2.1 Peptide design

Peptides for this study were selected to include different combinations of putative aggregation initiation sites, such as the apolar palindrome A113-GAAAAGA, β-sheet 1 (128–131), the sequence I138-IHF and the putative gatekeeper element helix 1 (144–154). Stabilizing (N153W) and destabilizing (R151G) mutants of PrP helix 1 were designed employing in-house written software [23]. To facilitate spectrophotometric determination of peptide concentrations, a tyrosine residue was added to the amino-terminus of peptides lacking tyrosine in their sequence.

2.2 Peptides

Peptides were purchased from Jerini AG (Berlin, Germany) and from Thermo Electron Corporation (Ulm, Germany) as HPLC-purified freeze dried powder containing trifluoroacetate counter ions. Peptides were protected by an amino-terminal acetyl group and by amidation at the carboxy-terminus to exclude charge effects from free termini. The sequences of the peptides under investigation are summarized in Fig. 1 .

2.3 Aggregation tests

Buffer conditions applied here were optimized prior to this study with respect to lag-time and duration of transitions. Peptides were dissolved in 20 mM sodium acetate buffer containing 20% DMSO p.a. (Sigma–Aldrich, St. Louis, MO, USA) and 0.01% sodium azide to prevent bacterial growth (Fluka, Switzerland), pH 4.5, sonicated three times for 5 min each, centrifuged at 14 000 × g for 1 h, and stored as frozen stocks at −20 °C. After UV spectrophotometric determination of concentration using an ε _mM (280 nm) of 1.28 mM⁻¹ cm⁻¹ (http://www.expasy.org/tools/protparam.html), all peptide samples were adjusted to a concentration of 50 μM, except for peptides encompassing helix 1. Initial absorptions in the samples were 0.0073 AU per residue with a standard deviation of 0.0026 AU. This small scatter – compared to the increases in absorbance measured in sigmoidal transitions upon aggregation – indicates the absence of aggregates at the beginning of measurements. Concentrations of helix 1 peptides were higher in order to be able to observe any transition (huPrP(140–158): 106 μM, huPrP(140–158)N153W: 94 μM; huPrP(140–158)R151G: 82 μM). Higher concentrations were not possible due to the limited solubility of selected peptides. To follow the time course of aggregation, the turbidity at 450 nm was measured every 3–6 h on an Uvikon 930 spectrophotometer (Kontron Instruments, Eching, Germany) for a total of 180 h in sealed 70 μl microcuvettes (Brand, Wertheim, Germany) to prevent volume- and concentration-changes due to solvent evaporation. Prior to each measurement samples were homogenized by careful agitation. Between measurements samples were stored at 23 °C. An identical cuvette filled with buffer without any peptide treated exactly as the samples was used a reference in turbidity measurements. Before each measurement, the photometer was zeroed using a cuvette with peptide free buffer solution.

2.4 Data evaluation

In order to obtain an estimate of the aggregation tendencies of the peptides, we have determined the approximate lag-time, the apparent rate of aggregation, and the midpoint of the aggregation transition. Lag-time was defined as the duration of approximately linear behaviour of the turbidimetric data and was determined by visual inspection. The approximate lag-time reported corresponds to the last data point prior to the aggregation transition. Assuming exponential growth data points past the lag-time were fit to an exponential growth function

where represents the first data point past the approximate lag-time, B the parameter corresponding to the signal amplitude, and t the time in hours. k was taken as the apparent rate of aggregation. The midpoint of the aggregation transition, t _aggr was obtained by subtracting the approximate lag-time from the midpoint of the sigmoidal transition t _m. t _m was obtained by fitting the observed data to the sigmoid function

where ΔA ₄₅₀ represents the difference in turbidity between the initial and the final equilibrium state, giving a measure for the extent of aggregation, t denotes the time in hours, and b represents the width of the sigmoidal transition. Fitting was performed using the program SigmaPlot 2001 (SPSS Inc., Chicago, IL, USA).

2.5 Whole mount electron microscopy

Samples from turbidity assays were stored for additional three weeks at 4 °C to allow fibril formation to complete. For whole mount electron microscope (EM) samples were spun down at 18 000 × g for 1 h. Five microlitres from each tube were applied to a pioloform-coated copper grid (Plano GmbH, Marburg, Germany) and allowed to adhere for 10 min. The grids were stained with 1% aqueous uranyl acetate for 1–2 min. Samples were examined in a Zeiss 902A transmission electron microscope (Carl Zeiss, Oberkochen, Germany) operated at 80 kV. Micrographs were taken using SO-163 EM film (Kodak, Rochester, NY).

