Amyloid self-assembly of hIAPP8-20 via the accumulation of helical oligomers, α-helix to β-sheet transition, and formation of β-barrel intermediates

Abstract

Self-assembly of human islet amyloid polypeptide (hIAPP) into β-sheet rich nanofibrils is associated with the pathogeny of type 2 diabetes. Soluble hIAPP is intrinsically disordered with N-terminal residues 8-17 as α-helix. To understand the contribution of the N-terminal helix to the aggregation process of full-length hIAPP, we investigated the oligomerization and fibrillization dynamics of the hIAPP fragment 8-20 (hIAPP8-20) with combined computational and experimental approaches. hIAPP8-20 was able to form cross-β nanofibrils in silico from isolated helical monomers via the helical oligomers and α-helices to β-sheets transition in the oligomers, as confirmed by transmission electron microscopy, atomic force microscopy, circular dichroism spectroscopy, Fourier transform infrared spectroscopy and reversed-phase high performance liquid chromatography. Our computational results also suggested that the critical nucleus of aggregation corresponded to hexamers, consistent with a recent mass-spectroscopy study of hIAPP8-20 aggregation. hIAPP8-20 oligomers smaller than hexamers were helical and unstable, while the α-to-β transition started from the hexamers. Converted β-sheet rich oligomers first formed β-barrel structures as intermediates before aggregated into cross-β nanofibrils. Our study uncovered a complete picture of hIAPP8-20 peptide oligomerization, aggregation nucleation via conformational conversion, formation of β-barrel intermediates, and assembly of cross-β protofibrils, thereby shedding light to the aggregation of full-length hIAPP, a hallmark of pancreatic beta-cell degeneration.

Graphical Abstract

hIAPP8-20, the helical fragment of human islet amyloid polypeptide, self-assembles into amyloid nanofibrils via the accumulation of helical oligomers, α-helix to β-sheet transition, and formation of β-barrel intermediates

Introduction

The self-assembly of peptides has attracted great attention due to its association with numerous degenerative diseases (e.g. Alzheimer[1], Parkinson[1], and type II diabetes (T2D)[2]) and its potential applications in nanotemplating and bio-nanotechnology[3–5]. Human islet amyloid polypeptide (hIAPP), a 37-residue peptide hormone co-secreted with insulin by pancreatic β-cells for glycemic control[6, 7], is the main component of the amyloid deposits found in the pancreas.[8, 9] Similar to other amyloid deposits[10–12], hIAPP amyloid nanofibers displayed a common cross-β core structure with the β-strands aligned perpendicularly to the fibril axis[7, 13, 14]. While hIAPP fibrils were found cytotoxic to pancreatic β-cells[15], accumulating evidence revealed that smaller, often transient and polymorphic oligomers preceding the formation of fibrils were more cytotoxic [16, 17]. Thus, characterizations of these oligomeric structures and their aggregation dynamics in the early aggregation stage of hIAPP are of significant importance both for understanding the cytotoxicity mechanism and for designing therapeutic strategies. However, the polymorphic and heterogeneous nature of these small amyloid oligomers remains a challenge to experimental characterizations[7].

Isolated hIAPP monomer in solution is intrinsically disordered with N-terminal residues 8–17 being highly helical[18], as also observed in other experimental and computational studies[18–24]. Structural models of hIAPP fibril derived from solid-state NMR and x-ray crystallography[14, 25] have both the amyloidogenic core sequence of residues 22–29[26] and N-terminal residues 8–17, also known as the secondary amyloidogenic core region[23], forming inter-peptide β-sheets in the cross-β core. Multiple experimental studies on the aggregation process of hIAPP reported the accumulation of α-helical intermediates before the final formation of β-sheet rich aggregates[7, 27–31]. Stabilization of helical conformations by negatively-charged membranes[27, 32, 33] or co-solutes[34, 35] accelerated the formation of β-sheet rich amyloid[27, 32, 33], suggesting the importance of helical intermediates in aggregation. Other hypothesis, contending that hIAPP aggregates through β-sheet conformers without involving α-helical intermediates, was also proposed [36, 37]. For instance, Dupuis et al. suggested β-hairpins as precursors for hIAPP aggregation based on their ion-mobility spectrometry–mass spectrometry (IMS-MS) experiments and molecular dynamics (MD) simulations[36, 37]. De Carufel et al. found that preventing helical folding with non-helical hIAPP analogs increased hIAPP cytotoxicity[19, 38], indicating the cytotoxicity of hIAPP might be irrelevant of helical intermediates. In addition, Maj et al. reported that helices in hIAPP could seed oligomer formation but were not necessary for amyloid formation[39]. Hence, the exact role of these helical intermediates in the hIAPP aggregation and the associated amyloid cytotoxicity remains still unclear, and so does the conversion of these helical structures to final β-sheets. Interestingly, the fragment of residues 8–20 (hIAPP8–20) corresponding to the helical region of hIAPP could independently aggregate into amyloid-like fibrils[23, 40]. Combing IMS-MS with atomic force microscopy (AFM), a recent study of hIAPP8–20 aggregation showed that the peptides first aggregated into various oligomers up to decomers[40]. Collision cross-section analysis revealed a morphological conversion as the oligomer size increased larger than a hexamer[40]. Understanding molecular details of the hIAPP8–20 aggregation process including the oligomerization dynamics, structures of various oligomers, and conformational transitions may help understand the contribution of the N-terminal helix to the aggregation process of full-length hIAPP.

