart from the WT or A31V inserts. However, plasmids isolated from the white colonies exhibited increased copy-numbers (by 2.5- to 3-fold) and sizes. Restriction analysis showed that white colonies were carrying plasmids over 1 kbp larger than expected. DNA sequencing revealed, in every plasmid analysed, the insertion of an IS10 mobile genetic element (De Palmenaer et al., 2008) interrupting the WH1(A31V)–mRFP coding sequence (Fig. S1). Insertion of this genetic element knocked out the expression of the fusion protein, as confirmed by Western blotting (Fig. 1D). These findings reflected the remarkable genome plasticity in response to stress and were the first indication of the differential toxicity linked to the amyloidogenic A31V variant of WH1–mRFP (see next section). Therefore, in all subsequent experiments single red colonies were used to inoculate culture medium.

Figure 1.

Expression of WH1–mRFPs in E. coli MC4100 cells.
A. Linear sketch of the plasmids used for WH1–mRFP overexpression under the tac promoter. WT or A31V (*) WH1 variants, and the cis (opsp)18 repeats are indicated.
B. Images of LB-ampicillin agar plates showing bacterial colonies transformed with plasmids pWH1(WT)–mRFP (left) and pWH1(A31V)–mRFP (right).
C. Plasmids purified from equivalent cell numbers in three independent IPTG-induced cultures (labelled 1–3) were quantified by UV light absorption (left), analysed by agarose gel electrophoresis of their supercoiled forms (1/15 of each DNA preparation; right) and digested with HindIII (bottom), a single cutter restriction enzyme that renders a defined banding pattern for each plasmid. Asterisks mark the restriction-resolved plasmid multimers from the cultures inoculated with white colonies (see B). Also, note their sizes, > 1 kbp larger than the parental pWH1(A31V)–mRFP, and the pLacI band visible just below.
D. SDS-PAGE analysis of total lysates (5 × 108 cells lane−1) from un-induced (t = 0 h) and IPTG-induced (t = 2 h) cultures. Western-blots performed with a monoclonal antibody against the His6-tag in WH1–mRFPs are aligned below. Different exposure times were required for detection of the basal (pre-induction) and post-induction expression levels.

Fluorescence microscopy analysis of the bacteria transformed with WH1(WT)–mRFP revealed that IPTG addition resulted in the emission of red fluorescence, with variable intensities, diffused across the whole cytoplasm, in about 25% of the cells (Fig. 2A). The bi-modal, all-or-nothing, expression levels achieved by the individual cells could be due to the stochastic variability intrinsic to transcription and translation (Davidson and Surette, 2008; Robert et al., 2010). Expression of the WH1(A31V)–mRFP variant led to the appearance of a couple of intense fluorescent foci, commonly located around one-quarter and three-quarters of the way along individual cells, in 17% of the bacteria. These foci resembled inclusion body protein aggregates (García-Fruitós et al., 2005). DAPI staining showed that these WH1(A31V)–mRFP inclusions were segregated from the bacterial nucleoid. However, when extra opsp repeats were cloned in cis, allowing for translation-coupled binding of WH1 to DNA with the multiplier effect of the high copy number (40–250 per cell) of the vector, a further 40% increase in the number of cells with fluorescence labelling was achieved (Fig. 2A, right). In addition, the foci observed in the presence of additional copies of the opsp repeats were larger than those observed in their absence (0.93 ± 0.15 µm2 versus 0.52 ± 0.08 µm2 respectively), but the fluorescence intensities, when corrected for the size of the inclusions, were similar (0.89 ± 0.26 versus 0.85 ± 0.20 respectively). In all of the subsequent in vivo experiments, the opsp DNA repeats were included to enhance the accretion of WH1(A31V) molecules into the inclusions.

Figure 2.

RepA-WH1(A31V) fused to mRFP accumulates as amyloid inclusions in the E. coli cytoplasm.
A. Images of the E. coli MC4100 cells transformed with the plasmids outlined in Fig. 1A, after a 2 h induction with IPTG. Cells were fixed and the DNA was stained with DAPI. Representative microscopy fields are shown, overlaying images obtained after excitation at 546 nm (red emission, mRFP) and 360 nm (blue emission, DAPI). At the lower left corner of each panel, the decimal proportion of red fluorescent cells/total number of cells counted is shown.
B. (a) Phase-contrast image of induced A31V–mRFP cells, after staining with 1 mM of the amyloidophylic fluorophore BTA-1. Arrows indicate dense cytoplasmatic inclusions. (b) mRFP intrinsic (red) fluorescence. (c) Staining with BTA-1 (green-blue). (d) Overlay of images (b) and (c), showing colocalization of the red and the greenish foci, which match with the inclusions in (a). Magnification bars: 2 µm.