2.6 Congo Red (CR) staining

CR staining of fibril preparations was performed in a variation of Puchtler's method after turbidity measurements. Fibril containing samples were centrifuged at 10 000 × g for 30 min, the pellets were suspended to HistoBond slides and allowed to dry at 50 °C. The dried samples were stained with saturated CR solution in 80% ethanol containing 1 M NaCl for 20 min. Unbound CR was removed by washing in distilled water followed by rinsing with 80% ethanol. The stained and dried samples were subsequently embedded in gummi arabicum and visually inspected for birefringence under a polarization microscope (B1-220, MOTIC, Wetzlar, Germany) equipped with a SLR camera (Canon 350-D, Japan. Note. The colours seen are slightly distorted due to the inability to perform correct white balance with the camera mounted to the microscope; pictures were only γ-corrected to avoid further distortions of colour).

2.7 Calculation of aggregation propensities

Theoretical calculations of aggregation propensities of various peptide sequences and the influence of selected residues within these sequences have been carried out with the software TANGO (http://tango.embl.de) at a pH of 4.5 [24, 25].

3.1 The sequence immediately preceding helix 1 has a higher aggregation propensity than β-strand 1 (see Table 1)

huPrP(128–136), which includes β-strand 1 and which was used here as paradigmatic model for β-strand self-association, showed a clear transition with a t _aggr of 28 h and an average apparent rate k of 0.024 h⁻¹. The difference in absorbance was between 0.1 and 0.15 AU. Such a clear transition was expected as single β-strands have a well-known tendency for self-association. huPrP(137–145), which corresponds to the loop between β-strand 1 and helix 1, displayed a much faster course of aggregation with a t _aggr of approximately 7–8 h and an apparent rate k of 0.15 h⁻¹ and a slightly reduced approximate lag-time. Both peptides are of the same length, and each of them contains one possible initiation site for aggregate formation: β-strand 1, which is part of the huPrP(128–136) peptide, has been implicated in fibril formation in studies on the peptide huPrP(118–135). huPrP(137–145) contains a region of bulky apolar residues immediately preceding helix 1 (I138-IHF), which showed a tendency to adopt extended conformations in NMR studies of peptides containing the helix 1 region of huPrP [27]. While huPrP(128–135) shows fine fibrillar aggregates, the aggregates of huPrP(137–145) appear to be highly ordered, in a nearly crystalline structure (see Fig. 2 ).

Despite being experimentally identified as a putative aggregation association site, this sequence is not detected by TANGO algorithm, which predicts aggregation propensities. This apparently is due to the positive charge of the His residue at the pH of 4.5. At higher pH, when histidine is uncharged, a small aggregation propensity is detected by the TANGO algorithm. Exchange of Tyr, which has similar steric requirements, for His, however, results in a high calculated aggregation propensity (see Table 2 A). Therefore, a comparison of several species specific polymorphism contained in this region with TANGO was performed on the IIYF background. Interestingly, I138-I (human) and L138-I (bovine and ovine) exhibited a comparable aggregation score in TANGO, while M138-I (mouse) and M138-M (syrian hamster) showed a significant decrease in the calculated aggregation propensity, with M138-M being the least aggregation prone sequence (see Table 2B). This might at least partially account for the difficulty of experimental TSE transmission from bovine and human sources to hamsters and mice. The exposed structural position of the IIHF-sequence further stresses its possible importance in the aggregation initiation and interaction between PrP^C and PrP^Sc (see Fig. 1). In fact, a recent study by Vanik et al., who used peptides corresponding to the sequences 23–144 of man, mouse, and hamster, revealed the same order of aggregation susceptibility as suggested by TANGO. Also, cross seeding experiments could demonstrate the importance of this sequence for the species barrier [29]. Further, the IIHF-sequence was identified as contact site between different prion proteins as an early step in the transconformation process in a recent molecular dynamics study published by DeMarco and Daggett [30]. As pointed out earlier, the structural conversion of PrP^C to PrP^Sc likely involves an order-to-disorder-to-order transition of the sequence prior to helix 2. In view of recent result it appears likely that aggregation actually starts from partially disordered states [31, 32], which are present even under equilibrium conditions [33-35]. It has been shown that sequences with neighbouring large, hydrophobic amino acids favour extended, β-like conformations even under strongly denaturing conditions [20]. Thus, even in the disordered state, I138-IHF is likely to populated sheet-like conformations due to intrinsic amino acid properties. Indeed, such a conformational preference could be detected in peptide models [27]. Therefore, I138-IHF can serve as an aggregation site either for homo- or for hetero-sequence aggregation, e.g., with the alanine-rich sequence [30].