In this work, we employed MD simulations with complementary transmission electron microscopy (TEM), AFM, circular dichroism (CD) spectroscopy, Fourier transform infrared (FTIR) spectroscopy and reversed-phase high performance liquid chromatography (RP-HPLC) to investigate the oligomerization and fibrillization dynamics of hIAPP8–20. Atomistic discrete molecular dynamics (DMD) simulations with implicit solvent - a rapid and accurate MD algorithm widely used to study protein dynamics, protein folding and aggregation[41–46] - were performed to study the assembly dynamics up to 20 hIAPP8–20 peptides. Our simulation results revealed that helical hIAPP8–20 monomers first assembled into various helix-rich oligomers from dimers to decomers by burying the hydrophobic surface of the amphiphilic helices. Low-molecular-weight helical oligomers smaller than the hexamers were metastable and underwent frequent association/dissociation dynamics. When the oligomer size increased above the hexamer, the dominant secondary structures became β-sheets. The computed critical nucleus size of the hIAPP8–20 hexamer was consistent with the mass-spectroscopy (MS) experiments[40], where the same hIAPP fragment was used. Most importantly, intermediate states in the form of β-barrels[47, 48] were observed after the α-helices to β-sheets transition in the oligomers. AFM and TEM imaging revealed the formation of single oligomers, and the accompanying FTIR and CD spectra confirmed conformational transitions of hIAPP8–20 from random coils and α-helices to β-strands over 24 h of incubation. Hence, our study uncovered a complete picture of hIAPP8–20 aggregation from isolated helical monomers to cross-β fibrillar aggregates via the accumulation of helical oligomers, conformational conversion from α-helices to β-sheets, and the formation of β-barrel oligomer intermediates.

Results and discussion

Eleven molecular systems with the number of hIAPP8–20 peptides ranging from 1 to 20 were studied in DMD simulations (Methods). In order to obtain enough conformational sampling and avoid potential bias associated with starting configurations, 20 independent DMD trajectories starting with different initial coordinates and velocities were performed for each molecular system and each independent simulation lasted 500 ns. The peptide concentration of each system was kept at ~6.5 mM by adjusting the dimension of the simulation box (Methods). The equilibrium assessment of the simulations was done before data analysis by examining the time evolution of several parameters, including potential energy, number of hydrogen bonds, and secondary structure contents. As illustrated by a typical trajectory of the largest 20-peptide system (Figure S1, randomly selected out of 20 independent simulation trajectories), simulations reached the steady states in the last 200 ns without major changes for the above energetic and structural parameters.

Conformational conversion from helices to β-sheets with increasing system size.

We first analyzed the conformation of an hIPP8–20 monomer, by computing the potential of mean force (PMF, i.e., the effective free energy) as a function of end-to-end distance and number of backbone hydrogen bonds using the last 200 ns trajectories of 20 independent simulations (Figure S1). There was only one free energy basin with the end-to-end distance and the number of hydrogen bonds corresponding to 1.5~2.5 nm and 4~10, respectively. The centroid structures of top 4 most populated clusters, located in the energy basin, represented that 47.5% of all conformations of hIAPP8–20 monomer mainly assumed the helical conformation. This result revealed that the isolated hIAPP8–20 monomers adopted the helical structures, in agreement with prior experimental studies[29, 31].

