Competing intrachain interactions regulate the formation of β-sheet fibrils in bovine PrP peptides

Abstract

At the heart of the pathogenesis of transmissible spongiform encephalopathies (TSEs), such as BSE, scrapie, and Creutzfeldt–Jakob disease, lies a poorly understood structural rearrangement of PrP, an abundant glycoprotein of the nervous and lymphoid systems. The normal form (PrP^C), rich in α-helix, converts into an aberrant β-sheet-dominated form (PrP^Sc), which seems to be at the center of the pathotoxic symptoms observed in TSEs. To understand this process better at a molecular level, we have studied the interactions between different peptides derived from bovine PrP and their structural significance. We show that two unstructured peptides derived from the central region of bovine PrP, residues 115–133 and 140–152, respectively, interact stoichiometrically under physiological conditions to generate β-sheet-dominated fibrils. However, when both peptides are incubated in the presence of a third peptide derived from an adjoining α-helical region (residues 153–169), the formation of β-sheet-rich fibrils is abolished. These data indicate that native PrP^C helix 1 might inhibit the strong intrinsic β-sheet-forming propensity of sequences immediately N-terminal to the globular core of PrP^C, by keeping in place intrachain interactions that would prevent these amyloidogenic regions from triggering aggregation. Moreover, these results indicate new ways in which PrP^Sc formation could be prevented.

A broad range of human diseases, including disorders such as systemic amyloidosis, Alzheimer’s disease, and transmissible spongiform encephalopathies (TSEs), are characterized by the accumulation of abnormal proteinaceous deposits known as amyloid (Tan and Pepys 1994; Kelly 1996, 1997; Dobson 1999). These structures are all formed by protease-resistant, long, unbranched protein fibrils of several microns in length and 4–12 nm in diameter (Booth et al. 1997; Sunde et al. 1997). Amyloidogenic proteins show no significant sequence or structural similarities. It is remarkable, however, that despite this fact all amyloid fibrils share a characteristic cross-β core structure, in which the β-strands are perpendicular to the long axis of the fibrils (Booth et al. 1997; Sunde et al. 1997). It is generally agreed that, in order to accommodate into such structures rich in β-sheet, the protein precursor must undergo at least partially unfolding from its native or physiologically active state (Kelly 1997; Cohen and Prusiner 1998; Dobson 1999). Recent studies show that many proteins that are not associated with disease can also be induced to form amyloid fibrils in vitro under appropriate conditions, implying that amyloid fibril formation can be an intrinsic generic property of polypeptide chains (Guijarro et al. 1998; Litvinovich et al. 1998; Dobson 1999; Gross et al. 1999; West et al. 1999; Villegas et al. 2000). Moreover, some of these disease-unrelated proteins have shown to be toxic when in an aggregated form, indicating the existence of common pathological pathways in diseases related to protein deposition (Bucciantini et al. 2002).

The observed intrinsic toxicity of protein aggregates independent of their sequence confirms that protein aggregation is a common problem that living organisms have to deal with to survive, and points toward the existence of evolutionary pressure to avoid protein aggregation (Chiti et al. 1999; Dobson 1999). Many factors must be involved in this protective mechanism. It is likely that the development during evolution of cellular elements that improve folding efficiency and decrease aggregation, but also, and probably even before that, the selection of sequences that can fold efficiently into a globular form in which the polypeptide chain and the hydrophobic residues are hidden in the interior, could have played an important role in avoiding aggregation (Chiti et al. 1999; Dobson 1999). One way of promoting or stabilizing such native forms would be the parallel evolution of neighboring sequences whose interactions successfully competed with those that might lead to aggregation.

