On the aggregation properties of FMRP – a link with the FXTAS syndrome?


A. Pastore, MRC National Institute for Medical Research, The Ridgeway, London NW7 1AA, UK
Fax: +44 20 8905 4477
Tel: +44 20 8816 2630
E-mail: apastor@nimr.mrc.ac.uk


Fragile X mental retardation protein (FMRP) is an RNA binding protein necessary for correct spatiotemporal control of neuronal gene expression in humans. Lack of functional FMRP causes fragile X mental retardation, which is the most common inherited neurodevelopmental disorder in humans. In a previous study, we described the biochemical and biophysical aggregation properties of constructs spanning the conserved region of FMRP and of two other human fragile X related (FXR) proteins, FXR1P and FXR2P. Here, we show that the same regions have an intrinsic tendency to aggregate and spontaneously misfold towards β-rich structures, also under non-destabilizing conditions. These findings pave the way to future studies of the mechanism of formation of FXR-containing ribonucleoprotein granules and suggest a possible link with the as yet poorly understood FXR proteins’ associated pathologies.

Structured digital abstract


circular dichroism


fragile X mental retardation protein


fragile X related


fragile X syndrome


fragile X associated tremor ataxia syndrome


K homology


messenger ribonucleoprotein


N-terminal domain


nuclear export signal


cytosine guanine triribonucleotide


thioflavin T


The fragile X mental retardation protein (FMRP) is an ∼ 70 kDa human protein encoded by the X-linked gene FMR1 which is expressed in different organs, most prominently in brain and gonads [1–3]. FMRP is a multi-domain protein which contains two Tudor domains connected to two protein K homology (KH) domains by an ca. 80 amino acids residue long linker. This region is followed by a putative intrinsically unstructured region which contains also arginine- and glycine-rich (RGG) motifs [4,5]. Co-presence of different nucleic acid binding domains in FMRP suggests that the protein has a prominent capacity to bind nucleic acids, in particular RNA, as experimentally confirmed both in vitro and in vivo [6,7].

The cellular role of FMRP is not well understood. Experimental evidence shows that FMRP binds co-transcriptionally to certain messenger RNAs forming messenger ribonucleoprotein (mRNP) particles, which are exported from the nucleus to the cytoplasm [8]. In the cytoplasm FMRP associates to microtubules, to polysomes and to mRNPs and permits the mRNP particles to be delivered to distal dendrite sites [9]. It has also been shown that messenger RNAs bound to FMRP are translationally repressed and that, in neurons, FMRP acts in an activity-dependent manner as an inhibitor of translation initiation ([10] and references therein).

Most studies on FMRP are related to its functions in brain neurons for two reasons. First, the lack of functional FMRP, due to transcriptional silencing of FMR1 gene, causes a neurodevelopmental disorder, fragile X mental retardation syndrome (FXS), the most common inherited mental disorder in humans. FXS is characterized by mild to severe mental retardation, autistic behaviour and, in male patients, macro-orchidism [11]. Second, alteration of FMRP expression, characterized by increased levels of FMR1 mRNA and decreased protein levels, can lead to a late onset neurodegenerative disorder, the fragile X associated tremor ataxia syndrome (FXTAS), with symptoms similar to Parkinsonism, and to premature ovarian failure in females [12,13]. Two other proteins, fragile X related (FXR) proteins 1 and 2 (FXR1P and FXR2P), can compensate partially for lack of FMRP in some organs of FXS patients, but not in brain and in gonads, thus emphasizing the crucial role of FMRP in correct spatiotemporal control of neuronal gene expression and for normal sexual maturation. FXR1P and FXR2P are structurally and functionally related to FMRP and they all together form the FXR protein family which includes members from different phyla [14]. All FXR proteins show a high degree of amino acid sequence conservation in their amino terminus and central region comprising Tudor and KH domains, whereas the carboxyl terminus has low sequence similarity with the only common denominator being the presence of RGG motifs.