BTA-1 is a fluorophore related to thioflavin T (Th-T) that is specific for amyloid cross-β structures, but possesses a superior ability to permeate cell membranes with no interference from DNA cross-binding (Cordeiro et al., 2001; and our own observations). These properties are derived from its neutral rather than positive charge (Wu et al., 2008). In bacteria expressing WH1(A31V)–mRFP, BTA-1 fluorescence colocalized with the core of the protein inclusions (Fig. 2B) showing their intrinsic amyloid nature.

RepA-WH1 amyloids impair cell fitness and are transmissible

To test whether the fluorescent intracellular inclusions were detrimental to bacterial fitness, time-lapse microscopy was performed. Consistently, cells exhibiting diffuse fluorescence (WT protein overexpressed) showed 1.7-fold increased generation times compared with non-fluorescent bacteria (Movie S1A), whereas cells carrying the fluorescent foci (A31V variant overexpressed) showed fivefold increased generation times (Movie S1B). The inclusion-free cells therefore became the major component of the resulting micro-colonies, compatible with the observation of weaker red coloration of A31V colonies compared with WT colonies (Fig. 1B).

The detailed observation of how bacteria containing WH1(A31V)–mRFP inclusions divide (Figs 3 and S2) revealed that 20–30 min before any equatorial constriction separating the two daughter cells (a morphological indication for cytokinesis) became evident, a small extra cytoplasmatic fluorescent particle appeared (arrows). In cells starting with two large sub-polar inclusions (the most common situation), the emergent fluorescent particle tended to locate towards the mid-cell, close to where the new cell pole will be located (Fig. 3, cells labelled A and B″). At the point of division, the central amyloid seed had grown in to a full size protein inclusion, temporally hampering septation and resulting in elongated cells (Fig. 3, cells A′ and A″). Cells overcame division arrest through the asymmetric inheritance of the central inclusion (see below), so that filaments composed of undivided bacteria were not observed, except after long incubation times. Another type of cell bore both sub-polar and mid-cell inclusions (Fig. 3, cell B). In this case, the emergent fluorescent particle located at the empty pole and a subsequent asymmetric cytokinesis resulted in daughter cells carrying either a single (Fig. 3, cell B′) or two sub-polar (cell B″) inclusions.

Figure 3.

Dynamics of WH1(A31V)–mRFP inclusions in vivo, as followed by time-lapse microscopy.
Two E. coli cells (labelled A and B) bearing inclusion particles are shown growing on LB agar at 37°C. Serial frames (odd rows, phase contrast; even rows, mRFP fluorescence) were taken at the indicated time intervals (lower right corner in the phase-contrast images) starting 15 min after IPTG induction. Daughter cells are labelled as prime (first generation) and double prime (second generation). Arrows point to emergent fluorescent particles, coloured according to their order of appearance (white → yellow). Magnification bars: 2 µm.

The fate of cells with a single aggregate deserves further attention (Fig. S2). If the inclusion was centred (cell A), cytokinesis asymmetrically distributed the maternal inclusion to one of the daughter cells, where it had a sub-polar location (cell A′). The transmission of amyloidosis to the other daughter cell was thus compromised, depending on the spontaneous, DNA-promoted, appearance of a seed (cell A″). This became a source of cured cells devoid of amyloid inclusions. If the single protein inclusion had a sub-polar location (Fig. S2, cells B and A′), the emergent particle appeared at the opposite sub-pole, which continued spreading amyloidosis (cell B′). On the other hand, the daughter cell (B″) carrying an inclusion close to the pole, which had already undergone two divisions, remained quiescent. Expanded generation times and quiescence of old pole-bearing cells are both aging-related phenotypes in E. coli cells carrying inclusion bodies (Stewart et al., 2005; Lindner et al., 2008; Winkler et al., 2010). However, such typical inclusion bodies exhibit more homogeneous sizes, discrete numbers and intracellular distributions, and elicit less acute cellular toxicities, than the WH1(A31V)–mRFP amyloid inclusions reported here.

