Single-molecular hetero-amyloidosis of human islet amyloid polypeptide

Abstract

Human amyloids and plaques uncovered post mortem are highly heterogeneous in structure and composition, yet literature concerning the hetero-aggregation of amyloid proteins is extremely scarce. This knowledge deficiency is further exacerbated by the fact that peptide delivery is a major therapeutic strategy for targeting their full-length counterparts associated with the pathologies of a range of human diseases, including dementia and type 2 diabetes (T2D). Accordingly, here we examined the co-aggregation of full-length human islet amyloid polypeptide (IAPP), a peptide associated with type 2 diabetes, with its primary and secondary amyloidogenic fragments 19–29 S20G and 8–20. Single-molecular aggregation dynamics was obtained by high-speed atomic force microscopy, augmented by transmission electron microscopy, X-ray diffraction and super-resolution stimulated emission depletion microscopy. The co-aggregation significantly prolonged the pause phase of fibril elongation, increasing its dwell time by threefold. Surprisingly, unidirectional elongation of mature fibrils, instead of protofilaments, was observed for the co-aggregation, indicating a new form of tertiary protein aggregation unknown to existing theoretical models. Further in vivo zebrafish embryonic assay indicated improved survival and hatching, as well as decreased frequency and severity of developmental abnormalities for embryos treated with the hetero-aggregates of IAPP with 19–29 S20G, but not with 8–20, compared to the control, indicating the therapeutic potential of 19–29 S20G against T2D.

Graphical Abstract

Amyloid fibrils are highly organized filamentous aggregates of peptides and proteins.¹ First observed clinically in Alzheimer’s disease (AD), Parkinson’s disease (PD) and type 2 diabetes (T2D) patients, amyloid fibrils share a common characteristic in possessing dense stacks of β-strands stabilized by hydrogen bonds along the fibril axis.² Each amyloid fibril typically consists of two or more protofilaments of β-sheets stretching over the length scale of micrometers, self-assembled by hydrophobic interactions. Amyloid fibrils are polymorphic mesoscopically, assuming versatile forms of twisted ribbons, helical ribbons, nanotubes or flat tapes up to tens of nanometers in diameter depending on the sequence of the amyloid protein, ionic strength, and aggregation time.³

The aggregation of amyloid proteins can be directly observed or inferred from transmission electron microscopy (TEM), X-ray diffraction, circular dichroism spectroscopy, Fourier transform infrared spectroscopy, atomic force microscopy (AFM),^4–12 as well as nuclear magnetic resonance spectroscopy. The kinetics of amyloid protein aggregation, when probed by a thioflavin T (ThT) or Congo red fluorescence assay, follows a sigmoidal trajectory to indicate the three stages of (primary and secondary) nucleation,^13–14 elongation and saturation, where disordered monomers aggregate to render α-helical and then β-sheet-rich oligomers, protofibrils, amyloid fibrils and plaques. While amyloid fibrils and plaques may take decades to form in vivo due to modulation by chaperone proteins, metal ions, pH and other structural and physiological factors, amyloid proteins can rapidly transform in vitro from functional monomers to fibrils within days (e.g., amyloid beta/Aβ and alpha synuclein/αS, whose aggregation is a hallmark of AD and PD, respectively),^15–17 or hours (e.g., human islet amyloid polypeptide/IAPP, whose aggregation is associated with T2D).^18–19

Amyloidosis is a hallmark of a range of debilitating neurological disorders and metabolic diseases. Accordingly, much research over the past two decades has been devoted to understanding the aggregation of amyloid proteins and their elicited toxicity. Towards therapeutic implementations, the design and delivery of peptides,²⁰ metal ions,^21–23 monoclonal antibodies,^24–26 small molecules^{4, 27–30} and, more recently, various nano- and polymeric materials^{18, 31–44} to target amyloid proteins has become a significant area of research. However, these efforts are overshadowed by a lack of cure for amyloid diseases, highlighting a crucial need for additional research and development of new strategies against amyloidosis.