To structurally investigate hIAPP8–20 oligomers, we computed the average secondary structure contents of random coils, helices, β-sheets in each of the multi-peptide systems during the last 200 ns of all independent simulations (Figure 1a). As the system size increased from two to ten, we observed the decrease of coils and helices, and the increase of sheets. When the number of simulated peptides was larger than ten, the overall secondary contents became saturated with the ratios of random coils, helices, and β-sheets fluctuating around 0.41, 0.26 and 0.21, respectively. We further analyzed for each residue the probabilities in the helical or β-sheet conformation in each system (Figure 1b). Helices were centered around residues 10–15, displaying a high helical propensity over 0.65 when the number of simulated peptides was less than eight. This result is consistent with a recent two-dimensional infrared spectroscopy (2D IR) study, where residues L12A13 mainly adopted the helical structure, but residues L16V17 did not in the monomeric state[39]. As the number of peptides increased to 10 and more, the average helical probability for above residues decreased to around 0.5. On the other hand, β-sheets started to emerge around residues 16–18 as the system size increased. For instance, as the number of peptides was larger than eight, the β-sheet propensities for residues L16, V17, H18 were over 0.4. Hence, as the number of peptides increased, the helices centered around residues 10–15 decreased while the β-sheets started to emerge and expand around residues 16–18.

Figure 1. Averaged conformational analysis of hIAPP8-20 oligomers in DMD simulations.

(a) The average secondary structure contents for random coil, β-sheet and helix in different molecular system with the number of aggregating peptides from 2 to 20. For each system, the propensities of each residue to form (b) helix and (c) β-sheet were computed. The mass-weighted size distributions of (d) oligomers, (e) β-sheet oligomers, and (f) invidual β-sheets were also computed. For each system, the analyses were done for last 200 ns trajectories of all 20 independent simulations.

To characterize oligomerization dynamics, we defined an oligomer as an aggregate of peptides inter-connected by at least one intermolecular heavy atoms contact with a cutoff distance of 0.55 nm, and the size as the number of peptides in the complex[43]. Using the last 200 ns trajectories of all independent simulations for a system, we computed the mass-weighted size distribution of oligomers (Figure 1d), which corresponded to the probability of finding a peptide in an oligomer of a given size. In all simulated systems, the peptides had a significantly high probability to be in the monomeric state. For instance, the monomer population in 2-, 4- and 6-peptide systems was 0.67, 0.43 and 0.38, respectively. When the number of peptides increased further, the monomer probability remained at ~0.30. For small molecular systems up to six peptides, non-monomeric oligomers with the size up to the system size displayed continuous probability distributions. But for larger systems, the size distributions of non-monomeric oligomers started to peak at around 7–9.

We further analyzed the probability of a peptide forming a given sized β-sheet oligomer in each system (Figure 1e). A β-sheet oligomer was defined as an aggregate of β-sheets inter-connected by at least one heavy atom contact, and the corresponding size was the number of peptides adopting β-strand structures in the complex. For small systems up to six peptides, the probability of forming β-sheet oligomers was low (mostly less than 0.06). As the system size increased to 8 and above, the β-sheet oligomers with 5–9 β-strands became predominantly β-sheet rich species with increased probabilities of ~0.07–0.12. We also computed the probability of each peptide forming individual β-sheets of a given size (i.e., the number of β-strands in a single β-sheet layer) (Figure 1f). In most systems, the size of the most populated β-sheet oligomer was equal to the size of the dominant β-sheet, suggesting that these β-sheet oligomers were mainly composed of a single β-sheet. β-sheet oligomers with sizes larger than 10 were also observed in the 18- and 20-peptide systems, but the dominating β-sheet sizes were mainly around 5–8, indicating that two-layer β-sheet oligomers could also be observed in these systems. We also computed the alignments of adjacent β-strands in the β-sheets (Figure S3), which revealed that β-sheets in the early hIAPP8–20 aggregates had both parallel and anti-parallel alignments of β-strands.

Biophysical characterization of the helices to β-sheets transition during the fibrillization of hIAPP8–20.

To experimentally investigate the early aggregation dynamics of hIAPP8–20 and to validate our computational predictions, we performed TEM, AFM, CD, FTIR and RP-HPLC studies. As shown in Figure 2a&b, our imaging revealed formation of single, barrel-like hIAPP8–20 oligomers of circa 7.0±0.9 nm of width by TEM (Figure 2a) and 3±1.0 nm of height by AFM (Figure 2b) over 6 h of incubation. Consist with our simulation results, the FTIR spectra (Figure 3a-c) displayed a gradual transition from random coils (peak 1,640–1,648 cm⁻¹, black trace) to β strands (peak 1,624–1628 cm⁻¹, purple trace) over 48 h of incubation. In addition, the CD data (Figure 3d) revealed a negative peak at 200 nm at 0 h and a negative peak at 218 nm at 48 h, reinforcing the structural transition from random coils to β strands.

Figure 2. Oligomer and fibril morphology characterizations.

(a) Transmission electron microscopy (TEM) and (b) atomic force microscopy (AFM) revealed formation of distinct barrel-like hIAPP8-20 oligomers of circa 3±1.0 nm of height by AFM and 7.0±0.9 nm of width by TEM over 6 h of incubation. (c) Over 48 h of incubation hIAPP8-20 formed amyloid fibrils. hIAPP8-20 concentration: 1 mM.

