Sequence-Dependent Self-Assembly and Structural Diversity of Islet Amyloid Polypeptide-Derived β-Sheet Fibrils

Determining the structural origins of amyloid fibrillation is essential for understanding both the pathology of amyloidosis and the rational design of inhibitors to prevent or reverse amyloid formation. In this work, the decisive roles of peptide structures on amyloid self-assembly and morphological diversity were investigated by the design of eight amyloidogenic peptides derived from islet amyloid polypeptide. Among the segments, two distinct morphologies were highlighted in the form of twisted and planar (untwisted) ribbons with varied diameters, thicknesses, and lengths. In particular, transformation of amyloid fibrils from twisted ribbons into untwisted structures was triggered by substitution of the C-terminal serine with threonine, where the side chain methyl group was responsible for the distinct morphological change. This effect was confirmed following serine substitution with alanine and valine and was ascribed to the restriction of intersheet torsional strain through the increased hydrophobic interactions and hydrogen bonding. We also studied the variation of fibril morphology (i.e., association and helicity) and peptide aggregation propensity by increasing the hydrophobicity of the peptide side group, capping the N-terminus, and extending sequence length. We anticipate that our insights into sequence-dependent fibrillation and morphological diversity will shed light on the structural interpretation of amyloidogenesis and development of structure-specific imaging agents and aggregation inhibitors.

species, amyloid peptides/proteins display structural diversity under certain physicochemical conditions, which is related to a variety of disease strains and phenotypes. For example, distinct "daughter" fibril morphologies and neuronal toxicities of amyloid β (Aβ) associated with Alzheimer's disease have been observed through seeding with selective "parent" seeds, which may have implications for different strains or pathological manifestations. 8 Understanding amyloid pathways and their possible conformations that could lead to pathogenesis and degeneration is a vital area of research.
Similar to naturally occurring proteins, short amyloid fragments can also exhibit polymorphic β-sheet structures when subtle structural modifications are applied. 9−11 This includes variation of environmental conditions and changing of amino acid sequences (e.g., single atom replacement of hydrogen to halogen, incorporation of β-amino acids or aromatic residues, rearrangement of the hydrophobic/hydrophilic residues). 12−21 This is due to the multiple noncovalent interactions involved in determining a cross-β structure such as hydrogen bonds (H-bonds), hydrophobic interactions, electrostatic attractions, van der Waals forces, and π−π interactions. 4,5,12,13,22−26 For example, Castelletto et al. showed that the formation of helical ribbon nanostructures by substitution of the α-alanine residues with β-alanine in the extended Aβ variant sequence (AAKLVFF), which was different to the cylindrical fibrils (KLVFF) or twisted tapes (AAKLVFF) reported in other works. 17 Hu et al. reported the important role of electrostatic repulsions between terminal charges in mediating the transformation from twisted ribbons into laminated nanobelts. 12 Coassembly of oppositely charged amyloid-inspired peptides into cylindrical nanostructures based on electrostatic attraction was reported by Cinar and colleagues. 25 Liang et al. reported a morphological transition of Aβ (16−22), Ac-KLVFFAE-NH 2 , from hollow nanotubes into fibers following the changes in β-strand registry at different pH conditions. 20 Because amyloid fragments share similar selfassembly mechanisms with natural proteins, an investigation of their structural diversities will contribute to the understanding into amyloid-related biological processes.