Ultrastructural features and distribution of RepA-WH1 intracellular amyloid inclusions

To study the fine structure of the intracellular amyloid inclusions, they were analysed by confocal laser microscopy (Fig. 4A) and immuno-electron microscopy (iEM) on thin cell sections (Fig. 4B). The fusion proteins were first targeted with an antibody against their N-terminal hexa-histidine tags (Fig. 1A), and then located using secondary antibodies conjugated either to a green fluorochrome or to colloidal gold nano-particles respectively. Confocal sections through the cells bearing WH1(WT)–mRFP revealed fluorescent spots throughout the cytoplasm that, in the case of the A31V variant, clustered into the inclusions. In iEM, the WH1(WT)–mRFP cells also exhibited the label dispersed throughout the cytoplasm (Fig. 4B, top). However, the electron-dense intracellular inclusions were selectively labelled when the WH1(A31V)–mRFP amyloidogenic variant was expressed (Fig. 4B, bottom). Cells of this type mostly exhibited the small amyloid seeds (Figs 3, S2 and 4A) as electron-dense, gold-labelled particles embedded in the less dense nucleoid domain (Fig. 4B, bottom; arrow in panel b). Mature amyloid inclusions were segregated from the nucleoid (identified by its nucleoprotein fibres; see also DAPI staining in Fig. 2A).

Figure 4.

Ultrastructural analysis of amyloid inclusions in E. coli MC4100 cells.
A. Confocal sections (Δz = 0.2036 µm) of bacteria expressing WH1(WT)–mRFP (left) or WH1(A31V)–mRFP (right). Rows: mRFP fluorescence (a, red); anti-His tag fluorescence (b, green); (c): the overlay of (a) and (b). Arrows indicate amyloid primordia. Magnification bars: 3 µm.
B. Immuno-electron microscopy of thin sections across bacteria expressing WH1–mRFP, either WT (top) or A31V (bottom). The boxed regions indicate the zoomed images to the right. Antibodies conjugated with colloidal gold spherical particles (ø ≈ 10 nm) decorate the primary antibody + His6(WH1) complexes, which are located on the electron-dense large protein aggregates in WH1(A31V), but are dispersed throughout the cytoplasm in WH1(WT). Arrow points to an amyloid primordium. Magnification bars: 200 nm.

Purified intracellular inclusions template RepA-WH1 amyloidosis in vitro

The amyloidogenic potential of the intracellular protein inclusions was then explored in vitro. About 50% of the E. coli expressed WH1(A31V)–mRFP protein were detected in the soluble lysate fraction, the other 50% had formed aggregates (Fig. S3A). The aggregates were purified via a protocol involving washing with detergents, extensive digestion with nucleases and sucrose gradient sedimentation. The final preparation was analysed by means of transmission electron microscopy (TEM; Fig. S3B) and its amyloid nature was confirmed by Th-T binding (Fig. S3C).

Sub-stoichiometric amounts of the purified WH1(A31V)–mRFP aggregates were added to either the WT or the A31V variant of soluble RepA-WH1 under in vitro amyloidogenic conditions optimal for the slow and ordered growth of fibres (Giraldo, 2007), but lacking the inducer, opsp-dsDNA. Aliquots were taken at time intervals throughout the incubation and the progression of protein amyloidosis was tracked by Th-T binding and analysis of fluorescence emission (Fig. 5A). The WH1(A31V)–mRFP aggregates, when compared with the samples incubated in their absence, accelerated the appearance of amyloid material by one week. It is noteworthy that seeding was exerted on both the homologous (A31V) and the heterologous (WT) WH1 variants, although for the latter to a more reduced extent, compatible with its milder amyloid character (Giraldo, 2007). TEM inspection of the final aggregates (Fig. 5B) revealed that, whereas both unseeded WH1 proteins produced small granular particles, the addition of WH1(A31V)–mRFP seeds resulted in hallmark large irregular aggregates (WT) and well defined fibres (A31V) (Giraldo, 2007). The latter neatly emerged from the WH1(A31V)–mRFP inclusions and were indistinguishable from the standard amyloid fibres promoted by opsp-dsDNA (Giraldo, 2007; Gasset-Rosa et al., 2008b) (Fig. 5B), indicating that once DNA has promoted the conformational transformation of WH1(A31V) in vivo this is preserved to template amyloid growth in vitro. These results are consistent with the evidence presented above on mother cell-to-daughter cell seed transmission, rather than stochastic growth of a nucleus from the soluble protein pool, as the mechanism for efficient spreading of RepA-WH1 amyloidosis in vivo.