Figure 3. Characterizations of hIAPP8-20 secondary structures over time.

(a-c) Fourier transform infrared (FTIR) spectra indicate a transition from random coils (peak 1,640-1,648 cm⁻¹) to β strands (peak 1,624-1,628 cm⁻¹) over 48 h of incubation. (d) CD analysis revealed a negative peaks at 200 nm at 0 h and a negative peak at 218 nm at 48 h, reinforcing the structural transition from random coil to β strands. hIAPP8-20 concentration: 1 mM.

With RP-HPLC we noticed a 6.5% depletion in the monomeric content and a slight rise in hydrophobicity for the sample of 6 h (254.4 mAU for the sample of 6 h vs. 272.1 mAU for the sample of 0 h; mAU: milli-absorbance unit). In contrast, for the sample of 24 h, 70% of monomers were depleted and converted into a range of heterogeneous structures with different hydrophobicity. For the sample of 48 h, the monomers completely disappeared and highly hydrophilic amyloid structures were recorded at earlier retention times (i.e., centered on the 7^th min) compared with that for monomeric hIAPP of 0 h.

Therefore, our experimental data confirmed that there existed a conformational conversion of hIAPP8–20 over time, from random coil/helical abundant to β-sheet dominant conformation during the early oligomerization phase.

hIAPP8–20 first aggregated into helical oligomers and then converted into β-barrel oligomers.

In addition to the averaged conformational analysis, we also examined the detailed conformational dynamics along simulation trajectories to investigate how hIAPP8–20 aggregated into β-sheet rich structures from helical monomers (representative trajectories for each molecular system in Figure 4 & Figure S5). We monitored the time evolution of the oligomers of different sizes (first columns in Figure 4 & Figure S5), together with sizes of the largest oligomer, the largest β-sheet oligomer, the largest β-sheet, and the mass-weight average β-sheet size (second columns of Figure 4 & Figure S5). The single-layer β-sheet aggregates in our simulations (Figure 1) could adopt the β-barrel conformation by closing the sheets with maximized backbone hydrogen bonds. Hence, we also plotted the total number of peptides forming β-barrel oligomers (i.e., the β-barrel size) along the simulation trajectories (Figure 4 & Figure S5). In addition, the time evolution of the random coil, helical and β-sheet structure contents was also included (third columns of Figure 4 & Figure S5). In small systems with 2–6 peptides (Figure S5a, b and Figure 4a), helical oligomers underwent frequent association and dissociation dynamics, and the relatively large oligomers (e.g. the largest oligomer whose size equal to the system size) were short-lived. Transient conformational conversions from helices to β-sheets could be observed in these relatively large oligomers. A similar conformational conversion behavior was observed for hIAPP11–25 dimers using all-atom MD simulations with explicit solvent[49]. In these molecular systems, these β-sheet rich oligomers were not stable and could dissociate back to smaller helical oligomers.

Figure 4. Aggregation dynamics of hIAPP8-20 in simulations.

In representive simulation trajectories of 4 (a), 8 (b), 12 (c), 16 (d) and 20 (d) peptides, the time evolution of populated oligomer of different sizes (colored coded according to mass-weighted probabilities),is shown in the first column. The time evolution of the sizes of the largest oligomer size (black), the largest β-sheet oligomer (red), the mass-weighted average β-sheet (blue), the largest β-sheet (purple) and the β-barrels (green) is plotted as the second column. The toal secondary structure contents are also shown as the funciton of time in the third column. Three representive snapshot structures along the simulation trajectories are presented to the right. Side-chains are shown as sticks and colored according to residue type (hydrophobic in white, hydrophilic in green, positively charged in blue, and negatively charged in red).

As the system size increased to 8 and larger, the conformational transition from helices to β-sheets was irreversible. Initially, hIAPP8–20 first aggregated into helical oligomers with sizes up to hexamers (e.g., corresponding to the initial phases where the helical content did not decrease and β-sheet structures were near zero in Figure 4 & Figure S5). Further growth of oligomers coincided with the increase of β-sheet content. In the later stage of these simulations, the largest β-sheet oligomer size was equal to the largest β-sheet layer size, suggesting that the peptides tended to form single-layer β-sheet structures consistent with the averaged conformational analysis above (Figure 1). These single-layer β-sheets could rearrange themselves by forming closed β-barrels (e.g., snapshots in Figure 4 & Figure S5), which underwent opening-and-closing dynamics during the simulations with the size of the largest β-sheet remained the same.