We have tested these ideas using the PrP protein. PrP constitutes an excellent model to study, given its connection with several pathological syndromes, such as CJD, GSS, scrapie, and BSE, through the formation of a misfolded infective and cytotoxic aggregated form termed PrP^Sc (Cohen and Prusiner 1998). The PrP protein is structurally divided into two main regions. The C-terminal half of the protein folds into a globular or compact domain mainly α-helical in nature, of which a number of structures are available (Riek et al. 1996; Donne et al. 1997; Lopez Garcia et al. 2000). The N-terminal half, however, seems to be mostly unfolded and retains some interesting properties, such as metal ion binding, which might be important in PrP biological function (Brown et al. 1997).

The residues located immediately before the globular or structured domain of PrP confer important properties on the protein that seem to be related to its infective and pathological cycle. In this way it has been suggested that at least some of the residues at the N terminus of the PrP structured domain might constitute, or be involved in, the barrier that prevents the transmission of PrP^Sc between certain species (Kaneko et al. 1995; Supattapone et al. 2001). This region seems also to be important in conferring typical PrP^Sc-like properties (prion replication or infectivity, cytotoxicity, and formation of proteinase-resistant aggregates), as suggested from studies on the PrP106 and PrP61 proteins or "miniprions" in which major regions of the protein have been removed (Supattapone et al. 1999, 2001; Baskakov et al. 2000). Moreover, studies carried out on peptides indicate that residues 106–126 might be involved in conferring neurotoxic properties to the prion protein (Forloni et al. 1993, 1996; Hope et al. 1996; Brown 2000). Previous studies have shown that a peptide corresponding to the Syrian hamster PrP residues 109–122 (SHa 109–122) spontaneously forms amyloid structures when incubated in vitro (Gasset et al. 1992). Moreover, this sequence was also able to induce the conversion of the otherwise unstructured SHa 104–122 peptide into amyloid structures (Nguyen et al. 1995a). Furthermore, a peptide spanning the entire region corresponding to peptides 104–122 and 129–141 (SHa 90–145) forms characteristic cross-β fibrils when synthesized as a single peptide (Nguyen et al. 1995b; Inouye et al. 2000).

Based on previous evidence on the behavior of small peptide sequences expanding the central region of the prion protein (end of unstructured region and beginning of the globular domain), together with the available structural information on the murine (Riek et al. 1996; Donne et al. 1997), human (Zahn et al. 2000), and bovine (Lopez Garcia et al. 2000) PrP proteins, we have devised a set of peptides that correspond to an expanded portion of the central region of bovine PrP, to investigate their structural properties both isolated and when incubated in combination. Our results show the presence of interactions between the different peptides that modify their structural properties and most importantly their aggregation behavior. In this context, we postulate that competing intrachain interactions might modulate the intrinsic propensity of the PrP protein to aggregate and perhaps to develop its pathologic characteristics.

Results

Design of peptides

Four different peptides based on the bovine PrP sequence were chosen to perform this study. The peptides are 115B1 (KPKTNMKHVAGAAAAGAVV), B1 (MKHVAGAAAA GAVV), B2 (MLGSAMSRPLIHF), and Hx1 (GSDYED RYYRENMHRYP). In Figure 1 ▶, the location of these within the central portion of the bovine PrP sequence and their correspondent homologous regions in the human and Syrian hamster sequences are illustrated. The numbering of residues within bovine PrP follows that published for isoform 1 (with 6 octarepeats; Goldmann et al. 1991), SwissProt accession no. P10279. The peptides 115B1, B1, and B2 are the bovine equivalent of the hamster peptides 104H1, H1, and H2, respectively, described by Nguyen et al. (1995a). Two of the chosen peptides (115B1 and B1) comprise some of the residues believed to exert specific neurotoxic effects (residues 106–126; Forloni et al. 1993^{, 1996;} Hope et al. 1996^; Brown 2000), whereas the other two correspond, respectively, to regions located between β-strand β-1 and helix 1 of the PrP protein (B2), and helix 1 plus flanking regions (Hx1).

Figure 1.