Human IAPP is one of the most aggregation-prone peptides known. It has been established that the residues 20–29 of human IAPP are highly amyloidogenic and cytotoxic.^45–47 Serine-to-glycine substitution at position 20 (S20G), a familial mutation implicated in the onset of T2D mellitus in the Japanese population, results in increased amyloidogenicity and cytotoxicity of IAPP.^48–49 In contrast, a single-point mutation at residue 26 from isoleucine to proline converted IAPP into a potent fibrillization inhibitor.⁵⁰ It has also been found that fibrils formed from the 19–29 S20G fragment are structurally similar to that of full-length human IAPP and hence may serve as a model for studying the toxicity of the latter.¹⁹ In addition to the residues 19–29, a secondary amyloidogenic domain of human IAPP was identified between residues 8 to 20.^51–52

While human amyloid fibrils and plaques are highly heterogeneous in structure and composition,^{1, 28, 53} data on hetero-amyloidosis remains extremely scarce. Much of the research in the field of amyloidosis has been conducted for single amyloid proteins that have limited applicability to in vivo conditions. The need for understanding hetero-amyloidosis is further warranted by peptide-based therapeutics, a major strategy for targeting toxic oligomeric or fibrillar amyloid proteins through a sequence-dependent specificity between the delivered peptide and the host. Accordingly, here we examined the co-aggregation of full-length human IAPP with its fragments IAPP 19–29 S20G and 8–20. In addition to the use of TEM and a ThT kinetic assay, high-speed atomic force microscopy (HS-AFM) at a scanning rate of 0.5 frame/sec was employed to record IAPP self-assembly and co-aggregation with its two major amyloidogenic fragments. This technique enabled the aggregation dynamics of IAPP to be revealed in real-time and at single-molecule resolution, providing information regarding the nature of IAPP hetero-amyloidosis that is unavailable from ensemble methods such as the ThT kinetic assay. Our data indicated that hetero-aggregation retarded fibril elongation but increased aggregation rate of IAPP, and altered IAPP fibril flexibility. Surprisingly, co-elongation of single and bundled IAPP fibrils, but not protofilaments, was observed for IAPP co-aggregating with 19–29 S20G or 8–20, a phenomenon unaccounted for by existing theoretical models.^54–57 Notably, an in vivo zebrafish embryonic assay revealed significant amelioration of IAPP toxicity due to hetero-aggregation of the peptide with 19–29 S20G, but not with 8–20. Together, this study provided first insights into hetero-aggregation down to single fibrillar and sub-second levels, as well as its biological implications which are crucial for the development of peptide-based therapeutics against T2D.

The effects of the amyloidogenic fragments on the aggregation kinetics, mesoscopic and atomic structures of full-length IAPP were examined using a ThT kinetic assay, TEM, stimulated emission depletion (STED) microscopy, HS-AFM, single-crystal X-ray diffraction (XRD) and discrete molecular dynamics (DMD) computer simulations (Scheme 1). Specifically, IAPP fibrillization in the presence and absence of the 11-residue 19–29 S20G (Ser20 replaced by Gly) and the 13-residue 8–20 IAPP fragments (Fig. 1A) was characterized to derive new insights into IAPP hetero-amyloidosis.

Scheme 1.

Techniques employed for characterizing the hetero-amyloidosis of IAPP with its amyloidogenic fragments IAPP8–20 and 19–29 S20G and their in vivo application with a zebrafish model. ThT: thioflavin T kinetic assay. TEM: transmission electron microscopy. Confocal: confocal fluorescence microscopy. STED: Stimulated emission depletion microscopy. XRD: X-ray diffraction spectroscopy. DMD: discrete molecular dynamics simulations. HS-AFM: high-speed atomic force microscopy.

Figure 1. Effects of fragments on IAPP fibril aggregation.

Panel D was derived from the ThT assay, while panels E and F were calculated from computer simulations. (A) Amino acid sequences of full-length human IAPP (37 residues), 19–29 S20G and 8–20 fragments. Red residue in the 19–29 S20G fragment sequence shows the early onset S20G familial mutation. (B) ThT assay revealed significantly promoted aggregation rate of IAPP in presence of 19–29 S20G. (C) 8–20 fragments promoted IAPP aggregation comparably to full-length IAPP. (D) Lag time derived from analysis of ThT fluorescence curves. * denotes statistically significant differences between sample mean to a control mean (ANOVA; * P ≤ 0.05; *** P ≤ 0.005). (E) A coarse-grained peptide model with different barriers between π-state (representing random-coil or helix) and β-state (representing β-sheet) was used in the simulation, where the two states were shown in the inset with stick representation. (F) The number of β-sheets was calculated for each simulation system and plotted as a function of simulation time (IAPP + IAPP corresponds to 2× IAPP concentration).