β-barrel oligomers, a novel class of oligomer aggregates first observed in a 11-residue peptide derived from the slow-aggregating αB crystallin[48], were proposed as toxic oligomers of amyloid aggregation due to their well-defined structures and compatibility to the “amyloid-pore” hypothesis of amyloid toxicity[50, 51]. Formation of β-barrel oligomers by Aβ40/42 was also supported by recent experimental studies using hydrogen exchange mass spectrometry[52] and NMR measurement[53]. IMS-MS study combined with computational modeling suggested that that Aβ fragments including Aβ24−34, Aβ25−35 and Aβ26−36 could also form β-barrels oligomers[54]. β-barrels was also observed in prior computational studies of other amyloid fragments - such as Aβ16–22[42, 55] [56], SOD1 (residues 147–153)[57], β2m(83−89)[58, 59] - using either all-atom or coarse-grained models. Systematic DMD simulations of seven 11-residue peptides derived from hIAPP, Aβ and α-synuclein demonstrated the formation of β-barrel oligomers as common aggregation intermediates en route from random coils to cross-β nanofibril[47]. Our results suggest that the helical fragment of hIAPP (i.e., hIAPP8–20) could also form β-barrel oligomer intermediates during the conformational transition from helices to β-sheets in oligomers.

Critical nucleus of hIAPP8–20 aggregation corresponded to hexamers.

A recent IMS-MS study of the early aggregation of the same peptide fragment as in our simulations detected oligomers with sizes ranging from two to ten[40], consistent with our observation of oligomer populations saturated around decomers in various simulations with up to 20 peptides (Figs. 1d, Figure 4, & Fig S5). During the course of our simulations, these oligomers were stabilized by burying most of their hydrophobic residues and exposing hydrophilic residues (e.g., snapshots in Figure 4 and Figure S5). Measurement of averaged collision cross-section (CCS) of these oligomers with mass-spectroscopy[60] suggested a possible conformational transition in hexamers with increasing oligomer sizes[40]. In our computational analysis, we used the radius of gyration squared (Rg²) to quantify CCS of oligomers of different sizes (Figure 5a). The computed Rg² of different oligomers fitted linearly with the experimental CCS values with a Pearson correlation coefficient of ~ 0.993.

Figure 5. Structural analysis of oligomers.

(a) The scatter-plot between experimental collision cross-section[40] and computational radius of gyration square (Rg²) for hIAPP8-20 oligomers with sizes of 1, 2, 3, 4, 5, 6, 8, and 10. Linear regression results into a correlation coeffient, R ~ 0.993. (b) The oligomer Rg² values as the funciton sizes was fitted by the homogenous growth relationship, y=A*n^2/3 (red line), with A ~ 53.31 Ų and the correlation coefficient ~ 0.966. (c) The probability of a peptide to form β-barrel oligomers, and (d) the probability distribution of β-barrel of different sizes were computed for each molecular system with different number of peptides.

We further fitted Rg² as a function of oligomer sizes, n, with the homogenous growth function of A*n^2/3 (A ~ 53.31, Figure 5b) [61]. Here, the fitting coefficient A was close to the Rg² value of an hIAPP8–20 monomer. Interestingly, we found that the Rg² values of oligomers smaller than an hexamer were located above the fitting line, but those of larger oligomers were below the fitting line – i.e., oligomers became more compact when hIAPP8–20 aggregated into hexamers or larger. This behavior is consistent with the observed helix to sheet transition from hexamers, and also in agreement with the IMS-MS derived postulation[40] that hIPP8–20 β-barrel formation started from a hexamer. We computed the probability for the peptide to form β-barrel oligomers in each molecular system (Figure 5c). β-barrels were observed for the simulations with at least 8 peptides. The mass-weighted size distribution of β-barrel oligomers was also computed for each system (Figure 5d). The most populated β-barrel oligomers were mainly composed of 6–9 peptides. Hence, our results and MS experiments[40] suggest that the conformational conversion of hIAPP8–20 from helices to β-sheets started in hexamers – i.e., the critical nucleus of hIAPP8–20 aggregation, and these β-sheet oligomers formed β-barrel oligomers as the aggregation intermediates.

Aggregation free energy landscape analysis.