Peptides used in this study and positioning in the PrP protein sequence. The sequences of bovine, human, and Syrian hamster PrP in the region under study are aligned. Because of N-terminus differences between species, the residue numbering differs, and that of bovine PrP follows that of Goldmann et al. (1991). Peptides used in this study are indicated above and conserved elements of secondary structure below the sequence alignment. The position of the neurotoxic, fibrillogenic peptide 106–126 (Selvaggini et al. 1993) is also indicated.

Structural characteristics of peptides derived from the junction region of bovine PrP

The peptides mentioned above were analyzed by different techniques to characterize their structural properties and their propensity to generate aggregates. The CD spectra of peptides 115B1, B1, and B2 under physiological conditions were characterized by a negative band at 198 nm and no positive peak (see Fig. 2A ▶), which is typical for unordered structures (Johnson 1990). FTIR amide I spectra of peptides 115B1 and B2 showed bands centered at 1648 cm⁻¹, characteristic of disordered structures (see Fig. 2B ▶). However, the amide I band of peptide B1 had three other components in addition to a major constituent at 1648 cm⁻¹, two at 1623 cm⁻¹ and 1690 cm⁻¹, which indicate the presence of antiparallel β-sheet structure, and one at 1663 cm⁻¹, which might be indicative of the presence of turns (Krimm and Bandekar 1986).

Figure 2.

Spectroscopic and electron microscopic characterization of prion bovine peptides. (A) CD spectra of 115B1 (dotted line), B1 (broken line), and B2 (continuous line) peptides (2 mM) in 20 mM HEPES, 100 mM NaF (pH 7.2). (B) FTIR amide I spectra of the bovine PrP peptides (2 mM) in ²H₂O buffer (20 mM HEPES, 100 mM NaCl, p²H 7.2). (C,D) Electron micrographs of B1 and Hx1 peptides, respectively. Clusters of fibrils are observed with B1 peptide, whereas Hx1 peptide shows only amorphous aggregates. Bar, 200 nm. (E) CD spectra of peptide Hx1 (1.5 mM) in 20 mM HEPES, 100 mM NaF (pH 7.2) in the presence of increasing concentrations of TFE (0%, 10%, 20%, 40%, 60%, 80%, 90%). (F) The change in the mean residue ellipticity (Balbach et al. 1997) at 208 and 222 nm of Hx1 plotted as a function of TFE concentration. All spectroscopic and electron microscopy observations were made on peptides that had been incubated at room temperature over 3 d.

When observed by electron microscopy (EM), samples of B1 peptide revealed clusters of fibrils of 7-nm diameter (Fig. 2C ▶) similar to those observed previously in samples of the homologous Syrian hamster SHa 109–122 peptide (Gasset et al. 1992). In contrast, samples containing either peptide 115B1 or B2 on their own showed no fibrils when observed under similar conditions (data not shown), in agreement with previous findings on Syrian hamster homolog peptides (Gasset et al. 1992; Nguyen et al. 1995a).

On the other hand, when compared with B1 and B2, the FTIR amide I spectrum of peptide Hx1 shows a shift in its major component toward 1650 cm⁻¹ (Figu. 2B ▶), which is compatible with Hx1 exhibiting a higher α-helical structure content. The additional component observed in the amide I spectrum of Hx1 at 1612 cm⁻¹ may reflect the presence of small aggregates, as detected by EM (Fig. 2D ▶), which formed after several hours of incubation under physiological conditions. When analyzed by CD, peptide Hx1 showed some helical content (minima at 208 and 222 nm) together with some random-coil contribution (minima at 200 nm; see Fig. 2E ▶), confirming the higher helical character already detected by FTIR analysis (see above; Fig. 2B ▶). This helical content increased gradually with the addition of TFE, following a broad sigmoidal pattern, which together with the presence of an isodichroic point at ∼206 nm, indicates a two-state transition between a random-coil and a helical conformation that is highly populated at higher TFE concentrations (Fig. 2F ▶).