In this work, we sought to determine the effect of amino acid sequences on the self-assembly behavior and fibril morphology of amyloidogenic peptides derived from islet amyloid polypeptide (IAPP), a 37 amino acid peptide hormone cosecreted with insulin to regulate the glucose levels. 27,28 The susceptibility to IAPP fibrillation due to the presence of a hydrophobic core has been implicated in type II diabetes through the deposition of amyloid plaques. 29 Herein, a series of amyloidogenic segments derived from the 21−29 residues of IAPP (Ac-NNFGAILSS) was designed, and the structures of amyloid fibrils were systematically characterized using transmission electron microscopy (TEM), scanning transmission electron microscopy (STEM), X-ray diffraction (XRD), atomic force microscopy (AFM), Thioflavin T (ThT) fluorescence assay, and circular dichroism (CD). By tailoring the amino acid residues at the C-terminal domain (Ser28, Ser29), which have been shown to determine the fibrillation of the hydrophobic core sequence (IAPP 20−29 residues), 30 we demonstrated the importance of H-bonds and side chain interactions in regulating peptide fibrillation and stabilizing distinct amyloid structures. In particular, we presented the formation of periodically twisted nanoribbons by Ac-NNFGAILSS, where different sizes of nanoribbons corresponding to the numbers of associated fibrils were observed. By tailoring the amino acid substitutions and sequence lengths, polymorphic fibrils were observed in a series of IAPP-derived fragments, where a correlation between molecular interactions and fibril helicity was hypothesized to emphasize the role of conformational constraints. The effects of N-terminal acetylation and sequence elongation on β-sheet strength, peptide hydrogelation, and fibril morphology were also studied. We anticipate our findings can help to understand the sequential determinants toward peptide self-assembly and amyloid polymorphism.

RESULTS AND DISCUSSION
Atomic structures obtained from the microcrystals of IAPP 21− 27 residues (NNFGAIL) revealed a tight main chain interaction between β-sheet layers. 31 This observation inspired us to investigate the morphological variability of peptides derived from the amyloidogenic core. We designed eight peptide sequences by tailoring the terminal residues (i.e., substitution, elongation, N-terminal capping/uncapping) derived from Ac-NNFGAILSS (Figure 1a), including Ac-NNFGAILTT, Ac-NNFGAILAA, Ac-NNFGAILVV, Ac-SSNNFGAIL, Ac-SSNNFGAILSS, Ac-TTNNFGAILTT, and H-SSNNFGAILSS, and studied the structural polymorphs of amyloid fibrillation (Figure 1b). Two major morphologies, the twisted and planar ribbons, were presented by tailoring the amino acid sequences. Polymorphic structures depending on the number of associated fibrils and β-sheet stacking that resulted in thin or thick ribbons were also investigated.

Article
As shown in Figure 2a, amyloid fibrils obtained from Ac-NNFGAILSS exhibited twisted ribbon-like structures with distinct axial periodicities (pitch: 85−100 nm). The high-angle annular dark-field (HAADF)-STEM image recorded a highcontrast image of the amyloid fibrils (Figures 2b and S1), where the strong scattering from the residual salts of the sodium phosphate buffer "stained" the ribbons and confirmed the diameters of uranyl acetate-stained fibrils obtained from TEM imaging. AFM topographic imaging (Figure 2c−e) revealed that the ribbon thickness was 11.4 nm and the height at the crossover was 6.4 nm. This agreed with the diameter measured from TEM imaging, in which the average size of individual fibrils was approximately 5−6 nm ( Figure S2). The length of the Ac-NNFGAILSS peptide was estimated to be approximately 3.4 nm using PyMOL software. Based on these findings, it was concluded that an individual fibril consisted of two adjacent β-sheets, with the hydrophobic inner surfaces packed tightly against each other.
Ribbon-like structures in Ac-NNFGAILSS exhibited different sizes based on the number of associated fibrils (Figure 2f). The smallest peptide ribbon had two fibrils (n = 2) aligned side-byside or in a twisted orientation, with diameters of 10.2 and 11.7 nm, respectively. Most twisted nanoribbons had a diameter of 14−16 nm, which corresponded to the association of three fibrils (n = 3). As shown in Figure S3, the nanoribbons exhibited a periodical spacing of 93.4 nm, which was consistent with the AFM result ( Figure 2e). Such large spacing has frequently been seen when fibrils are loosely coiled, implicating the flexibility of intersheet interactions. 32 Larger ribbons with diameters of 24.4 and 30.2 nm and periodical spacing of 97.6 and 88.2 nm corresponded to the association of n = 4 and n = 5 fibrils, respectively.