A – Aggregation propensity of the IIHF-region: influence of the charge pattern on aggregation propensities. B – Effects of species specific sequence variations: I138 and I139 vary considerably among species and have a strong effect on the aggregation propensity (Note. Variant H140Y has been used to estimate species specific effects; see panel A). C – Influence of single residue substitutions on the aggregation propensities of the sequence huPrP(113–131) and selected sub-sequences (huPrP(A113–A120), huPrP(V121–V122), and huPrP(G123–G131), which includes β-strand 1). It shall be mentioned that the isolated sequences A113-GAAAAGA and Y128-MLG (strand 1) per se do not have any aggregation propensity in TANGO.

3.2 Helix 1 has no tendency to form aggregates, it can delay but not prevent fibril formation in longer peptides

Further sigmoidal transitions could be observed for peptides huPrP(128–145), huPrP(137–145) and huPrP(137–157), although the differences in absorbance were less pronounced, which may be related to differences in morphology and size between peptide aggregates. In contrast to the abovementioned peptides, peptides consisting solely of helix 1 (huPrP(140–158) and its stabilized variant N153W), excluding the flanking sequences, showed no aggregation tendency whatsoever. Only a destabilized variant, huPrP(140–158)-R151G, showed a very slow, shallow transition which could be observed after 101 h (see Table 1). Here, the intense network of salt-bridges and favourable electrostatic interactions within helix 1 is distorted, obviously enabling aggregation, presumably by interaction of aromatic residues.

However, adding the sequence of helix 1 to strongly aggregating peptides like huPrP(128–145) did not abolish fibril formation, but nevertheless delayed the formation of aggregates to an average lag-time of 30 h for huPrP(137–157), compared to 8 h for huPrP(137–145), and 16 h for huPrP(128–145), which is approximately of the same length. The growth phase of aggregates measured by t _aggr, however, is much longer (see Table 1).

This leads to the possibility that helix 1 might remain in a helical conformation during the early stages of the transconformation process. This is again supported by the study of molecular dynamics data recently published by DeMarco and Daggett, which showed no conversion of helix 1 during the simulated protofibril formation [30]. Later in the course of the conformational transition, facilitated by conformational changes in its surroundings, the energetically unfavourable conversion of helix 1 into an extended structure could take place. This is in line with studies showing the existence of helical PrP dimers in the early stages of the SDS induced prion transconformation, which later form higher oligomers rich in β-sheet elements [36]. As previously hypothesized [27], helix 1 might play the role of a “gatekeeper” element in the aggregation process, by acting as an energetic barrier for the transconformation of PrP.

3.3 The alanine-rich sequence requires flanking parts of the sequence to form fibrillar aggregates

The palindromic alanine-rich sequence A113-GAAAGA is believed to be causal for neuro-inflammation induced by the peptide PrP(106–126). Interestingly, huPrP(106–126) did not show an observable transition that would indicate aggregate formation within 180 h under our conditions.2 The same holds true for huPrP(110–120), which contains almost exclusively the alanine-rich region, and for huPrP(120–127), which contains additional hydrophobic residues. Prolonged incubation of huPrp(106–127) and huPrP(110–127) finally leads to aggregate formation. The aggregates of huPrP(106–127) exhibit a typical fibrillar morphology in the ER, almost resembling a disordered arrangement of protofibrils, whereas huPrP(110–127) forms amorphous aggregates (see Fig. 2). Regardless of the morphology, in both cases the aggregates showed the typical red-green birefringence in a CR stain, indicating the presence of amyloid (see Fig. 2). However, it was not possible to perform CR-staining nor to detect aggregates in EM-micrographs for huPrP(110–120).