The incubation of peptide 115B1 together with peptide B1 in equimolecular amounts shows some increase in intermolecular β-sheet structure as indicated by the presence of two additional components at 1623 and 1690 cm⁻¹ when compared with the arithmetic addition of the amide I FTIR spectra corresponding to the isolated peptides (Fig. 3A ▶). This β-sheet induction is, however, quite small when compared with that observed in Syrian hamster PrP-derived homologous peptides. Furthermore, and in contrast to the findings on the homologous hamster peptides, mixing peptide B2 with B1 in a ratio 1:1 led to a barely detectable decrease in the overall β-sheet content of the mixture, as judged by FTIR spectroscopy (Fig. 3B ▶). Most interestingly, equimolecular mixtures of the two unstructured peptides, 115B1 and B2, which individually had no propensity to form β-sheet structure or fibrils, resulted in a strong induction of β-sheet and loss of random-coil structure, as revealed by the presence of a major amide I FTIR component appearing at 1623 cm⁻¹ and a minor one at 1690 cm⁻¹ (Fig. 3C ▶). Furthermore, the incubation of these two peptides led to the formation of long, twisted fibrils (Fig. 3E ▶). The addition of a third peptide B1 did not have any effect on the behavior of the two peptides 115B1 and B2 as observed by FTIR (Fig. 3D ▶).

Figure 3.

Infrared spectroscopy and electron microscopy characterization of bovine prion peptide mixtures. FTIR spectra in the amide I region of mixed prion peptides (1 mM each peptide) in ²H₂O buffer (20 mM HEPES, 100 mM NaCl, p²H 7.2): (A) 115B1/B2, (B) B1/B2, (C) 115B1/B1, and (D) B1/B2/115B1. In each panel, the continuous line represents the observed spectrum of the peptide mixture, the dotted line the expected spectrum based on the summation of the spectra of the individual peptides in the mixture, and the discontinuous line represents the difference between the observed and the expected spectra. All spectra were recorded after incubation at room temperature for 3 d. (E) Electron micrograph of mixed peptides 115B1/B2; bar, 200 nm. (F) HPLC chromatograms of the fibril-forming peptide mixture 115B1/B2 (main panel) and of individual peptides 115B1 and B2 (inset panel, broken and continuous lines, respectively) on a Jupiter 5-μ C18 300-Å column. The peptides were eluted with a linear AB gradient 0%–50% B in 22 min (eluent A, 0.1% TFA; eluent B, 0.1% TFA in 50% acetonitrile/50% water), 1 mL/min.

We were particularly interested in the composition of the β-rich fibrils formed after mixing peptides B2 and 115B1, as neither of them was able to form fibrils on their own. Accordingly, we pelleted the fibrils, washed them extensively, dissolved them in 100% HFIP overnight, and then subjected them to reversed-phase HPLC separation. The chromatogram of the fibril-forming peptide mixture showed three distinct major peaks (Fig. 3F ▶). Peaks 1 and 2 corresponded to peptides 115B1 and B2, respectively (see Fig. 3F ▶, inset). Integration of these peaks indicates that the fibrils were composed approximately of equal proportions of the two peptides. Peak 3 was only observed in the fibril-forming peptide mixture and not during the analysis of individual peptides. This third peak may be caused by a strong intermolecular interaction between peptides 115B1 and B2 that is resistant to the denaturing conditions used to break down the fibrils prior to HPLC analysis.