FXR proteins are components of different types of nuclear and cytoplasmic ribonucleoprotein granules in which they often co-localize [1,15–19]. FXR proteins can form hetero-oligomers in vitro and when over-expressed in cellular systems [20–22], although in mammalian cells they are believed to preferentially homo-oligomerize [21]. The oligomerization properties of FXR proteins are likely to have significant importance for regulation of their cellular functions as shown in different model systems. In Drosophila neurons, for instance, the mobility of certain mRNAs is controlled by FMRP in a concentration-dependent manner [23]. High levels of transfected Fmrp in mouse embryonic Fmr1 KO STEK cells induce formation of cytoplasmic stress granules in which mRNAs are trapped into repressed mRNP granules [24,25]. In a recent study, we investigated the oligomerization properties of human FXR proteins and showed that, in vitro, multi-domain constructs from the highly conserved N-terminus have an elevated tendency to aggregate [26]. They self-assemble not just by forming dimers but through a more complex pattern of self-association which proceeds in a continuous way from the monomer to large molecular weight aggregates via formation of dimeric species. We proposed that this behaviour is typical of ‘complex-orphan proteins’, i.e. proteins which exist in the cell as part of large molecular assemblies. When produced in isolation, they have an elevated tendency to self-associate.

To further characterize the nature of aggregation of FXR proteins, we have carried out a study of their aggregation properties using different approaches. We identified by in silico analysis potential hot-spots of aggregation/fibrillation and showed that they all cluster in the protein N-terminus. We then studied the aggregation behaviour of various constructs from FMRP, FXR1P and FXR2P using complementary biophysical techniques. We demonstrate that not only do all constructs have an intrinsic tendency to aggregate but they also undergo an irreversible conformational transition towards β-enriched structures which are typical of amyloidogenic diseases such as Alzheimer’s, Parkinson’s and Huntington’s diseases. The transition occurs spontaneously also under native-like conditions without any need for fold destabilization. We propose that our findings could be relevant for understanding granule formation and could have a link with the pathogenesis of FXTAS.


FXR proteins present aggregation and amyloidogenic hot-spots

To understand the aggregation properties of FXR proteins, we first analysed their sequence for predicting potential aggregative and amyloidogenic hot-spots in polypeptides. The results all suggest the presence of several aggregation hot-spots which cluster in the highly conserved N-terminal ∼ 400 amino acids (Fig. 1A) whereas no sequence was identified in the C-terminal region of any of the FXR sequences, including the most ancestral FMRP homologue dFMR1 from Drosophila melanogaster. Two putative amyloidogenic regions are present in the NDF of the human FXR proteins (sequences YVIEYA and TYNEIV). Two more hot-spots amongst the nine detected in FMRP by waltz with reliability higher than 90% correspond to the residues 166–174 located in the linker region connecting the Tudor and KH domains and to residues 303–308 (LIQEIV) in the second KH domain.

Figure 1.

 Sequence alignment of FXR proteins and indication of fibrillogenic regions. The alignment was produced and colour coded according to clustalw2 [27]. Extra rows were added below for the rulers relative to human FMRP, FXR1P and FXR2P. The regions predicted as fibrillogenic by the waltz software [40] are indicated as red crosses.

Taken together, this in silico analysis suggests the presence of several potential aggregation foci all grouped in the highly conserved N-terminus.

Temperature induces a conformational transition of FMRP domains

We tested experimentally the role of the different regions in aggregation/fibrillation using single-and multiple-domain constructs from human FXR proteins which we knew from previous extensive characterization are stable to degradation [26] (Fig. 2). Although monodispersed at sufficiently low concentrations, we had previously demonstrated that they all have a strong tendency to aggregate [26]. To check whether aggregation is associated with misfolding, we probed their secondary structure as a function of temperature by far-UV circular dichroism (CD). We first analysed the secondary structure content at different temperatures of FMRP Nt-KH1 (residues 1–280), the longest of the FMRP fragments we could obtain in a stable form (Fig. 3A). The spectrum of Nt-KH1 is typical of an α–β fold at 20 °C, as expected from the presence in the construct of Tudor and KH domains [4,28]. At higher temperatures (40–45 °C), the conformation starts to change (Fig. 3B). At 55 °C, the spectrum becomes typical of an all-β protein indicating a profound structural rearrangement with a minimum around 215 nm. The transition is irreversible, as the CD spectra of samples treated at 45 °C remain unaltered after the temperature is decreased to 37 or to 20 °C.