The periodical spacing of these larger ribbons remained at 85−100 nm, which was similar to those of the smaller Ac-NNFGAILSS ribbons composed primarily of three fibrils. In fact, bundle formation of β-sheet-forming peptides with finite

ACS Nano
Article fibril width and helicity has been reported to stem from a competition between free energy gain from attraction between ribbons and penalty due to elastic distortion from the further growth of intrinsically twisted ribbons. 33−35 These suggested that the larger ribbons were formed by bundling two or more smaller ribbons during the aging process ( Figure 3a). In addition, right-handed helical structures were found in higherorder twisted structures (Figures 2d and 3b); for example, a tilt angle (γ) of n = 5 ribbons was estimated to be 17.7°( Figure  2d). Formation of superhelical structures resembled the quadruple helix reported by Muraoka et al., in which the helix was composed of two smaller helices, each of which was composed of two individual fibrils. 36 Given that fibril association is a time-dependent and self-propagating process that forms the final energy-favorable structures, 32,37 the growth of Ac-NNFGAILSS fibril was further examined at different time points. As shown in Figure S4, no noticeable twisted ribbons were observed within 6 h, whereas elongated twisted fibrils formed with the increasing aging time. The majority of twisted ribbons were seen to favor an association of n = 3 fibrils over time, suggesting that this was the most thermodynamically stable structure for the Ac-NNFGAILSS assembly.
The kinetics of β-sheet formation by Ac-NNFGAILSS was analyzed by ThT fluorescence assay, where the sigmoidalshaped profile was noted in the process of peptide fibrillation ( Figure S5). The role of β-sheets was highlighted through the addition of high concentration of urea (4 M) to break the Hbonding networks and weaken the β-sheet structures; this not only reduced the number of mature Ac-NNFGAILSS ribbons, but they were also less twisted ( Figure S6). The microcrystal structure of Ac-NNFGAILSS β-sheet structure was obtained by XRD using a dried aligned stalk that was mounted in parallel with the X-ray beam. The 2D diffraction pattern showed a strong meridional reflection at 4.7 Å, which corresponded to the spacing between H-bonded β-strands within a β-sheet ( Figure 4a). The equatorial rings at ∼10 Å (with two concentric rings at 11.1 and 9.0 Å) and 18.2 Å corresponded to the short-and long-range intersheet distances, respectively. Multiple concentric rings (i.e., 9.0 and 11.1 Å) have been seen in thick amyloid fibers consisting of either two or more concentric cylindrical β-sheets or two or more single cylindrical fibers wound around each other. 38 This was in agreement with the varied sizes of Ac-NNFGAILSS nanoribbons observed by TEM imaging (Figure 2f). The outer and inner diffused rings also supported the heterogeneity in the fibrillar structure.
CD and Fourier transform infrared (FTIR) spectra were used to demonstrate the molecular details of interstrand and intersheet interactions of the ribbons. As shown in Figure 4b, the negative and positive maximums at 225 and 206 nm, respectively, suggested a β-sheet structure. The negative peak was slightly red-shifted from the typical β-sheet absorbance at 218 nm, which was due to the aromatic interactions during peptide aggregation. 39−42 The FTIR spectrum obtained from freeze-dried peptide fibrils exhibited strong absorbance at 3285, 2955, and 2875 cm −1 , which corresponded to the N−H stretching, −CH 3 antisymmetric, and symmetric stretching,

ACS Nano
Article respectively. Characteristic bands known as the amide I and II regions are located between 1700 and 1500 cm −1 , which are mainly due to the CO and C−N stretching and N−H bending modes from amide groups of the peptide structure ( Figure S7). The vibrational bands in the range of 1400 to 1300 cm −1 could likely be assigned to the C−N or C−O stretching and N−H bending modes of the peptide structure. 43 In particular, the strongest absorbance at 1628 cm −1 and the absence of 1690 cm −1 in the amide I region (Figure 4c) suggested a parallel β-sheet structure of the peptide fibrils. 44 Other pronounced vibrational bands displayed in the amide I (1667 and 1600 cm −1 ) and amide II (1522 and 1516 cm −1 ) regions confirmed a distinct β-sheet structure. 