Inhibition of β-sheet formation by an adjacent α-helical region

The results above indicate that two neighboring regions of bovine PrP, which are individually unstructured under physiological conditions, have a propensity to form fibrils rich in β-sheet when combined in solution in the absence of other PrP sequences. Why, then, does this not happen more readily in the context of the whole protein? We hypothesized that interactions between one of these regions, corresponding to the peptide B2, and the adjacent α-helical region (helix 1) in the native structure might inhibit the nonnative, β-sheet-forming interactions. Accordingly, we simultaneously incubated peptide Hx1 (which includes helix 1) together with B2 and 115B1 in equimolecular amounts and under physiological conditions. The results show the absence of the β-sheet FTIR components that are otherwise observed in mixtures B2:115B1 (i.e., 1623 and 1690 cm⁻¹) when Hx1 is added to the mixture (Fig. 4 ▶), together with a higher absorbance at frequencies close to 1650 cm⁻¹, which might indicate stabilization of helical structure. These results clearly indicate that Hx1 strongly inhibits the formation of β-sheet between the peptides B2 and 115B1, stabilizing at the same time its helical structure. In agreement with this observation, fibrils were virtually undetectable when samples containing equimolecular mixtures of the three peptides were analyzed by EM (data not shown).

Figure 4.

Effect of Hx1 peptide on interactions between peptides 115B1 and B2 monitored by Infrared spectroscopy. Amide I FTIR spectrum of an equimolar mixture of peptides Hx1, B2, and 115B1 (1 mM each) in ²H₂O buffer (20 mM HEPES, 100 mM NaCl, p²H 7.2). The continuous line represents the observed spectrum. The dotted line represents the spectrum expected if there were no interaction between peptide Hx1 and the fibrillogenic mixture (115B1 and B2). This was calculated by summation of the FTIR spectrum of the Hx1 peptide alone and that of the mixture of 115B1 and B2 (see Fig. 3 ▶). The discontinuous line represents the difference between the observed and the expected IR spectra. All IR spectra were recorded after incubation of the three peptides at room temperature for 3 d.

Discussion

In this work, we have studied the intermolecular interactions of isolated peptides corresponding to central sequences of bovine PrP to model interactions that might occur within the native polypeptide chain. The peptides were derived from the region of bovine PrP that links the N-terminal unstructured region with the C-terminal globular domain, which is believed to participate in the major conformational changes that promote the conversion to PrP^Sc and the onset of prion pathogenesis (Kocisko et al. 1995; Supattapone et al. 1999, 2001; Baskakov et al. 2000). Other groups have shown previously that single peptides derived from region 90–145 of hamster PrP can form fibrils rich in β-sheet (Gasset et al. 1992; Nguyen et al. 1995b; Inouye et al. 2000) reminiscent of those found in TSE amyloid plaques. Here we show that fibril formation can occur by interaction between nonfibrillar peptides derived from the homologous region of bovine PrP. To our knowledge, this is the first time that the formation of fibrils rich in β-sheet has been reported to occur by combining two otherwise unstructured peptides under physiological conditions. Bearing in mind that these peptides (115B1 and B2) correspond to adjacent regions of the bovine PrP protein, these results imply that nonnative intramolecular interactions can produce potentially pathogenic outcomes. The question, then, is how is this deleterious process avoided in the native structure?

When we inspect the NMR-determined structure of bovine PrP (Lopez Garcia et al. 2000), we observe that the regions corresponding to peptides B2 and Hx1 are closely associated in the native protein. In Figure 5 ▶, we have attempted to schematize the structure to illustrate the general arrangement of the residues under discussion. For example, Leu 149 in B2 (equivalent to Ile 138 in human PrP and Met 138 in mouse and Syrian hamster PrP, respectively) interdigitates between the Tyr 161 and Arg 162 side chains (Tyr 150 and Arg 151 in other species) in the proximal face of helix 1 (Fig. 5 ▶). Moreover, a hydrogen bond is formed between Pro 148 and Tyr 161, whereas Arg 147 makes close contacts with Tyr 161 and Phe 152 with Glu 157. In the native protein, the interactions between region B2 and helix 1 are further stabilized by the short antiparallel β-sheet formed between strands β1 and β2, located at the N terminus of B2 and at the residues immediately C-terminal of Hx1, respectively, and also by the backbone linkage of the C terminus of B2 to Gly 153 at the N terminus of Hx1. These interactions with helix 1 in the native state, therefore, can be seen to inhibit the natural tendency of region B2 to form a β-sheet with region 115B1.