Figure 2.

 Summary of the modular structure of FXR proteins (A) and of the constructs used in this study (B) using a smart-like [41] representation. NDF stands for the N-terminal domain which contains two Tudor domains. KH stands for protein K homology domain. Linker indicates the region between the second Tudor domain and KH1. NES stands for nuclear export signal. The domain boundaries used are the same as in [26].

Figure 3.

 Spectroscopic study of temperature-induced conformational changes of FMRP. (A) Far-UV CD spectra of Nt-KH1 of FMRP recorded at 20 °C (black line) or 55 °C (dotted line) and expressed in molar ellipticity of Nt-KH1. (B) Temperature course at 220 nm, in molar ellipticity, of FMRP NDF (curve a), Nt (curve b) and Nt-KH1 (curve c). The rate of temperature increase was 1 °C·min−1. (C), (D) Far-UV CD spectra at 20 °C (black lines) or 55 °C (dotted lines) of NDF and Nt respectively. (E), (F) Time course of the α-to-β transition of FMRP NDF (curve a), Nt (curve b) and Nt-KH1 (curve c) induced by incubation at 45 and 50 °C, respectively. All spectra were recorded at protein concentrations of 5 μm.

The spectra of shorter FMRP fragments lacking the KH domain, NDF (residues 1–134) and Nt (residues 1–217), were examined to assess the role of the individual domains (Fig. 3C,D). These fragments also undergo structural rearrangements characteristic of an irreversible increase of β content above 45 °C, suggesting that they are individually able to misfold (Fig. 3B–D). To test whether pre-incubation at a fixed moderate temperature could also cause the observed β-enriched structural rearrangement, the three FMRP fragments, at the same protein concentration (5 μm), were independently incubated at different temperatures. In a time course measurement at 45 °C, NDF underwent a conformational change after 3 h pre-incubation (Fig. 3E). For comparison, the two longer constructs Nt and Nt-KH1 incubated at 45 °C did not reach, over the same time, the minimum CD signal (Fig. 3E) observed for the corresponding samples at 55 °C. A similar experiment was performed at 50 °C and resulted in a faster conformational transition compared with 45 °C (Fig. 3F). At this temperature, the intensities of the NDF and Nt-KH1 spectra reached a maximum after 40 and 120 min, respectively. Nt underwent a conformation transition at 50 °C which was not, however, complete during the time course of the experiment (3 h). This suggests that, under the same experimental conditions, the region C-terminal to the NDF, comprising the linker between NDF and KH1, has a prominent role in aggregation.

Taken together, these data show that different domains of the conserved region of FXR proteins rearrange their structure upon temperature treatment. In all cases examined this rearrangement occurs with very similar modalities and results in a significant enrichment of the β content.

Tendency to misfold is a conserved feature of FXR proteins

To extend our studies to other FXR proteins, we used the human FMRP paralogues FXR1P and FXR2P. The secondary structure of FXR1P Nt-NES was first examined at different temperatures since for this protein we could obtain a long construct which spans the whole conserved region (residues 1–380). The far-UV CD spectrum of this construct shows that an α-to-β conformational transition occurs under experimental conditions similar to those used for the FMRP domains (Fig. 4A). Interestingly, at the protein concentrations used for the assay (5 μm), the β-enriched conformation of FXR1P Nt-NES persists up to 80 °C. This behaviour is similar to that observed for FMRP Nt-KH1 and suggests that the KH2 and NES regions do not significantly influence solubility. Similar studies with shorter FXR1P fragments as well as with FXR2P fragments all underwent similar α-to-β conformational transitions, although overall FXR2P seemed to be more prone to aggregation (Fig. 4B).