ThT kinetics assay on IAPP-fragment co-aggregation.

The ThT assay indicated that 19–29 S20G significantly accelerated IAPP aggregation (Fig. 1B). Specifically, even at a low fragment/IAPP molar ratio (1:5) the nucleation (lag) time of 25 μM IAPP was noticeably reduced from 6.1±0.4 h to 2.6±0.9 h (Fig. 1D). Further increases of the fragment/IAPP ratio to 1:1 and 2:1 decreased the nucleation time to 2.4±0.2 h and 1.9±0.8 h, respectively.

In addition, the ThT fluorescence intensity in the saturation phase of IAPP co-fibrillated with 19–29 S20G was up to 50% lower than that of 25 μM IAPP (Fig. 1B). A similar observation was reported by Westermark et al. for full-length IAPP S20G vs. full-length wide-type IAPP.⁵⁸ Moreover, in contrast to full-length IAPP, no concentration-dependent increase in total ThT fluorescence was observed for hetero-aggregation. TEM analysis revealed that amyloid fibrils that formed in the presence of 19–29 S20G had markedly different morphology from the IAPP control (Fig. 2B vs. 2A). Specifically, the fibrils in the presence of the fragment appeared to be highly twisted soft helical ribbons⁵⁸ (Fig. 2B) with altering periodicities of 100–370 nm (Table S1, Supporting Information). In contrast, pure IAPP displayed rich polymorphism of mature amyloids (Fig. 2A). Moreover, the twisted fibrils of pure IAPP were found to have a uniform periodicity of 110 ± 17 nm, and a 4× longer persistence length than amyloids formed in the presence of 19–29 S20G (Table S1). This greatly elevated flexibility of IAPP fibrils in the presence of 19–29 S20G may imply pathogenic consequences in the context of T2D, an aspect to be examined in future in vivo studies.

Figure 2. Effects of fragments on IAPP fibril morphology.

Panels A-L were acquired from TEM, confocal and STED imaging as well as X-ray diffraction, panels M-O were obtained from computer simulations, and panels P-S are cartoon representations of IAPP-fragment hetero-amyloids. (A) IAPP fibril and IAPP co-aggregated with (B) 19–29 S20G and (C) 8–20 at equimolar concentration of 25 μM over 24 h at room temperature. Red dotted line charts represent periodicity analysis of twisted fibril using grayscale histogram plotting along the fibril axis (white arrows indicate the analyzed fibrils). (D-F) Confocal fluorescence microscopy of a ThT-labelled IAPP fibril, a ThT-labelled IAPP + 19–29 S20G co-aggregated fibril, and a ThT labelled IAPP + 8–20 co-aggregated fibril, respectively. (G-I) Stimulated emission depletion (STED) super-resolution microscopy of IAPP fibrils with and w/o the fragments confirmed co-aggregation of (G) non-labelled IAPP with Alexa 647-labelled (H) 19–29 S20G and (I) 8–20 fragments into single hetero-peptide fibrils. X-ray diffraction pattern of (J) IAPP, (K) IAPP + 19–29 S20G and (L) IAPP + 8–20 amyloids. (M) The double U-shape model of an IAPP fibril representing two IAPP molecules mated into a steric zipper around the fibril axis. 8–20 and 19–29 amino-acid sections are highlighted in purple and orange, respectively. Possible binding sites of (N) 19–29 S20G (orange) and (O) 8–20 (purple) fragments to full-length fibril (grey) during fibrillization based on the complementarity of amino-acid sequences. Schematic peptide arrangements in homogeneous (P) IAPP fibril and heterogeneous (R) IAPP + 19–29 S20G and (S) IAPP + 8–20 fibrils, where the grey rectangles represent full-length IAPP, orange 19–29 S20G fragments, and purple 8–20 fragments.