Figure 5.

Purified WH1(A31V)–mRFP cellular inclusions seed the growth of amyloid fibres in vitro.
A. Kinetics of amyloid growth, as followed by Th-T fluorescence emission. Pure soluble WH1 (WT or A31V; 35 µM) seeded (filled symbols) or not (empty symbols) with 1 µg of purified WH1(A31V)–mRFP aggregates (Fig. S3). Samples were incubated as described previously (Giraldo, 2007), without inducer dsDNA. Aliquots were taken at the indicated intervals, incubated with Th-T (60 µM) and the fluorescence emission spectra were acquired (λexcit = 440 nm). Emission at 481 nm was plotted against time (error bars: SD; three replicates).
B. TEM visualization of the protein aggregates obtained after 4 weeks' incubation. Top right: standard WH1(A31V) amyloid fibres control grown in the presence of opsp-DNA. Magnification bars: 200 nm.


In this study we show that the expression of a hyper-amyloidogenic mutant variant (A31V) of a plasmid DNA replication protein (RepA-WH1) fused to a red fluorescent reporter (mCherry/mRFP) results in the generation of amyloid protein aggregates exhibiting a toxic phenotype in E. coli, which is evident by a large expansion in the generation time of bacteria. Such amyloid inclusions are vertically transmitted to the progeny (i.e. coupled to cell division). Purified WH1(A31V)–mRFP aggregates can template in vitro the growth of amyloid fibres from soluble WH1(A31V) protein. These fibres are morphologically identical to the fibres grown using dsDNA as a cofactor (Giraldo, 2007; Gasset-Rosa et al., 2008b), indicating the close relatedness of these two types of amyloid assemblies.

A significant feature of the RepA-WH1(A31V) protein is the striking ability of small dsDNA ligands of defined sequence (opsp) to promote, through transient binding, the assembly of the protein into amyloid fibres in vitro (Giraldo, 2007; Gasset-Rosa et al., 2008b). In terms of the ensemble concept for native protein structures (Boehr et al., 2009), allosteric binding of dsDNA to WH1 (Díaz-López et al., 2003) would be determinant in selecting a subset of aggregation-prone protein conformations among those fluctuating at equilibrium, thus driving the whole process towards amyloid assembly. We have successfully conferred the ability of DNA to promote WH1(A31V) amyloidosis in E. coli cells, by showing that cloned repeats of the specific opsp dsDNA sequence enhance protein accretion to the intracellular inclusions in vivo (Fig. 2A, right). It is remarkable that, allowing for a single mismatch, the E. coli genome (http://genolist.pasteur.fr/Colibri) carries up to 48 dispersed copies of the opsp sequence, which seem able to trigger WH1(A31V) aggregation in the absence of plasmid-borne opsp repeats, although with a reduced frequency (Fig. 2A, left). It follows that sub-stoichiometric (catalytic) levels of inducer DNA would be sufficient to promote amyloidosis in the presence of a vast excess of WH1(A31V) molecules. The dispersion of the copies of the opsp sequence present in the E. coli genome made it unfeasible to study WH1 amyloidosis in the absence of a specific inducer ligand. The potential of nucleic acids to act as promoters of amyloidogenesis in vitro has been extensively documented for proteins involved in neurodegenerative amyloid diseases, such as PrP prion (Cordeiro et al., 2001; Nandi et al., 2002; Deleault et al., 2003; Geoghegan et al., 2007; Silva et al., 2008) and α-synuclein (Cherny et al., 2004; Hegde et al., 2010), together with other physiologically relevant cofactors such as glycosamineglycans or phospholipids (Deleault et al., 2007; 2010; Abid et al., 2010; Wang et al., 2010). These findings underline the importance of studying amyloidogenesis in a biologically relevant context (Bellotti and Chiti, 2008). This is the first time that the inducer effect of nucleic acids, via RepA-WH1(A31V) amyloidosis, has been shown both in vitro and in vivo. It is noteworthy that the same protein sequence stretch directly contributing to RepA-WH1 amyloidosis (Giraldo, 2007) was recently found to be involved in establishing the RepA-RepA assemblies that negatively regulate plasmid replication (Gasset-Rosa et al., 2008a). However, direct evidence for an amyloid structure in such assemblies is lacking. Although most amyloidoses described to date involve cytoplasmatic protein deposits, the first nuclear-localized yeast prion has just been reported for the transcription factor Sfp1p/[ISP+] (Rogoza et al., 2010). The general biomedical relevance of our bacterial model system for DNA-promoted protein amyloidosis may be more far-reaching, since protein amyloid deposits in the cell nucleus are commonly associated with human neurodegenerative diseases caused by proteins bearing polyglutamine repeats, as is the case for Huntington's disease (De Rooij et al., 1996) and diverse spinocerebellar ataxias (Bauer and Nukina, 2009), and is also true for transmissible spongiform encephalopathies (Mangéet al., 2004). A role for nucleic acids in promoting protein amyloidosis in these diseases cannot therefore be excluded.