Figure 4.

 Spectroscopic studies of FXR1P and FXR2P. (A) Far-UV CD spectra expressed in molar ellipticity of the FXR1P Nt-NES at 20 °C (continuous line) or 55 °C (dotted line) using 5 μm protein concentration. (B) Temperature course of FXR1P Nt-NES at 220 nm, using 5 μm concentrations. The recording rate was 1 °C·min−1.

These data indicate that not only FMRP but also the other human FXR paralogues have a strong and well conserved tendency to misfold.

Misfolding of FMRP domains occurs also under non-destabilizing conditions

Since the duration of temperature treatment plays a role in the observed process, we tested if prolonged incubation could lead to α-to-β conformational transition also at 37 °C, i.e. close to the physiological temperature at which FMRP functions in human cells. Initially, incubation of FMRP Nt-KH1 (5 μm) at 37 °C did not cause a significant secondary structure perturbation, but a conformational transition to a β-enriched structure was observed after a 45-h incubation (Fig. 5A). The lag time decreased to 16 h at concentrations threefold or sixfold higher (15 and 30 μm respectively), as expected for a concentration-dependent phenomenon such as aggregation (Fig. 5B,C). The final intensity of the recorded CD signal is very similar to that recorded at 55 °C, suggesting not only that a similar process takes place at both temperatures but also that the final states are structurally comparable.

Figure 5.

 Following the conformational transition of FMRP Nt-KH1 at 37 °C and different incubation times as a function of protein concentration. (A), (B), (C) Comparison of the FMRP Nt-KH1 CD spectra before (continuous line) and after (dotted line) incubation at 37 °C using 5, 15 and 30 μm protein concentrations, respectively. (D) Size exclusion chromatography elution profile of FMRP Nt-KH1: the continuous line chromatogram derives from freshly prepared Nt-KH1, the broken line chromatogram is the profile of the same sample kept at 4 °C for 16 h, and the dotted line chromatogram corresponds to a sample incubated at 37 °C for 16 h.

Freshly prepared FMRP Nt-KH1 samples (30 μm) are monomeric and monodispersed, and if stored at 4 °C they remain mainly monomeric with a small but detectable increase of dimeric species as time proceeds, i.e. after 16 h incubation. The size exclusion chromatograms of these samples incubated at 37 °C over the same time show the appearance of high molecular weight species which are absent both in fresh samples and in samples incubated at 4 °C (Fig. 5D). We can conclude that recombinant Nt-KH1 of FMRP has an intrinsic tendency to aggregate in vitro also at physiological temperature in native-like conditions.

FMRP has an intrinsic tendency to form protofibrils

To characterize the nature of the aggregates, we examined the longest fragments from FMRP (Nt-KH1) and from FXR1P and FXR2P (Nt-NES) using the fluorescence signal of thioflavin (ThT) dye that is indicative of formation of amyloid-like structures [29]. After addition of ThT to diluted FXR protein solutions (5 μm), samples were incubated for 15 min at increasing temperatures using 5 °C intervals and their fluorescence was monitored. No fluorescence could be detected during incubation at temperatures below 65 °C. At this temperature, a small increase of fluorescence emission at 482 nm was observed. Treated samples were then incubated at room temperature and fluorescence was measured at different time points reaching a maximal emission at 482 nm after 12 h (Fig. 6A). The following measurement after 24 h did not show further increase and a net decrease was observed after 96 h, probably caused by fibre sedimentation (data not shown).

Figure 6.

 Testing the fibrillogenic properties of FXR proteins. (A) ThT fluorescence assay on FXR2P Nt-NES treated over the temperature range 20–65 °C, increasing the temperature by 5 °C every 15 min and using 5 μm protein concentration. The fluorescence observed at 20 °C (black curve) decreases at temperatures between 45 and 55 °C (red curve), and increases after exposure to 65 °C (orange). The fluorescence signal reached a maximum after 12 h (green curve). (B) Transmission electron micrograph of negatively stained FMRP Nt (1–217) protofibrils pre-incubated at 50 °C for 3 h. (C) Negatively stained aggregates were observed for the FMRP Nt-KH1 construct (1–280) that was incubated for an extended time at 37 °C. (D) Electron micrograph of FXR1P Nt-NES aggregates obtained after 2 h incubation at 50 °C. The scale bars correspond to 100 nm.