Figure 5.

Schematic structure of native bovine PrP, illustrating the potential interactions between peptides used in this study. The two-dimensional representation is based on the NMR structure (Lopez Garcia et al. 2000) but with displacement and simplification for clarity. Amino acid residues are illustrated on the structure within the region of interest, with those that differ between species underlined and italicized. Peptides used in this study are indicated by jointed lines.

This interaction could be destabilized by mutation, reduction in pH, denaturation, and so on, facilitating the formation of intramolecular β-sheet structure between the adjacent B2 and 115B1 sequences. Of itself, this might not be pathogenic, as the nonnative protein could be rapidly turned over. However, our results show that the B2–115B1 interaction can promote the formation of a macromolecular fibril as a result of alternating heterotypic interactions. In the intact protein, this process might correspond to the destabilization of B2–Hx1 interaction in one molecule by interaction with the abnormal 115B1–B2 β-sheet in another, leading to the propagation of misfolding from one monomer to another, which eventually might result in the formation of protease-resistant, pathogenic PrP^Sc. It is likely that this would happen through the formation of an initial nucleus stabilized by intermolecular interactions, from which amyloids and prion macromolecular aggregates seem to form (Harper and Lansbury 1997).

Nevertheless, one must avoid overinterpreting this result in too simplistic a manner. Although recent work has shown that the thermodynamically favored β-forms of PrP are normally precluded by the kinetically favored pathway of α-form folding, and that denaturing conditions at low pH can tip the balance in favor of the former (Baskakov et al. 2001), the details of prion misfolding and aggregation remain largely unknown at the molecular level. What is becoming clear is that PrP is capable of several, alternative pathways of misfolding, some of which may relate to strain-specific differences in the pathological properties of prions, whereas others may be in vitro artifacts (Peretz et al. 2001). Recently, it has been reported that fully oxidized PrP is able to form two distinct nonnative isoforms: a β-oligomer and an amyloid isoform (Baskakov et al. 2002). The former is not considered an intermediate on the pathway to fibrillar aggregation, nor a substructure of the amyloid isoform; in fact, it must unfold and refold to give the latter. Although the formation of amyloid-like fibrils from the peptides described here is highly indicative, it would be premature to conclude definitively whether the interactions we see represent those that occur in amyloid formation in vivo, rather than in β-oligomer or other aberrant structures.

Our findings, however, help to provide a molecular explanation for the outcome of experiments in which truncated forms of PrP, in which the region corresponding to Hx1 has been deleted, but in which the region corresponding to 115B1 and B2 has been retained, showed an increased tendency to form protease-resistant, β-sheet-dominated fibrils in vitro, to support PrP^Sc formation in vivo and to be neurotoxic (Supattapone et al. 1999, 2001; Baskakov et al. 2000). They are also consistent with the hypothesis that disruption of the salt bridges stabilizing helix 1 might promote the formation of the abnormally folded PrP^Sc (Morrissey and Shakhnovich 1999; Speare et al. 2002).

As a corollary, one would predict that agents or sequence modifications that destabilized the pathogenic interaction between regions B2 and 115B1 or that stabilized the native-like interaction between Hx1 and B2 would prevent the formation of PrP^Sc in vivo. The challenge now is to identify the specific contacts between B2 and 115B1 in fibrils and to develop structure-based approaches to the inhibition of their interaction. In addition, this work also suggests new approaches for the design of protein variants resistant to aggregation and amyloid formation complementary to other strategies previously reported based on the stabilization of secondary structure motifs (Villegas et al. 2000).