Regarding the intracellular distribution of WH1(A31V)–mRFP amyloid particles, similar fluorescent spots have previously been described for multi-copy plasmids lacking a partition system, when labelled with fluorescent protein reporters tightly bound to DNA (Pogliano et al., 2001; Anand and Khan, 2010). However, such plasmid-linked foci usually have better defined circular contours and are smaller in size than the WH(A31V) inclusions. It is unlikely that the expression plasmids used in our study would determine the clustering of protein aggregates, thus acting as a physical carrier for the transmission of amyloid particles because, whereas WH1(WT)–mRFP appeared dispersed throughout the cytoplasm, WH1(A31V)–mRFP aggregated into inclusions (Fig. 2A), despite identical DNA binding properties between the two proteins (low affinities and high off-rates) (Díaz-López et al., 2006; Giraldo, 2007; Gasset-Rosa et al., 2008a). It is likely that the large fraction of volume occupied by the nucleoid is the main determinant, through physical exclusion, of the distribution of protein aggregates throughout the cytoplasm (Winkler et al., 2010). Whether other factors, such as DnaK-J and ClpB chaperones (Rokney et al., 2009; Winkler et al., 2010), are involved in DNA-promoted RepA-WH1 amyloidosis by generating transmissible seeds through the fragmentation of intracellular inclusions, as is the case for their eukaryotic orthologues Hsp70-40s and Hsp104 in the spreading of prions in yeast (Shorter and Lindquist, 2008; Tyedmers et al., 2010), remains to be determined. It is possible that due to the small size of the transmitted WH1(A31V)–mRFP particles they could evade detection by conventional fluorescence microscopy, as it is the case for yeast propagons which are generated in large numbers (Byrne et al., 2009; Sindi and Serio, 2009). However, we observe a discrete number (usually one) of emergent small WH1(A31V)–mRFP particles in the newly formed poles of bacterial cells (Figs 3 and S2), which then grow to the size of mature inclusions.

To our knowledge, the toxicity of the WH1(A31V)–mRFP inclusions, in terms of their effect on the rate of cell division, is the highest reported for any inclusion body (IB) in E. coli to date (Stewart et al., 2005; Lindner et al., 2008; Winkler et al., 2010). IBs are bacterial intracellular protein aggregates commonly considered to be inert deposits. However, the presence of protein molecules within these IBs retaining a functional conformation is feasible (Martínez-Alonso et al., 2009). The pathways through which the WH1(A31V)–mRFP inclusions exert their toxicity will be the subject of future research and might provide new insight into molecular mechanisms common to other proteinopathies.

The amyloid character of the intracellular inclusions described in this study, their cell-to-cell transmissibility and their ability to nucleate and act as templates for the assembly of fibres, qualify RepA-WH1 as a minimalist model system to study the determinants of protein amyloidogenesis in general and, due to its associated toxicity, more specifically of amyloid proteinopathies. Furthermore, the fully synthetic nature of RepA-WH1 amyloids makes them suitable candidates for the bottom-up design of more complex biological assemblies (Deplazes, 2009; De Lorenzo, 2010). Although the RepA-WH1 amyloids described here are transmissible, they clearly differ from yeast prions in lacking both infectivity and a selectable, gain-of-function phenotype (Wickner et al., 2007; Tessier and Lindquist, 2009). Engineering valuable epigenetic characteristics into the RepA-WH1 amyloid scaffold is a future target for synthetic biology.