To verify the morphology of the end-states of aggregation, we used transmission electron microscopy and examined samples after the conformational transition achieved either by direct exposure to 50 °C or by prolonged incubation at 37 °C, at two different protein concentrations (5 and 30 μm). Negatively stained protofibrillar and fibrillar assemblies were observed for FMRP Nt, FMRP Nt-KH1 and FX1RP Nt-NES (Fig. 6B–D). The FMRP Nt protofibrils appeared homogeneous, with an apparent uniform diameter (7 nm) and variable lengths that very rarely exceeded 100 nm (Fig. 6B). Interestingly, we also observed dense networks of long linear and unbranched fibrils with a 10-nm diameter, which displayed repeating segments and twists. The FMRP Nt-KH1 samples contained globular particles with an average diameter of 24 nm, often decorated with stain, as well as clustered deposits of fibrils with an average diameter of 6 nm (Fig. 6C). FX1RP Nt-NES aggregates have a curved appearance, with an apparent average diameter of 10 nm (Fig. 6D). They also clustered together and were often found to be decorated with the uranyl acetate stain. Taken together these results confirm a marked tendency of FXR constructs to fibrillation.


We have shown here that different fragments of FXR proteins not only have a strong tendency to aggregate as previously described [26] but also undergo an irreversible conformational transition which leads to a significant increase in their β-structure content. Several conserved putative aggregation and amyloidogenic hot-spots were predicted by in silico analysis of the FXR amino acid sequences. They are all grouped in the highly conserved (more than 70–80% identity and 80–90% similarity) N-terminal half of the proteins which is also the region involved in most of the interactions with the FXR cellular partners [30], suggesting that the aggregation hot-spots could have a prominent role in determining the hetero- and self-assembly behaviours of the full-length proteins. By combining CD spectroscopy and size exclusion chromatography, we have established a clear link between FMRP aggregation and misfolding, as observed by the concentration dependence of the conformational transition. Relatively small variations of protein concentration also lead to an increase of the rates at which the conformational transition occurs. An appreciable ThT fluorescence increase, irreversible β-enriched structural transitions and electron microscopy analysis support formation of ordered fibrillar aggregates. We observe a very similar behaviour for the two FMRP paralogues, FXR1P and FXR2P, for which the conformational transition occurs with modalities very similar to FMRP.

Interestingly, the observed transition towards β-enriched conformations occurs also at physiological temperature under non-destabilizing conditions. This behaviour is very interesting for a protein such as FMRP which contains multiple globular domains. While understanding how and when misfolding occurs is easier for intrinsically unfolded proteins, such as the Alzheimer Aβ peptides or α-synuclein, studies of globular proteins have traditionally involved the use of ad hoc mutations and/or destabilizing conditions, such as high temperature, molecular crowding or high pressure. These conditions lead to destabilization of the structure and access to fibrillogenic regions normally buried in the hydrophobic core. Only recently a small but steadily increasing number of examples are being described in which misfolding occurs in native-like conditions. This is the case for instance of the globular Josephin domain of ataxin-3, the protein responsible for the misfolding Machado–Joseph disease: we have recently shown that Josephin aggregation and misfolding is promoted by exposed hydrophobic patches involved in recognition of its natural partner ubiquitin, thus suggesting a link between normal function and misfolding [31]. Likewise, the globular domain of the prion protein contains a seeding region, H2H3, which retains its fold during the early stages of unfolding [32]. It has been suggested that in many proteins related to conformational diseases aggregation/amyloidogenic regions coincide with interaction surfaces [33–35].