45 Based on these results, the 3D structure of the Ac-NNFGAILSS fibril was depicted using PyMol software ( Figure 4d) and was consistent with the parallel, face-to-face β-sheets of NNFGAIL (21−27 residues) 31 and NFGAILS (22−28 residues) or FGAILSS (23− 29 residues). 46 Interestingly, untwisted structures (i.e., planar ribbons) were observed for Ac-NNFGAILTT, where the C-terminal serine residues (Ser28, Ser29) of the Ac-NNFGAILSS segment were substituted by threonine. Analyses of TEM (Figure 5a,b) and AFM images (Figures 5c and S8) showed that the individual Ac-NNFGAILTT ribbon had a diameter of 8−10 nm and thickness of 9.4 nm. To confirm the morphology of Ac-NNFGAILTT ribbons, the peptide sample was imaged at a perpendicular orientation to the elongation axis of the ribbons before tilting between ±44°. As shown in Figure 5d,e, the ribbon thickness was reduced at higher angles, which indicated the planar ribbons instead of cylindrical structures. Compared to Ac-NNFGAILSS ribbons ( Figure S4), long and untwisted ribbons of Ac-NNFGAILTT were observed immediately during the cooling process, showing a pronounced tendency for higher numbers of associated fibrils ( Figure S9). As shown in Figure  5f, the XRD pattern showed a similar cross-β structure with the perpendicularly arranged meridional reflection at 4.7 Å and equatorial rings at 17.6, 10.8, and 9.1 Å. It was intriguing to find that the intersheet spacing of Ac-NNFGAILTT ribbons was smaller than the spacing obtained from Ac-NNFGAILSS (18.2 and 11.1 Å) and suggested the tighter intersheet contact. Additional rings in the equatorial region were reflections from higher-order periodicities, indicating that the Ac-NNFGAILTT ribbons possessed a highly crystallized structure. These reflection bands were difficult to assign, which was in part due to overlap with the diffused equatorial reflection from sample heterogeneity. The FTIR spectrum showed a similar interstrand peptide stacking as Ac-NNFGAILSS ( Figure S10), in which the strongest absorbance at 1628 cm −1 and the absence of 1690 cm −1 suggested the parallel β-sheet structure of the ribbons. CD spectrum showed the β-sheet structure of Ac-NNFGAILTT ribbons, which exhibited both negative and positive peaks at 228 and 212 nm, respectively (Figure 5g). The larger red shift ( Figure S11a) in the negative absorbance of the CD peak was observed in Ac-NNFGAILTT compared to Ac-NNFGAILSS (225 nm, Figure 4b), which confirmed a stronger intersheet aromatic interaction facilitated by the phenylalanine

ACS Nano
Article (Phe23) residue. The stronger β-sheet registry of Ac-NNFGAILTT was also confirmed by the ThT assay ( Figure  S11b), which showed an increased fluorescence intensity compared to that of Ac-NNFGAILSS.
The different morphologies of Ac-NNFGAILTT and Ac-NNFGAILSS fibrils were unexpected, given the high structure similarity of serine and threonine. We speculated that the presence of the methyl group in threonine of Ac-NNFGAILTT likely caused a restricted conformational freedom and stronger hydrophobic interactions that balanced the charge effect at the C-terminus. This was supported by XRD (Figures 4a and 5f) and CD data (Figures 4b and 5g), which suggested a reduced intersheet spacing and stronger main chain interactions by threonine substitution. AFM analyses (Figures 2d and S8b)

ACS Nano
Article showed an increased peptide stacking of Ac-NNFGAILTT over Ac-NNFGAILSS in the vertical direction, with the stronger lamination likely explaining the difficulty to introduce or maintain twisted structures. It was therefore proposed that the tighter β-sheet packing favored lateral associations into planar nanoribbons. The effect of serine and threonine side chains resulting in different biological products and conformations has been reported. 47−50 For example, studies by Martı́nez-Saéz and Corzana et al. showed striking differences in the ψ s torsion angle depending on the glycosidic linkages with serine or threonine in the O-linked glycoproteins. 47,50 The former gave an alternate conformation (ψ s = −97 or −180°), whereas the latter exhibited a rigid, eclipsed conformation at ψ s = 63 or 120°, which was explained by the steric repulsion at the methyl group side chain of threonine.