Experimental procedures

WH1–mRFP +/− (opsp)18 gene fusions

Two oligonucleotides containing the opsp-dsDNA sequence (bold) (Giraldo, 2007), 5′-CAATGACAAGTCAATGACAAGT (opsp-A) and TGTTCAGTTACTGTTCAGTTAC-5′ (opsp-B), were phosphorylated (T4-PNK), annealed and ligated (T4-DNA ligase) to build concatemers. After end-filling (Klenow DNA polymerase), the ligation product ladders were visualized on 2% agarose gels. Concatemers were cloned into the expression vector pRGrectac-NHis-WH1 (WT and A31V, for short pWH1(WT/A31V); Giraldo 2007), previously linearized with SmaI. The monomeric variant of DsRed mCherry was amplified by PCR (Pfu DNA polymerase) from the plasmid pRSETb–mCherry (Shanner et al., 2004) using the following primers (mutations in upper case):

inline image

The PCR product was fused downstream of repA-WH1 in pWH1(WT/A31V), either with or without the cloned (opsp)18 repeats (see above). Plasmids were linearized with HindIII and then recombined with the PCR fragment using the In-Fusion Dry-Down PCR cloning kit (Clontech). The resulting in-frame fusions included a (Gly-Ser-Ser)2-Gly flexible linker between the two protein partners. Constructs were CaCl2-transformed in E. coli JM109 (recA) cells and checked by restriction analysis and automated DNA sequencing (http://www.secugen.es/).

Expression of WH1–mRFPs

All in vivo experiments were carried out in the common E. coli K-12 strain MC4100 (Ferenci et al., 2009). pLacI, a pACYC184 (CmR) derivative containing a cloned 1212 bp BstYI (lacI) fragment, was used as a donor for the LacIq repressor. Bacteria grown overnight on Luria–Bertani (LB) agar plates plus ampicillin (100 µg ml−1) were then inoculated in the same (liquid) medium and grown at 37°C to early exponential phase (OD600 = 0.2). The cells were then induced with 0.5 mM IPTG for 2 h. Expression analysis was carried out by SDS-PAGE, followed by Western-blot detection of the His6-tagged WH1–mRFPs. This was performed using an anti-His murine monoclonal primary antibody (1:5000 dilution; Sigma) and a secondary anti-mouse HRP-conjugated antibody (1:10 000) and chemiluminescent detection (ECL plus; GE Healthcare). After expression, plasmids from 5 × 109 cells (1.0 OD600 ≈ 1.5 × 109 cells ml−1) were purified (High Pure Plasmid Isolation kit; Roche) and then analysed by UV-absorption spectrophotometry and agarose gel electrophoresis.

Fluorescence microscopy

Routine microscopic observation of the bacterial cells was performed using a live cell imaging AF6000 LX equipment (Leica), consisting of a DM16000B inverted microscope, a HCX PLAN APO (100×/NA 1.40) oil immersion objective and a Hamamatsu C9100-02 CCD camera. The following excitation and suppression filters were used: mCherry (BP 546/12, BP 600/40); DAPI (BP 360/40, BP 470/40); Alexa 488 (BP 480/40, BP 527/30); and BTA (BP 467/37). Phase-contrast images of the fields were also recorded. Confocal laser microscopy was carried out using a Leica imaging device (TCS-SP2-AOBS, including a DMIRE2 inverted microscope). The software provided by Leica was used for the superposition of the views resulting from the illumination of a given field at different regions of the spectra and for image quantifications.

For observations performed on fixed bacteria, 0.5–1.0 ml culture aliquots were spun-down and the cells were washed with the same volume of PBS buffer (10 mM Na3PO4, 150 mM NaCl, 15 mM KCl pH 7.4). Cells were collected by centrifugation at 1600 g and the cell pellet was then fixed in fresh 4% paraformaldehyde (PFA) in PBS buffer, for 30 min at room temperature. After fixation, the cell pellets were washed twice in PBS. In some experiments, cells were stained with 5 µg ml−1 DAPI (4′,6-diamidino-2-phenylindole–2HCl; Serva) for 15 min during fixation. 10 µl aliquots of cells in PBS were air-dried on poly-l-lysine coated slides and mounted with Vectashield (Vector Laboratories).

For BTA-1 staining, cells were fixed in PFA and washed in PBS, as described above. Then they were suspended in BTA [2-(4′-methylaminophenyl) benzothiazole (Sigma); 1 mM in ethanol], incubated for 30 min at room temperature, washed in PBS and mounted.