To further characterize the hIAPP8–20 aggregation dynamics, we computed the potential of mean force (PMF, i.e. the effective aggregation free energy landscape) as a function of the oligomer size (n_oligomer) and the average number of residues in either helical or β-sheet conformation per peptide (n_helix & n_β-sheet) for the largest system in our computational study (20 peptides). To capture the early aggregation processes, we used all 500 ns trajectories from 20 independent simulations. As shown in Figure 6, isolated monomers (n_oligomer = 1) were mostly helical (n_helix ~ 5–8) but could also unfold (n_helix ~ 0, n_β-sheet ~ 0). The low free energy basin with n_oligomer ~ 2–5, n_helix ~ 4–6, and n_β-sheet ~ 0 denoted the initial helical oligomers (e.g., snapshot 1 in Figure 6), which could inter-exchange with each other due to similar free energies within the same basin. With n_oligomer ~ 6–9, there was another deep broad free energy basin with n_helix ~ 0–4 and n_β-sheet ~ 1–7, corresponding to oligomers underwent conformational conversions from helices to β-sheets. The snapshots labeled 2–5 in Figure 6, along with the aggregation dynamics of representative trajectories in Figs. 4 & S5, demonstrated conformational transitions and formation of β-barrels in these oligomers. Especially, the free energy barrier in Figure 6d at n_oligomer = 6 separating small non β-sheet oligomers from larger β-sheet rich oligomers underscores the hexamer as the critical nucleus of hIAPP8–20 aggregation. We also observed a weakly populated large β-rich aggregates with the basin located at n_oligomer ~14, n_β-sheet ~ 5 and n_helix ~1. The corresponding snapshot structure (labeled 6 in Figure 6) indicates that the aggregated structures were comprised of two-layered β-sheets, resembling the canonical cross-β structure of amyloid fibrils. With the addition of more peptides, we expect a continuous growth of the cross-β aggregates with decreased free energies.

Figure 6. Aggregation free energy landscape.

The 2D PMF as a function of the oligomer size (n_oligomer) and the average number of residues per chain adopting either helical (n_helix, panels a & b) or β-sheet conformations (n_β-sheet, panels c & d) were computed for the largest simulation system with 20 peptides. (e) Snapshot structures corresponding to aggregates labelled as 1-6 in the PMF plots are shown to the right.

Based on simulations of 20 peptides, we computed inter-peptide contact frequency maps between either the backbone (Figure S6a) or the sidechain (Figure S6b) atoms of different residues. The backbone contact frequency map (Figure S6a) revealed that the formation of either parallel or anti-parallel β-sheets was mainly driven by interactions between the hydrophobic residues F15, L16 and V17. The sidechain contact frequency map (Figure S6b) suggests that hIAPP8–20 oligomers such as β-barrels and cross-β oligomers were stabilized by interactions among residues L12, A13, N14, F15, L16 and V17. Overall, hydrophobic residues F15, L16 and V17 played a crucial role in hIAPP8–20 oligomerization and fibrillization.

Conclusion

Here we combined rapid all-atom DMD simulations with complementary TEM, AFM, FTIR, CD and RP-HPLC experiments to investigate the aggregation dynamics of hIAPP8–20, the helical fragment of full-length hIAPP. Our systematic DMD simulations demonstrated that hIAPP8–20 first aggregated into helical oligomers, as observed in prior experimental studies of hIAPP aggregation[28, 62]. Due to the amphiphilic properties of the helices, helical oligomers up to hexamers were observed by burying their hydrophobic residues. With similar free energies, these helical oligomers could undergo frequent association and dissociation dynamics. Conformational changes from helices to β-sheets took place within the oligomers. In helical oligomers smaller than hexamers, these transitions were reversible. Irreversible helix-to-sheet transition was observed only in larger oligomers. Experimental characterizations with TEM, AFM, and RP-HPLC demonstrated the formation of oligomers, while CD and FTIR consistently confirmed such helix-to-sheet transitions of hIAPP8–20 occurred prior to the formation of amyloid fibrils. The computationally determined hexamers as the critical nucleus of hIAPP8–20 aggregation also agreed with the recent IMS-MS study of hIAPP8–20 aggregation[40]. Analyses of both conformational dynamics along simulation trajectories and the aggregation free energy landscape revealed that these β-rich oligomers formed β-barrels – closed single-layer β-sheets – as the intermediates toward the formation of multi-layer β-sheet oligomers or the canonical cross-β aggregates resembling nanofibers of amyloid aggregation. Complementary to our previous study of unstructured amyloid peptides that form β-barrel oligomers as aggregation intermediates toward self-assembling into cross-β aggregates[47], the observation of β-barrel formation by helical hIAPP8–20 suggested that β-barrel oligomers might be a common intermediate of amyloid aggregates independent of the secondary structure of precursor proteins. Together, our results uncovered a complete picture of how helical hIAPP8–20 aggregated into nanofibers via the formation of helical oligomers, conformational conversion from helices to β-sheets, and the formation of β-barrel oligomer intermediates.