We then further examined the side chain effects of the terminal residues on fibril formation. Peptide sequences with serine substitution by valine and alanine were synthesized separately, and fibril growth was performed under the same experimental conditions. ThT assay (Figure 6a) showed the varied aggregation propensity in the order: Ac-NNFGAILVV > Ac-NNFGAILTT > Ac-NNFGAILSS ≥ Ac-NNFGAILAA, which correlated with the twisted/untwisted morphologies. As proposed above and depicted in Figure 6b, stronger β-sheet registry led to untwisted ribbons of Ac-NNFGAILVV ( Figure  6c) and Ac-NNFGAILTT (Figure 5a), whereas twisted fibrils were formed in Ac-NNFGAILSS ( Figure 2a) and Ac-NNFGAILAA (Figure 6d). In particular, the aggregation propensity as determined by the ThT assay (Figure 6a) could correlate with the residue hydrophobicity, in which valine possessed the highest hydrophobicity while serine and alanine were more hydrophilic. 51−53 On the other hand, increased fibril thicknesses were noted in Ac-NNFGAILTT and Ac-NNFGAILVV, which was due to the enhanced side group

ACS Nano
Article interaction that increased the peptide stacking in the vertical direction. Compared to Ac-NNFGAILTT, the Ac-NNFGAILVV fibrils exhibited an enhanced heterogeneity in fibril diameter. This could be due to the rapid aggregation of Ac-NNFGAILVV upon dissolving in the buffer and likely preexisting oligomer seeds during peptide purification that accelerated the assembly process. The importance of Cterminal residues was also manifested by Ac-NNFGAIL where the two serine residues were removed from Ac-NNFGAILSS, and weak β-sheet formation was observed in this truncated sequence by ThT assay ( Figure S12).
Extended peptide sequences with additional serine or threonine residues at the N-terminal region (i.e., Ac-SSNNFGAILSS and Ac-TTNNFGAILTT) were next synthesized to study the intersheet molecular interactions and the resulting fibril structure. AFM height profiles (Figure 7a−d) showed the thickness of individual fibril was 4.5 and 4.8 nm for Ac-SSNNFGAILSS and Ac-TTNNFGAILTT, respectively. As confirmed by the ThT assay, the extended peptides exhibited stronger aggregation propensity and triggered water gelation due to the strengthened β-sheets (Figure 7d,e). To analyze the hydrogel properties, rheological studies were performed to determine the storage modulus (G′) and loss modulus (G′′) at varying frequencies of applied oscillatory stress, which correspond to the elastic and viscous responses of the peptide hydrogels, respectively. As shown in Figure S13, G′ was higher than G′′ in Ac-TTNNFGAILTT, indicating the elastic hydrogel formation. In line with the findings from Figures S4, S6, and 6a, substitution of serine with threonine enhanced the aggregation propensity of amyloid peptides ( Figure S13c,d), leading to a higher value of G′ for Ac-TTNNFGAILTT than Ac-SSNNFGAILSS ( Figure S13d).
The Ac-TTNNFGAILTT fibrils were untwisted and had a diameter of 9.8 nm, which was likely a lateral association of two fibrils ( Figure S14). Most Ac-SSNNFGAILSS fibrils also favored an untwisted structure (diameter: 8.5 nm), although some twisted fibrils were presented (Figure 8a,b). An uncapped peptide (H-SSNNFGAILSS) was prepared to examine the effect of charge interactions on peptide assembly, in which the opposite charges eliminated the electrostatic repulsions at the termini (Figure 8c). FTIR spectra show the parallel β-sheet structure for H-SSNNFGAILSS and Ac-SSNNFGAILSS (Figure 8d,e). The more defined FTIR peaks at 1540 cm −1 (amide II) and 1632 cm −1 (amide I) with the decreased absorbance at 1667 cm −1 (amide I) suggested the stronger order of the uncapped peptide fibrils. As shown by the TEM images in Figure 8c, the uncapped peptide boasted a marked difference from the capped sequence (Figure 8a) by exhibiting untwisted structures with strong lateral associations (i.e., 4−10 or more associated fibrils). ThT assay showed an enhanced βsheet formation of H-SSNNFGAILSS compared to the capped sequence (Figure 8f), which is due to the electrostatic attraction between −NH 2 and −COOH that promoted peptide selfassembly into large nanostructures. The role of electrostatic interactions was also confirmed by adding salt to the Ac-SSNNFGAILSS solution to shield the terminal charge, in which an enhanced β-sheet formation was observed in the presence of higher concentrations (50 and 150 mM) of sodium chloride (NaCl) ( Figure S15).