The 8–20 fragment accelerated IAPP aggregation near the efficiency of its cognate full-length peptide (Figs. 1C&D). Similarly to 19–29 S20G, co-fibrillization of IAPP with 8–20 had no clear concentration-dependent effect on the ThT fluorescence at saturation. This can be attributed to the fact that no increase in ThT fluorescence was observed for 19–29 S20G and 8–20 aggregated at 1 mM while fibrils were seen with TEM (data not shown). Similarly, the fibrillization of fragment 8–20 in the literature was mainly characterized by AFM, X-ray diffraction, or TEM.^{52, 59–60} Hence the ThT assay, while informative for probing peptide aggregation kinetics, may not accurately reflect the total β-sheet content in IAPP hetero-aggregates. Interestingly, in contrast to the 19–29 S20G, twisted IAPP amyloid fibrils formed in the presence of the 8–20 fragments were 70% more rigid than the control, while non-twisted fibrils were 18% more rigid compared to pure IAPP aggregates (values of persistence length see Table S1). However, the overall polymorphism of IAPP fibrillated in the presence of the 8–20 fragments (Fig. 2C) was comparable to that of pure IAPP (Fig. 2A).

Aβ16–22, a short Aβ fragment between residues 16–22 that was shown to inhibit fibril formation of full-length Aβ1–42 and thus proposed to be a therapeutic agent for the treatment of AD,^61–63 had little effect on IAPP aggregation (Fig. S1), in contrast to 19–29 S20G and 8–20. This indicates that 19–29 S20G and 8–20 were more selective over Aβ16–22 towards binding to their complementary regions of full-length IAPP.

Structural analyses of IAPP-fragment co-aggregation.

STED microscopy confirmed that both 19–29 S20G and 8–20 co-aggregated with full-length IAPP to form hetero-peptide amyloid fibrils (Figs. 2D–I). The presence of amyloid fibrils was confirmed using β-sheet sensitive ThT dye (Figs. 2D–F), and super-resolution STED microscopy (spatial resolution: 20 nm; an order or magnitude higher than confocal fluorescence microscopy) verified the presence of Alexa 647-labelled 19–29 S20G (Fig. 2H) and 8–20 (Fig. 2I) segments inside the fibrils. XRD revealed the signature pattern of amyloid fibrils, at 4.7 Å for inter-strand spacing and 8–10 Å for inter-sheet distance, which was not affected by hetero-aggregation with the fragments (Figs. 2J–L). The greater flexibility of IAPP + 19–29 S20G and increased rigidity of IAPP + 8–20 compared to IAPP control, based on their values of persistence length (Table S1), can be explained by a proposed model of fragment binding to full-length IAPP according to the complementarity of amino-acid sequences (Figs. 2M–O). As the 19–29 S20G residues were buried in the interior of a double U-shaped IAPP fibril (highlighted in orange color in Figs. 2M&2N), the segments bound the corresponding sections of cognate full-length IAPP to alter the fibrillar morphology of the resulting hetero-aggregates (Fig. 2R vs. 2P). In contrast, the outer binding sites of 8–20 segment (Figs. 2O&2S) did not dramatically change the fibril contour.

Multiscale discrete molecular dynamics simulations of IAPP-fragment co-aggregation.

We performed both atomistic and coarse-grained (CG) DMD simulations to gain a molecular insight into the effects of different amyloidogenic peptide fragments on IAPP aggregation and fibrillization. Using all-atom simulations, IAPP dimerization was studied in the absence or presence of equimolar fragments. In the mixture, fragments 19–29 S20G remained predominately as coils while fragments 8–20 were helical (Fig. S2A inset and Fig. S2B). Compared with the IAPP dimer control, the presence of 19–29 S20G or 8–20 fragment increased the β-sheet content of IAPP in the primary amyloidogenic region (i.e., residues 22–29) (Fig. S2A). The presence of 19–29 S20G also increased the β-sheet content and reduced the helical content of IAPP in the N-terminus (especially, residues 8–16), but the helical 8–20 increased the helical propensity of IAPP in the same region. The all-atom simulations indicated that 8–20 encountered a higher energy barrier associated with helix unfolding than 19–29 S20G (Fig. S2) towards aggregation.