For immunofluorescent staining, cells were fixed in freshly prepared 2% PFA, made in PBS buffer containing 0.5% Triton X-100, for 30 min at room temperature. After three washes with the same solution, pellets were suspended in GTE buffer (50 mM glucose, 10 mM EDTA, 20 mM Tris-HCl pH 7.5). Then, 15 µl aliquots were air-dried on poly-l-lysine coated slides, incubated with freshly prepared lysozyme (8 µg ml−1 in GTE), for 5 min at room temperature, and washed twice with PBS for 10 min. Next, they were incubated with blocking solution (PBS plus 0.05% Tween-20, 2% BSA), for 30 min at room temperature. Slides were then incubated with the anti-His monoclonal primary antibody (1:400 dilution in blocking solution), at 4°C overnight. The preparations were then washed five times for 10 min in PBS containing 0.05% Tween-20 and incubated with Alexa 488-coupled anti-rabbit secondary antibody (Molecular Probes; 1:100 dilution in blocking solution) for 1 h at room temperature in the dark. After five washes with PBS plus Tween-20, samples were air-dried and mounted as described above. Negative controls were performed omitting the incubation with the primary antibody.

For time-lapse microscopy of living bacteria, 100 µl of melted aerated 1.5% agar-LB (supplemented with ampicillin) was placed onto a glass slide, overlaid with a coverslip and left to solidify. Then the coverslip was removed, IPTG was added and 10 µl of the cultures were immediately placed onto the agar layer and covered with a clean coverslip. Specimens were maintained in the microscope incubation chamber, under controlled humidity and temperature (37°C), throughout all the experiment. Images were digitally captured at the specified time intervals.

Transmission electron microscopy and immuno-electron microscopy

Purified protein inclusions and aggregates assembled in vitro were examined after negative staining with uranyl acetate, as indicated previously (Giraldo, 2007). TEM was carried out using a JEOL JEM-1230 transmission electron microscope, operated at 100 kV. Photographic films (Kodak) were digitized using an AGFA Arcus 1200 scanner.

For iEM, cultures of exponentially growing cells were fixed in 2% PFA in PBS buffer for 1 h, washed three times in PBS for 10 min and then embedded in 2% agarose. The agarose blocks were cut into small pieces, washed in PBS, dehydrated through an ethanol series (from 30% to absolute ethanol), embedded in London white resin and cured at 60°C. Immuno-gold labelling was performed on ultra-thin sections of cells collected on the top of Formvar-coated nickel grids. These were successively floated on drops of: 20 mM glycine in PBS for 15 min; PBS containing 0.05% Tween-20 for 10 min; blocking solution (PBS plus 0.05% Tween-20, 2% BSA) for 30 min; and anti-His primary antibody (1:400 dilution in blocking solution), for 2 h at room temperature. After three washes (10 min each) in PBS-Tween-20, grids were incubated with an anti-mouse secondary antibody conjugated with 10 nm gold particles (Sigma; 1:50 dilution in blocking solution), and then washed three times in PBS-Tween-20 and once in bi-distilled water. Grids were then air-dried and contrasted with 5% aqueous uranyl acetate for 30 min, washed in bi-distilled water and air-dried before being observed by EM (see above). Negative controls were performed omitting the incubation with the primary antibody.

Purification of soluble and aggregated WH1(A31V)–mRFP

Two Erlenmeyer flasks (2 l vol) with 400 ml of LB medium plus ampicillin (100 µg ml−1) were inoculated with E. coli MC4100 cells carrying pWH1(A31V)–mRFP + opsp18. Cultures were grown at 30°C to OD600 = 0.8, when IPTG was added to 0.5 mM and the temperature was shifted to 37°C. The cells were then incubated overnight (OD600 = 4.0). Bacteria were harvested, washed with cold 0.9% NaCl and the cell pellet was resuspended in 15 ml of a buffer containing 1.0 M NaCl, 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 1% Brij-58 (Sigma), 1 mM pNH2-benzamidine and 10% glycerol. Lysozyme (1 mg) was then added and the suspension was incubated at 37°C for 30 min before sonication (two cycles, 30 s each, using a 1 cm wide tip). The resulting cellular lysate was clarified by ultracentrifugation at 20 000 r.p.m. (1 h, 4°C) in a Beckman Ti60 rotor. Approximately 50% of WH1(A31V)–mRFP was detected as soluble protein in the supernatant and was purified by immobilized metal affinity chromatography (Ni2+-IMAC), as indicated for WH1(A31V) (Giraldo, 2007).