Finally, the serine residues of Ac-NNFGAILSS were shifted to the N-terminal region (Ac-SSNNFGAIL) to examine the effect of terminal serine residues. As shown in Figure S16, Ac-SSNNFGAIL exhibited a reduced aggregation propensity compared to Ac-NNFGAILSS, which indicated weaker interactions at Ser19 and Ser20 compared to the C-terminal serine. This observation was consistent with previous findings showing the important role of Ser28 to the fibrillation of IAPP (20−29, SNNFGAILSS) sequence. 30

CONCLUSIONS
In summary, the fibrillation behaviors of eight IAPP-derived amyloidogenic peptides were systematically investigated, where the fibril structures/morphologies could be rationally tailored by varying the peptide sequences ( Table 1). The IAPP 21−29 fragment (Ac-NNFGAILSS) aggregated into twisted ribbonlike structures, with the resultant diameter correlating with the number of associated fibrils. The spontaneous association of individual fibrils into larger twisted ribbons and superhelices was also observed. Increasing the β-sheet strength or intersheet contacts through additional H-bonds and/or hydrophobic interactions led to a reduction of torsional strain in the βsheet and formation of untwisted planar ribbons. The ThT assay indicated a correlation between fibril morphology and βsheet strength, in which untwisted planar structures displayed a stronger propensity to form β-sheets. This hypothesis was supported by amino acid substitution (i.e., serine substitution by threonine, valine, and alanine), sequence elongation (i.e., additional N-terminal serine and threonine), and terminal capping. The role of electrostatic interaction toward peptide assembly and the peptide sequence toward hydrogelation were also addressed. Since polymorphism is a common feature due to the variable amyloid fibril assembly pathways, and the fact that a correlation has been drawn between the fibril assembly/ morphology and cellular toxicity, 54,55 we believe that our work will provide understanding of the amyloid-derived peptide selfassembly and the structural basis of amyloid polymorphism.

MATERIALS AND METHODS
Materials. Fmoc-protected amino acids for peptide synthesis were purchased from AGTC Bioproducts Ltd. Wang resins were used for peptide synthesis. Sodium phosphate monobasic (NaH 2 PO 4 ), sodium phosphate dibasic (Na 2 HPO 4 ), and Thioflavin T were purchased from Sigma-Aldrich (UK). Carbon grid (200 mesh copper) and uranyl acetate solution were purchased from Electron Microscopy Sciences. Milli-Q water (18.2 MΩ.cm) was used for all the experiments.
Solid Phase Peptide Synthesis. Peptides were synthesized by standard Fmoc chemistry using Wang resins. Protected amino acids were added to the growing peptide chain with the activating reagent 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU). Following addition of the N-terminal amino acid, the Fmoc group was removed under 20% piperidine in DMF deprotection conditions. The peptides were cleaved by trifluoroacetic acid/triisopropylsilane/deionized water (95:2.5:2.5, v/v/v). The crude peptide was precipitated by cold ether several times, and reverse-phase high-performance liquid chromatography (HPLC, Shimadzu) was applied to obtain the target peptides. The Phenomenex C18 Gemini NX column was 150 × 21.2 mm and had a 5 μm pore size and 100 Å particle size.
Preparation of Peptide Fibrils. Peptide powders were dissolved in sodium phosphate buffer (20 mM, pH 7.5) at desired concentrations. The solutions were heated to 70°C for 10 min to fully dissolve the powders before being cooled at room temperature.
Transmission Electron Microscopy Imaging of Negative Stained Amyloid Fibrils. For TEM sample preparation, 10 μL of the aged fibril solution was dropped on a carbon film grid for 5 min, and the residual liquid was removed with a piece of filter paper. After that, the TEM grid was stained with 5 μL of 1 wt % uranyl acetate for 5 min. The excess staining agents were removed with filter papers. TEM imaging was performed on JEOL 2100F TEM with an acceleration voltage of 200 kV, and the images were recorded with an Orius camera. The fibril diameters, twist spacing, and twist angles were estimated from at least four measurements using ImageJ, depending on the fibril morphologies.