Our results bear a number of interesting consequences. First, the strong tendency to aggregate of FXR proteins could help us to understand the driving forces that lead to granular formations and eventually understand more about their functional role. The findings presented in this study also suggest interesting possibilities for the ability of this family of proteins to contribute to both early life syndromes such as FXS (for instance through destabilizing mutations) and aggregation-related neurodegeneration later in life; such could be the case of FXTAS. The latter is a particularly interesting possibility since it could shed new light onto a still poorly understood syndrome: although RNA aggregation is thought to be an important driving force for formation of the pathological neuronal intranuclear RNP inclusions observed in FXTAS patients, little is known about the factors which determine their formation and stability [36]. The current view is that FXTAS is the end-point of a process that begins in early development and reaches its maximum late in life [37]. rCGG expansion in the 5′UTR region of FMR1 mRNA is required for formation of neuronal inclusions in FXTAS patients, which consist also of other mRNAs and of different proteins amongst which FXR1P and FXR2P [38]. Although FMRP, which is expressed in FXTAS patients, has not so far been identified amongst the components of the inclusions, we cannot exclude at this stage that its absence is not simply due to lack of sensitivity of the detection methods used. We suggest as a working hypothesis that the aggregative and misfolding tendency of one or more of the FXR proteins could contribute to pathology, thus adding FXTAS to the family of misfolding diseases. While more work needs to be done to test this hypothesis, we expect that important information may come from cellular studies of the effects of molecular crowding [39] on the FXR folding, homo- and hetero-association when surrounded by other cellular components.

Experimental procedures

Bioinformatic analysis

The amino acid sequences of human (Q06787, P51114, P51116), mouse (P35922, Q61584, Q9WVR4), chicken (Q5F3S6), frog (P51113, Q6GLC9, P51115), zebra fish (Q7SYM7, Q7SXA0, Q6NY99) and fruit fly (Q9NFU0) FXR proteins were aligned using program clustalw2 [27]. These sequences were also searched for putative determinants of aggregation and amyloidogenesis by the following consensus prediction tools: aggrescan (http://bioinf.uab.es/aggrescan/) for prediction of hot-spots for aggregation in polypeptides; pasta (http://protein.cribi.unipd.it/pasta/) for prediction of amyloid-like structure aggregation; amylpred (http://biophysics.biol.uoa.gr/AMYLPRED/) to predict features related to the formation of amyloid fibrils; tango (http://tango.switchlab.org/) for prediction of sequence-dependent and mutational effects on the aggregation of the peptides and proteins; and waltz (http://waltz.switchlab.org/) for predicting amyloidogenic regions in protein sequences.

Cloning, protein expression and purification

The constructs studied in this paper were produced according to procedures previously described [26]. In short, clones of human FMR1, FXR1 and FXR2 were used as templates for DNA amplification by PCR. PCR amplicons encoding different fragments of the conserved region of FXR proteins were cloned into a modified pET-24 vector (Novagen, Gibbstown, NJ, USA) encoding an amino terminal Trx (thioredoxin)-His6-tag and a tobacco etch virus (TEV) protease cleavage site.

Escherichia coli BL21 STAR (DE3) cells transformed with plasmids encoding different FXR fragments were grown at 37 °C in Luria–Bertani medium containing appropriate antibiotic. Protein over-expression was induced with 0.2 mm isopropyl thio-β-d-galactoside after the cell culture had reached D600nm = 0.8; the growth was continued for an additional 5 h at 28 °C. The cells were harvested by centrifugation, resuspended in a lysis buffer containing 20 mm Tris/HCl (pH 8.0), 150 mm NaCl, 10 mm imidazole, 0.2% Igepal CA-630 (Sigma–Aldrich, St Louis, MO, USA), 2 mmβ-mercaptoethanol, supplemented with the Complete EDTA-free protease inhibitor cocktail (Roche, Indianapolis, IN, USA), and lysed by ultrasonication. The recombinant peptides were then purified from the soluble fraction of the centrifuged cell lysate by immobilized metal-affinity chromatography (IMAC) using Ni-NTA (Ni2+-nitrilotriacetate) metal-affinity chromatography matrix (Qiagen, Yokyo, Japan), and dialyzed against 50 mm Tris/HCl (pH 8.0), 1 mm dithiothreitol, 0.5 mm EDTA; the recombinant Trx-His6-tag was removed by cleavage with TEV protease (Invitrogen, Carlsbad, CA, USA). The FXR peptides were further purified by IMAC, anion exchange chromatography (MonoQ HR 5/5) and size exclusion chromatography (Superdex 200 HR 16/60 or Superdex 200 10/30) using elution buffer consisting of 50 mm Tris/HCl (pH 8.0), 2 mmβ-mercaptoethanol. The purity of the recombinant peptides was higher than 95% as verified by SDS/PAGE and by mass spectrometry. Protein concentration was measured by UV absorbance at 280 nm using theoretical extinction coefficients calculated by protparam.