Although detailed aggregation pathways of hIAPP8–20 are likely different from full-length hIAPP without the presence of N-terminal disulfide bond, the primary amyloidogenic core of residues 22–29 and following sequence in the C-terminal, they share some common features, such as observations of helical intermediates and conformational conversion of these helices to final β-sheet rich fibrils. Due to their high helical propensities, residues 8–17 in the full-length hIAPP could also retain their helices in their early oligomeric aggregates [7, 27–31]. Our results of the helix-to-sheet transition of peptides in the oligomers as the rate-limiting step of aggregation are also consistent with the hypothesis of β-sheet conformers being the amyloid aggregation precursors[36, 37]. Replica exchange MD simulations combined with IMS-MS experiments demonstrated the intrinsic propensity of hIAPP monomers to form β-hairpin conformations, which could take place in the oligomers. Depending on experimental conditions of aggregation such as concentrations and the presence of co-solvents or membranes[39, 63, 64], the appearance of helical intermediates may vary but the formation of β-sheet rich oligomers should be the obligatory step toward amyloid aggregation. Hence, our study of hIAPP8–20 aggregation sheds a new light on the aggregation of full-length hIAPP.

Materials and methods

Molecular systems setup.

In this work, we systematically investigated the conformational ensembles and aggregation dynamics of hIAPP8–20 with the amino acid sequence of ⁸ATCATQRLANFLVHSS²⁰ by using all-atom discrete molecular dynamic (DMD) simulations with implicit solvate on 11 molecular systems with the number of peptides ranging from 1 to 20. For each system, 20 independent DMD trajectories were performed starting with different initial coordinates and velocities and lasted 500 ns. The concentration of hIAPP8–20 in each system was maintained at ~6.5 mM by adjusting the simulation box size. This concentration, much higher than those in the experiments, was used to accelerate the assembly process and decrease computational cost. A previous benchmark aggregation simulations of a peptide fragment from insulin at four different concentrations (3.3, 8.3, 16.6, and 83 mM) demonstrated that final aggregates reached similar structures with steric-zipper-like features, irrespective of the initial peptide concentration[65]. Details of all the simulations are summarized in Table 1.

Table 1.

Details of the molecular systems in our DMD simulations, including the system size (N_peptide), corresponding dimension of the cubic simulation box, number of DMD runs (N_run), duration of each DMD simulations, and accumulative total simulation times.

hIAPP8-20 (8ATCATQRLANFLVHSS20)
Npeptide	1	2	4	6	8	10	12	14	16	18	20
Dimension, nm	6.3	8.0	10.0	11.5	12.7	13.7	14.5	15.3	16.0	16.6	17.2
Nrun	20	20	20	20	20	20	20	20	20	20	20
Time, ns	500	500	500	500	500	500	500	500	500	500	500
Accumulative time, µs	10.0	10.0	10.0	10.0	10.0	10.0	10.0	10.0	10.0	10.0	10.0

Details of DMD simulations.

All simulations were performed at 300K by using the DMD algorithm[66–68], where the continuous potential functions in classic molecular dynamics (MD) were modeled by discrete step-wise functions. With significantly enhanced sampling efficiency, DMD simulation has been widely used by in studying protein folding[44, 69], amyloid aggregation[42, 43, 70, 71], nanoparticle peptides interactions[72]. The Medusa forcefield[44, 73] adopted in DMD is similar to that used in MD simulation and has been fully discussed in our previous studies[42, 43]. Both bonded interactions (i.e., covalent bonds, bond angles, and dihedrals) and non-bonded interactions (i.e., van der Waals, solvation, hydrogen bond, and electrostatic terms) were included. The units of time, length, and energy in our simulation were 50 femtosecond, 1 Å, and 1 kcal/mol, respectively. The temperature was maintained by the Anderson thermostat[74].

Analysis methods.