High-Resolution TEM/STEM Imaging of Unstained Amyloid Fibrils. TEM samples were prepared by dropping 10 μL of the aged fibril solution on an ultrathin carbon holey grid (400 mesh copper, Electron Microscopy Sciences) for 5 min, and the excess liquid was removed with filter paper. Electron microscopy characterizations of the unstained amyloid fibrils were performed on a Philips CM300 TEM at 300 kV and a Cs-corrected TEAM 0.5 TEM/STEM at 50 kV using a HAADF detector, which resulted in Z-contrast images.
X-ray Diffraction. Peptide solutions (25 mM) were prepared in phosphate buffer (20 mM, pH 7.5) and aged for at least 2 days prior to XRD sample preparation. The aged solution (10 μL) was dropped and suspended between two waxed stalks spaced around 2 cm. The droplet was left to dry at room temperature for at least 24 h. The diffraction pattern was collected at the Advanced Light Source Beamline 8.3.1 (λ = 1.12709 Å) with a Pilatus 6 M detector. Exposure times were 1.00 s with a 1.00°rotation at specimen-to detector distances of 350 mm. The 2D diffraction images were radially integrated into a 1D pattern by Nika 2D SAS software package for Igor Pro. 56 Atomic Force Microscopy Imaging. For AFM sample preparation, 5−10 μL of the aged fibril solution was dropped on an oxygen plasma-treated silicon wafer (about 1 cm × 1 cm) for 5 min. The residual liquid was removed. AFM imaging was performed on a 5500 Microscope (Keysight Technologies, previously Agilent) in tapping mode in air. A HQ:NSC15/Al BS tip (μmasch) was applied for the topography imaging (tip radius of 8 nm, resonance frequency of 325 kHz, force constant of 40 N m −1 ). Circular Dichroism. Peptide solutions were prepared in phosphate buffer (20 mM, pH 7.5) and aged for at least 2 days prior to CD analyses. The samples were loaded in a 0.01 cm quartz cell, and CD measurement was performed with a Jasco-715 spectrometer, with the scanning speed of 100 nm/min, resolution of 2 nm, data pitch of 0.1 nm, accumulation time of 2, and response time of 4 s. The data unit was converted from machine units (millidegree, θ) into molar ellipticity [θ] using the equation where P is the path length in centimeters and conc. is the molar concentration of the peptide samples.
Fourier Transform Infrared Spectroscopy. The IR spectra of the fibril powders were obtained on FTIR spectrometry (Shimadzu) in attenuated total reflection mode, with the wavenumber scanned between 4000 and 650 cm −1 , response time of 8 s, and step of 1 cm −1 .
Thioflavin T Assay. To determine the critical aggregation concentration, different concentrations of peptide solutions were incubated with 10 μM of ThT in 384-well plates. The plate was left for at least 12 h at room temperature in the dark before fluorescence measurement. Fluorescence spectra were obtained on a PerkinElmer Enspire plate reader and SpectraMax M5 multimode microplate reader (Molecular Devices Ltd.) for Figure 7, with an excitation wavelength at 440 nm and emission between 455 and 700 nm. For measuring the kinetics of ThT fluorescence, peptide solutions (2 mM) were prepared in the phosphate buffer (20 mM, pH 7.5) containing 10 μM of ThT in 384-well plates prior to fluorescence measurement (ex. 440 nm, em. 485 nm).
Rheology. Peptide solutions (10 mM) were aged in phosphate buffer (20 mM, pH 7.5) for at least 2 days prior to the rheological measurement. Rheological properties were assessed using an Advanced Rheometer AR2000ex with AR Instrument Software (TA Instruments) fitted with a Peltier temperature control system. Oscillatory parallel plate rheological measurements were made at 25°C using a 10 mm diameter steel plate with a 500 μm gap distance in the initial setup. The dynamic moduli of the peptide gels were obtained by measuring the G′ and G″ as a function of controlled angular frequency in the range from 0.1 to 50 rad/s, with a constant strain value of 0.1%.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10