These results are somewhat surprising, as several other studies implicated the apolar palindromic sequence AGAAAAGA in fibril formation and cytotoxicity of peptides around the 110–120 region [12, 37, 38]. Nevertheless, our data do not contradict the importance of the palindromic sequence for the conformational transition of PrP. Molecular dynamics studies showed the importance of a template for the re-folding of this sequence [30], whereas our studies only tested for auto-aggregation within the same sequence. In particular, protofibril models derived from such simulations showed a β-core made up from residues 116–119, adding as a third strand to the native β-sheet, the elongated native sheet itself and another strand containing residues 135–140, underlining the importance of I138-IHF for the pathogenic process. Importantly, inter-prion contacts in this model are formed between residues 116–119 (alanine-rich) and residues 135–140, corresponding to the above proposed IIH(F) aggregation site.

Extending the 110–126 peptide on the amino terminal side to form huPrP(106–126), however, restores the alanine-rich peptide's (huPrP(110–120)) ability to form amyloid-like fibrillar plaques (see Fig. 2). This points to a role of the 106–110 region as another possible initiator site, which is further underscored by the importance of the 90–112 region for prion infectivity [39]. So, while the palindromic region might not be able to form amyloid fibrils by itself, it may be included into fibrillar structures, whose formation is initiated by other regions of the peptide. Further evidence for this behaviour could be found in TANGO calculations of aggregation propensities (see Table 2C). The isolated AGAAAAGA showed a low TANGO score due to the presence of flexibility increasing Gly residues [20], which have also been shown to be evolutionary conserved in order to prevent aggregation in other cases [40]. However, elongation of the sequence to AGAAAAGAVV, as well as the introduction of the pathogenic A117V mutation [41] increased the calculated aggregation tendency of the peptide indicating that the aggregation propensity is introduced by the valine residues. An explanation for this behaviour can be found in the conformational propensity of the palindromic region. Conformationally flexible, short Ala-rich peptides are known to preferentially populate polyproline type II-like conformations, which are close to extended conformers in the Ramachandran space [42]. The presence of rather bulky Val residues, however, might shift the conformational distribution in the direction of β-sheets [43]. The propensity for poly-proline II (PPII) like conformations also offers an explanation for the need for a template in the aggregation process, as PPII helices show an intrinsic tendency to interact and form coiled-coils, such as in collagen. Proline rich peptides, such as iPrP13 (DAPAAPAGPAVPV), which has been designed to inhibit aggregation into β-sheets [44], could therefore present an ideal template. Indeed Soto et al. could show that iPrP13 interacts with the AGAAAAGA palindrome and interferes with the prion transconformation – possibly by presenting a stronger interacting template for the palindrome than natural interaction partners like I138-IHF. Further experimental evidence can be deduced from the work of Chabry et al., who found that peptides encompassing the palindromic region (113–141, 115–141, 117–141 and 119–141) could inhibit the conversion of full length prion protein in a dosis-dependent manner, while the only marginally shorter peptide 121–141 showed no inhibitory effect [45].

In summary, our study results in two major findings, namely that the bulky, apolar, and highly surface-exposed I138-IHF sequence is a possible initiation site for aggregation and that the aggregation properties of the neurotoxic peptide huPrP(106–126) are centred at residues 106–110 and VV(121–122), respectively. Recent molecular dynamics simulations of the PrP^C to PrP^Sc conversion [30] as well as studies concerning the species barrier [29] also point to the importance of those regions in the aggregation initiation process. The crucial role of these structural elements is further underscored by the existence of several pathogenic mutations. Our TANGO calculations indicate that the important M129V polymorphism, which is located in β-strand 1 and is responsible for phenotypic heterogeneity of prion diseases [46] and susceptibility for the classical and new variant Creutzfeldt-Jakob Disease (CJD) [47], changes the aggregation propensity in the VV(121–122) region. In addition, the Gerstmann–Sträussler–Scheinker syndrome (GSS)-causing G131V mutation, also located in β-strand 1, could be shown to increase local conformational plasticity, thereby facilitating prion transconformation [47]. Thus, if β-strand 1 becomes exposed during structural rearrangements, it may also efficiently enhance aggregation as shown here. The only amyloidogenic element described here which does not harbor any pathogenic mutations is the I138-IHF sequence. However, the differences in sequence between various species suggest a role of this region in the species barrier. We have shown that IIHF has the ability to aggregate fast compared to, e.g., β-strand 1. With β-strand 1 being hidden in PrP^C, this makes the partially surface-exposed I138-IHF motive the most important identified aggregation promoting element in PrP. In combination with recent data from the literature a concerted operation of several amyloidogenic elements appears a likely mechanism in formation of inter-prion protein contacts. Interaction of different elements may also explain the conformational heterogeneity observed in prion strains.