The secondary structure of hIAPP8–20 oligomer was determined by using the dictionary secondary structure of protein (DSSP) method[75]. Following our prior work[43], two chains were considered to form a β-sheet if (i) at least two consecutive residues in each chain adopted the β-strand conformation and (ii) they formed at least two backbone hydrogen bonds. A hydrogen bond was considered to be formed if the distance between backbone N and O atoms was ≤3.5 Å and the angle of N−H···O ≥120°[76]. The anti-parallel/parallel β-strand ratio was calculated by the number of hydrogen bonds between any two adjacent anti-parallel/parallel β-strands. The size of a β-sheet was the number of β-strand in a β-sheet layer. The mass weighted β-sheet size ${\overline{n}}_{\beta -\mathit{\text{sheet}}-\mathit{\text{size}}}$, was determined by the following equation ${\overline{n}}_{\beta -\mathit{\text{sheet}}-\mathit{\text{size}}}=\left(\sum _{i=1}^{n_{\beta }}{n}_i^2\right)÷\left(\sum _{i=1}^{n_{\beta }}{n}_i\right)$, where n_β denotes the number of β-sheets, and n_i is the size of the ith β-sheet. Two peptides belonged to a same oligomer if they were inter-connected by at least one inter-molecular heavy atom contact (the cutoff of 0.55 nm). The oligomer size was referred to as the number of peptides in an oligomer. A β-sheet oligomer was defined as multiple β-sheets inter-connected by at least one heavy atom contact pair. The total number of β-strands in a β-sheet oligomer was defined as β-sheet oligomer size. If every β-strand was connected by two neighboring β-strands and formed a closed single β-sheet layer, it was then defined as a β-barrel oligomer. We used the network-based algorithm to identify β-barrel oligomers along the simulation trajectories[42]. The potential of mean force (PMF, i.e., free energy surfaces) was constructed using the relation −RT ln P(x, y), where P(x, y) was the probability distribution of two selected reaction coordinates, x and y, computed from simulations.

Atomic force microscopy.

Beta barrel and oligomer formation of hIAPP8–20 were observed using single-molecule atomic force microscopy (AFM). For this hIAPP was dissolved in Milli-Q water at 400 µM concentration, incubated for 6 h at room temperature (20 °C). Then, a droplet of 20 μL of hIAPP was deposited and incubated for 2 min on freshly cleaved mica, rinsed with Milli-Q water, and dried with air. Imaging was performed using a Cypher AFM (Asylum Research) operating in the AC Mode.

Transmission electron microscopy.

Beta barrel and oligomer formation of hIAPP8–20 were further confirmed using transmission electron microscopy (TEM). For this hIAPP was dissolved in Milli-Q water at 400 µM concentration, incubated up to 48 h at room temperature (20 °C). At specific time points of incubation (i.e., 6 h and 48 h), 5 µL of peptide-containing solution was pipetted onto a glow discharged (15 s) copper grid (400 mesh; ProSciTech), followed by 1 min of adsorption. Excess sample was then drawn off using filter paper and the grid was washed by Milli-Q water with the excess drawn off. The grid was stained with a drop of 1% uranyl acetate for 30 s, then the excess stain was drawn off and the grid was air dried. Imaging was performed by a Tecnai G2 F20 transmission electron microscope (FEI, Eindhoven, The Netherlands) operated at a voltage of 200 kV. Images were recorded using a Gatan UltraScan 1000 (2k×2k) CCD camera (Gatan, California, USA) and Gatan Microscopy Suite control software.

Fourier transform infrared spectroscopy.

The FTIR spectra of hIAPP were obtained using a Shimadzu IRTracer-100 spectrophotometer. For this 5 µL of 400 µM hIAPP8–20 (from 0 h to 48 h of incubation time) were placed on a sample holder, air dried and their spectra were acquired between 600 and 4,000 cm⁻¹ at 20 ºC with resolution of 4 cm^-1. A blank spectrum was acquired and subtracted from the sample spectra. Peak fitting and secondary structure analysis were performed using PeakFit v4.12 (SeaSolve Software Inc.).

Circular dichroism (CD).

CD spectra of hIAPP8–20 (1 mM; incubation time: 0 h and 48 h) were obtained for the wavelength range of 190–260 nm with a 1 nm step size at room temperature. The CD spectra were taken using a Chirascan Plus qCD instrument (Applied Photophysics). The measurement was performed in triplicate and average spectra of 3 measurements were analyzed.

Reversed-phase high performance liquid chromatography (RP-HPLC).

hIAPP8–20 (1 mM, incubated for 0, 6, 24 and 48 h) were dissolved in 10% acetic acid and ran on a 4.6-mm (internal diameter) × 50-mm (length) monolithic reversed-phase C18 high-performance liquid chromatography (HPLC) column (Chromolith SpeedROD; Merck Millipore, Darmstadt, Germany) using an ÄKTAmicro HPLC system (GE Healthcare, Little Chalfont, UK). The mobile phase consisted of buffer A (0.1% trifluoroacetic acid; Thermo Fisher Scientific) and buffer B (80% acetonitrile, 0.1% trifluoroacetic acid; Thermo Fisher Scientific). The peptide mixtures were loaded onto the column and separated using the following chromatographic conditions: 2–15% buffer B over 0.25 min (2 mL/min), 15–30% buffer B over 4 min (2 mL/min), 30–40% buffer B over 8 min (2 mL/min), 40–45% buffer B over 10 min (2 mL/min), 45–99% buffer B over 2 minutes (1 mL/min), 99–100% over 2 min (1mL/min), and re-equilibrated 6 min in 2% buffer B at 2 mL/min.