The conformation of IAPP monomers and dimers were well studied in prior computational simulations.^64–69 Although secondary structure contents were force filed dependent, IAPP mainly adopted unstructured conformations with some helical structure in the N-terminal and transient β-sheet structures in amyloidogenic core region (IAPP 8–20 and IAPP 19–29),^64–69 our all-atom DMD simulations are consistent with those previous simulation studies. For example, residues around A8–F15 displayed greater propensities to form helices than the rest region of IAPP and weakly populated β-sheet structures were mainly formed by residues around L16-S28, which was consistent with previous IAPP dimer simulation studies.^67–69

To capture the hetero-fibrillization process, we performed CG DMD simulations with peptide models possessing low, intermediate and high energy barriers between aggregate-incompetent (π) and competent (β) states to represent IAPP, 19–29 S20G and 8–20, respectively (Fig. 1E, Methods in the Supporting Information). Starting from isolated peptides in the π-state the time evolution of the total number of peptides forming β-sheets was computed for the different molecular systems to mimic the ThT assay (Fig. 1F). Indeed, IAPP + IAPP (refers to as double IAPP concentration) and IAPP + 19–29 S20G displayed similar lag times and elongation rates, while IAPP + 8–20 rendered a longer lag time than IAPP + IAPP. This is consistent with the ThT assay showing that 19–29 S20G accelerated IAPP aggregation more efficiently than 8–20 (Figs. 1B–D).

High-speed atomic force microscopy (HS-AFM) and in silico evidence of retarded and unidirectional IAPP co-elongation.

It is important to note that ensemble aggregation kinetics represents a convolution of protein nucleation, elongation and dissociation⁷⁰ and does not accurately discern the presence of amyloid fibrils until the latter reach a threshold in quantity.⁷¹ Accordingly, we further examined the effects of the fragments on IAPP fibril formation using a single-fibril elongation assay. Namely, HS-AFM was utilized to monitor the effects of the two types of fragments on single IAPP fibril elongation (Figs. 3&4; videos S1–S3, for the elongation of IAPP, IAPP + 19–29 S20G, and IAPP + 8–20, respectively). Similar to the observation by TEM, HS-AFM analysis (Fig. 3A) confirmed that pure IAPP fibrils possessed different morphologies and various height profiles between 9 and 20 nm (Figs. 3D,E), suggesting predominant formation of mature higher-order fibrils despite the presence of the mica substrate. Moreover, we observed a convergence of fibril morphologies resulting from the co-aggregation of IAPP with the 19–29 S20G fragment (Fig. 3B). Fibril heights of 9 ± 1.7 nm (an accurate representation of fibril width due to the finite tip size, which renders the lateral sizes in AFM images to appear larger due to tip “broadening”) were recorded by HS-AFM (Figs. 3D,F), corresponding to fibrils assembled from 2 to 3 protofilaments (Figs. 3K–M). In addition, enhanced surface roughness resulted from the formation of small aggregates (Figs. 3B,F). A similar phenomenon was observed for IAPP co-aggregation with the 8–20 fragment (Figs. 3C,G); however, the overall fibril polymorphism was comparable to the control, as indicated by the various height profiles of the fibrils (Fig. 3D,G). Analysis of fibril kymograph projections (Figs. 3H–J and Figs. S3,S4) revealed that the elongation rates of fibrils co-aggregated with the fragments were slower than the self-assembly of full-length IAPP (Fig. 4A). The elongation rate constant (E) of 1 μM IAPP was 0.52 ± 0.19 nm/s, whereas it was 0.12 ± 0.07 nm/s and 0.10 ± 0.03 nm/s in the case of IAPP (0.5 μM) co-aggregated with the 19–29 S20G and 8–20 fragments (0.5 μM), respectively. Surprisingly, co-elongation of multiple mature IAPP fibrils was observed (Figs. 3K–M; video S4), indicating a new form of tertiary protein aggregation driven by H-bonding between the cooperating fibrils. The bending of the bundled fibrils may result from accumulative mismatches between the elongation rates of the fibril constituents over time, which is consistent with the TEM observations (Fig. 2B).