Aggregated WH1(A31V)–mRFP protein was resuspended in 15 ml of a buffer including 0.1 M NaCl, 20 mM Tris-HCl pH 7.5, 10 mM MgCl2, 0.5% Na-deoxycholate, 0.5% N-dodecyl-N,N-dimethylammonio-3-propanesulphonate (SB-12), 1 mM DTT, 1 mM pNH2-benzamidine and 10% glycerol. The suspension was sonicated and centrifuged as before. The pellet was resuspended in 15 ml of the same buffer (excluding the detergents), supplemented with 5 units of DNaseI (Worthington) and 0.5 mg of RNaseA (Roche), and then incubated for 1 h at 37°C. The protein suspension was sonicated and centrifuged again and the pellet was resuspended in 10 ml of 0.1 M Na2SO4, 4 mM MgSO4, 20 mM Tris-HCl pH 8 and 10% glycerol. After a further sonication/centrifugation round, aggregated protein was homogenized in 1.5 ml of the same buffer. One-hundred-microlitre aliquots of this suspension were layered on sucrose discontinuous gradients (60–50–40–30–20% in the same buffer, 200 µl of each solution) cast in 2 ml Eppendorf tubes as described previously (Liebman et al., 2006). The tubes were centrifuged at 7500 r.p.m. for 16 h at 4°C, then 100 µl aliquots were carefully removed from the top of the tube and 10 µl samples were analysed by SDS-PAGE. WH1(A31V)–mRFP inclusions were found to peak at fraction 8 in the gradient. The protein concentration in the inclusions was estimated by dissolving 10 µl of the purified aggregates in 6 M Gu-HCl and 20 mM Tris-HCl pH 8, determining the absorbance in the UV-Vis spectrum and then using the equation C (mg ml−1) = 1.55 × A280 − 0.76 × A260 (Simonian and Smith, 2006). From the A280/A260 ratio (1.82), purified protein inclusions were judged to be free of nucleic acids.

Fluorescence spectroscopy analysis of the binding of thioflavin-T (Th-T) to amyloids

To confirm for the amyloid nature of the purified protein inclusions, 50 µl of a 4 mg ml−1 suspension was added to 440 µl of a buffer including 0.1 M Na2SO4, 4 mM MgSO4 and 20 mM Tris-HCl pH 8. Then 10 µl of a 3 mM Th-T (Sigma) stock (final concentration: 60 µM) was added. To study the kinetics of amyloid seeding by the purified inclusions, 450 µl of buffer and 10 µl of the Th-T stock were supplemented with 40 µl of the amyloidogenic samples at the indicated times (0, 1 h and 1, 7, 14, 21, 28 and 35 days). This was performed as described previously (Giraldo, 2007); briefly, 100 µl aliquots including 35 µM of purified soluble WH1(WT) or WT(A31V) in 0.1 M Na2SO4, 30 mM Tris-HCl pH 8, 4 mM MgSO4, 7% PEG4000 and 3% MPD were supplemented (or not, as indicated) with 1 µg of pure WH1(A31V)–mRFP inclusions. It is noteworthy that inducer opsp-dsDNA, an absolute requirement for ordered fibres (Giraldo, 2007), was absent from the incubations.

In all cases, samples incubated with Th-T were placed in Hellma F-QS cuvettes (2 × 10 mm path length) and allowed to equilibrate at room temperature for 5 min before analysis. A Jobin-Ivon SPEX Fluorolog-3 spectrofluorometer, with the temperature controlled at 22°C by a Peltier module, was used for spectra acquisition. The excitation wavelength was set to 440 nm and the excitation and emission slits were set to 5 and 10 nm respectively. The integration time was 1 s and the spectra were acquired in 1 nm steps. Data were plotted and analysed using KaleidaGraph v.3.6 software (Synergy Software, CA, USA).


We are grateful to Roger Y. Tsien for the gift of pRSETb–mCherry, to Ana M. Serrano, María T. Seisdedos, Fernando González and Mercedes Carnota for technical assistance and to Salvador Ventura and Daniela Rhodes for the critical reading of this manuscript. This work was supported by grants from the Spanish Ministry of Science and Innovation (MICINN: BFU2006-00494, BIO2009-06952 and CSD2009-00088) and the Autonomous Government of Madrid (CAM: P-BIO-0214-2006).