Aggregation studies by CD and size exclusion chromatography

CD spectra were recorded using a Jasco J-715 spectropolarimeter equipped with a thermostatted cell holder controlled by a Jasco Peltier element, at different temperatures, over a wavelength range from 260 to 190 nm in quartz cuvettes (Hellma) of path length appropriate to protein concentration of the samples, i.e. 1 mm for 5 μm (0.15 mg·mL−1), 0.2 mm for 15 μm (0.5 mg·mL−1) and 0.1 mm for 30 μm (1 mg·mL−1). Thermally induced denaturation transitions were monitored by CD absorption at 220 nm from 10 to 95 °C, in 1-°C steps and with an equilibration time of 1 min·°C−1. Reversibility was tested by performing an inverse temperature scan. The purified recombinant proteins were in 20 mm Tris/HCl (pH 8.0), 1 mmβ-mercaptoethanol. To monitor progression of protein aggregation over time, protein samples were incubated at 37 °C and CD spectra were recorded at different time points (1, 6, 16, 24, 45, 72 h, 1 week).

Analytical size exclusion chromatography was carried out by injecting 100 μL of samples (30 μm) into a Superdex 200 10/300 GL column.

ThT fluorescence assays

The ThT assays were performed by consecutively incubating at 20, 30, 40, 50, 55, 60 and 65 °C, for 15 min at each temperature without shaking, the purified protein solution diluted to 5 μm in a buffer containing 20 μm ThT, 20 mm Tris/HCl (pH 8.0), 1 mmβ-mercaptoethanol. After the last heating step, at 65 °C, the sample was kept at room temperature (20 °C) and spectra were recorded after 1, 2, 3, 4, 12, 24 and 96 h. Fluorescence was measured using an ISS PC1 (Interconnect Systems Solution) spectrofluorimeter. All measurements were carried out at 20 °C over a 60 s time course with excitation at 440 nm (0.4 nm slit width) and emission at 482 nm (1.5 nm slit width). For each measurement 10 scans were recorded. The measurement of the fluorescence of the reaction buffer treated at 65 °C showed only a weak peak at 520 nm.

Transmission electron microscopy

A sample volume of 4 μL was spotted onto freshly prepared carbon-coated and glow-discharged copper grids (FormVar). Upon adsorption to the grid surface for 30 s, the sample was washed briefly with milli-Q water and subsequently stained with 1% (w/v) uranyl acetate for 30 s. Micrographs of negatively stained areas were taken with a JEOL 1200 transmission electron microscope operating at 100 kV and at a magnification of 27 800× on electron microscope films (Kodak) and developed with Phenisol developer (Ilford) and Hypam fixer (Ilford) for 5 min each.


We thank Steve Martin for help with CD and fluorescence studies, Lesley Calder for support with electron microscopy analysis and Steve Howell for mass spectrometry analysis. We are grateful to Cesira de Chiara and Laura Masino for critical discussion and assistance in graphic elaboration of CD results. We acknowledge support from the MRC (Grant ref. U117584256). Kris Pauwels is the recipient of an EMBO long-term postdoctoral fellowship (ALTF 512-2008).