Figure 3. Effects of fragments on single IAPP fibril elongation revealed by high-speed atomic force microscopy (HS-AFM).

HS-AFM images during IAPP fibrillization (A) without and with (B) 19–29 S20G and (C) 8–20 fragments indicated increased formation of small aggregates in heterogeneous samples. (D) Fibril height histogram (n = 35, 68 and 42 for IAPP, IAPP + 19–29 S20G and IAPP + 8–20, respectively). White dotted lines in panels A-C represent high line profiles presented in panels E, F and G. Representative kymographs of (H) pure IAPP, (I) IAPP + 19–29 S20G and (J) IAPP + 8–20 fibril elongation. (K) Time-lapse AFM images (also see video S4) of simultaneous co-elongation of (L) bundled protofibrils (IAPP + 19–29 S20G) and (M) relevant height profiles.

Figure 4. A quantitative analysis of single fibril elongation assay.

Panels A, C, E, F, G and H were derived from HS-AFM experiments, while panels B and D were obtained from computer simulations. (A) Elongation rate constant. Apparent elongation rates obtained from the final fibril length measurements at the end of the elongation assay (Fig. S4A). (B) The number of peptides in one fibril as a function of simulation time, where the initial simulation state and final state are shown in the insets. The fibril growth exhibited two states, namely, elongation state and dwell state, marked by the solid and dashed arrows, separately. (C) The rate of pause-free elongation indicates underestimation of the fibril elongation rate which arise from growth intermittency (n = 19, 14 and 28 for IAPP, IAPP + 19–29 S20G and IAPP + 8–20, respectively) (see Figs. S3,4). (D) The apparent elongation rates were obtained from the linear fitting of multiple fibril growths in panel B. Typical snapshots of fibril structures are shown in the insets. (E-H) Statistical analysis of stepwise IAPP fibril growth or IAPP co-aggregated with the fragments: (E) dwell (pause) time (n = 16, 12 and 22 for IAPP, IAPP + 19–29 S20G and IAPP + 8–20, respectively), (F) step time (n = 19, 14 and 28 for IAPP, IAPP + 19–29 S20G and IAPP + 8–20, respectively), and (G) step size (n = 19, 14 and 28 for IAPP, IAPP + 19–29 S20G and IAPP + 8–20, respectively). (H) Correlation between step size and step time. Error bars denote the mean ± SD. * denotes statistically significant differences between sample mean to a control mean (ANOVA; * P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.005; **** P ≤ 0.001).

The measured elongation rate for full-length IAPP at a concentration of 1 μM was comparable to that of IAPP1–37 (16 and 40 nm/min,⁷² concentration: 2 μM) and Aβ1–42 (0.4–2.5 nm/s, concentration: 2.5 μM in 10 mM phosphate) obtained with HS-AFM,⁷³ and slower than the elongation of Aβ25–35 (5–100 nm/s, concentration: 10–50 μM) determined by scanning-force kymography.⁷⁴ Consistent with the observation by Huang et al.,⁷² we recorded predominantly uni-directional elongation of IAPP fibrils in both self and hetero-aggregation (Fig. S5), also in agreement with other studies reporting fast- and slow-growing ends of amyloid fibrils.^73–76

Similar to previous HS-AFM^{74, 77} and total-internal-reflection fluorescence microscopy⁷⁶ studies of αS and Aβ, we noticed the intermittent nature of IAPP elongation dynamics (Figs. 3H–J and Fig. S4). Intermittent elongation of amyloid fibrils, or pausing, has been proposed in the literature as consequential to the addition of a kinetically trapped state at the fibril end that is off pathway from the normal monomer-to-fibril addition reaction.⁷⁶