Materials and methods

Peptide synthesis and purification

The bovine PrP peptides used in this study: 115B1 (KPKTNMK HVAGAAAAGAVV), B1 (MKHVAGAAAAGAV), B2 (MLGS AMSRPLIHF), and Hx1 (GSDYEDRYYRENMHRYP), were synthesized on an Applied Biosystems 430 automated peptide synthesizer. Peptides were purified to homogeneity by reverse-phase HPLC as described elsewhere (Yang et al. 1994). Because residual TFA gives rise to a strong absorption band near 1670 cm⁻¹ in the infrared spectrum that overlaps with the conformation-sensitive amide I band of the peptide backbone, all trifluoroaceate counterions were replaced by chloride ions by lyophilization three times from 50 mM hydrochloric acid. The calculated molecular mass of the peptides was verified by electrospray mass spectrometry. Peptides were maintained in stock solutions in hexafluoroisopropyl alcohol (HFIP), and the concentration was quantitated by amino acid analysis.

Circular dichroism

Circular dichroism (CD) spectra of single peptides were recorded on a Jasco J720 spectropolarimeter (Jasco UK) using quartz cuvettes of 0.01 cm path length, from 195 to 250 nm at room temperature. Peptide concentrations were 1.5 mM to 2 mM in 20 mM HEPES, 100 mM NaF (pH 7.2). The use of NaF avoided excessive light absorption by chloride ions below 200 nm (Nguyen et al. 1995a). Qualitative secondary structure assignments were based on the following: α-helix, minima at 208 and 222 nm, maximum at 190 nm; β-sheet, minimum at 218 nm, maximum at 195 nm; random coil, minimum at 198 nm, no positive peak (Johnson 1990).

Infrared spectroscopy

FT-IR spectra for single peptides as well as for peptide mixtures were recorded on a Bio-Rad FTS 175C spectrometer (Bio-Rad Laboratories Europe) equipped with an MCT detector cooled with liquid nitrogen, and purged with a continuous flow of nitrogen gas. Peptide samples (at concentrations similar to those described above) were prepared in ²H₂O, 20 mM HEPES, 100 mM NaCl, p²H 7.2 (electrode reading corrected for isotope effects). Samples were placed between a pair of CaF₂ windows separated by a 50-μm Mylar spacer. For each sample, 256 interferograms were collected at 2 cm⁻¹ spectral resolution. Buffer spectra obtained under similar conditions were subtracted from all the samples before analysis. Second-derivative analysis of the amide I band was used to locate the frequencies of the different spectral components. Spectra were processed using WIN-IR software (Bio-Rad).

Reversed-phase high performance liquid chromatography

Analytical chromatographic measurements were performed on a Gilson HPLC system controlled by Unipoint software. Runs were carried out on a Jupiter RP C18 column (5 μm, 300 Å, 4.6 × 150 mm i.d.), and the peptide mixture B2/115B1 in 100 μL and 1 mM each was prepared as for FTIR spectroscopy. After 3 d incubation at room temperature, the peptide mixture was ultracentrifuged for 1 h at 100,000g and 20°C in a TL-100 centrifuge using a TLA-100 rotor (Beckman Coulter UK, Ltd.). The resulting pellet was washed three times with 100 μL of buffer (20 mM HEPES at pH 7.2). The final pellet was resuspended in 200 μL of HFIP and incubated overnight to allow fibril dissociation. Just before injection, the HFIP was evaporated, and the resulting pellet was dissolved in a mixture of water, acetonitrile, and acetic acid (4:4:1). Separations were performed at an eluent flow rate of 1 mL/min. The mobile phase A was 0.1% TFA in water, and the mobile phase B was 0.1% TFA in 50% acetonitrile. Retention times were determined using a linear gradient 0%–50% B from 2 to 22 min. Individual peptides were analyzed under the same chromatographic conditions, using 120 nmoles for each injection. The retention times of single peptides were compared with those obtained from the analysis of the mixture.

Electron microscopy

Suspensions of peptides from preparations used for FTIR or CD spectroscopy (3 μL) were applied to carbon-coated copper grids, blotted, negatively stained with 1% uranyl acetate, air dried, and then examined in a JEOL JEM1010 transmission electron microscope, operating at 80 kV.