We also monitored the growth of single fibrils in CG simulations by tracking the time evolution of the number of peptides in a given fibril (Fig. 4B). The fibril growth exhibited a similar intermittent increase in silico (referring as the elongation state and pause state noted by line arrow and dash arrow in Fig. 4B, respectively) as fibril traces shown in Fig. S4A. Examination of the simulation trajectory (Fig. S6) indicated that the heterogeneous distribution of soluble peptides resulted in these two states during fibrillization. During the elongation phase, there were a locally high concentration of monomeric peptides around the tracked fibril (highlighted in Fig. S6 on the left) and a locally low concentration of peptides near the fibril in the pause phase (highlighted in Fig. S6 on the right). By analyzing the growth of multiple fibrils from independent simulations, the apparent elongation rates were obtained for IAPP control and IAPP mixed with the fragments (Fig. 4D), with typical fibril structures shown as insets. Consistently with the experimental elongation assay (Fig. 4A), the elongation rate for IAPP self-assembly was higher than IAPP co-aggregated with fragments from the simulations. In addition, co-aggregation of IAPP with the fragments significantly prolonged the pause phase of fibril elongation, increasing its dwell time from 53 ± 32 s to 188 ± 127 s and 156 ± 136 s in the presence of the 19–29 S20G and 8–20 fragments, respectively (Fig. 4E). This phenomenon may arise from the incorrect docking of either the fragments alone or monomer-fragment complexes to the end of an elongating protofibril, which was then followed by dissociation of the ‘faulty’ fragment/complex or a conversation of the latter into the elongation-competent conformation. Notably, pure IAPP elongation intermittence arose from the same scenarios, and escape from this metastable state would require a transition across an energy barrier either from the trapped state to the correct conformation state, or to the monomeric state. Our observations suggest that the energy barriers of this trapped state were higher in the case of hereto-aggregation. This is consistent with the study by Laurence et al. of Aβ1–42 elongation at different temperatures, where increased temperature (higher free energy) decreased the fraction of time that a fibril end spent in the pause state.⁷⁶ Although we did not observe statistically significant differences in the step time (Fig. 4F), we did find that the step size was shortened in the case of IAPP co-aggregation with 8–20 fragments (Fig. 4G). Interestingly, there was a strong correlation between the step size and step time (R² = 0.85) for the fibrillization of pure full-length IAPP (Fig. 4H). However, no such dependence was found in the presence of the fragments (R² = 0.32 and 0.28 for IAPP + 19–29 S20G and IAPP + 8–20, respectively). Based on this observation and the fibril elongation trajectories with pause sections artificially removed (Fig. S4B), we conclude that the rate of full-length IAPP elongation was constant but their co-aggregation with the fragments varied over time, likely due to the entropic gains of the latter heterogeneous systems.

Pause-free projection of fibrillization trajectories avoided underestimation of fibril elongation rate due to intermittencies (Fig. S4B). The pause-free elongation rate of IAPP was determined to be 0.99 ± 0.50 nm/s, whereas IAPP with 19–29 S20G and 8–20 fibrillated at the rate of 0.50 ± 0.32 nm/s and 0.36 ± 0.26 nm/s, respectively (Fig. 4C). According to previously reported mass-per-length (MPL) measurements (using scanning TEM, IAPP protofibrils possessed a mass of ~20 kDa/nm.^{53, 78–79} Similar to other amyloids, IAPP fibrils contain β-sheets with a cross-β alignment relative to the long fibril axis⁸⁰ and a spacing of 4.7 Å between β-sheets strands (Figs. 2J–L).⁷⁸ Thus, it is expected that the MPL of IAPP fibrils to be ƞ·MW/0.47 kDa/nm,⁸¹ where MW is the molecular weight of the peptide and ƞ the number of molecules in each β-sheet spacing. This indicates that ~2.44 IAPP monomers constituted each cross-β repeat. By assuming that fibrils elongated due to the addition of monomers to one end, we converted pause-free elongation rates to absolute elongation rates by E_a=2.44·E/0.47 monomers per second. Thus, the absolute elongation rate of pure IAPP (1 μM) was about 5.1 monomers·s⁻¹, while co-aggregation of IAPP with 19–29 S20G and 8–20 fragments at the same total molar concentration of monomers (1 μM) was considerably lower, at 2.6 and 1.9 monomers·s⁻¹, respectively.

Effect of hetero-aggregation on IAPP toxicity in vivo.

The use of zebrafish embryos as an in vivo model has led to remarkable progress in toxicology and genetic studies,⁸² and has recently been applied to the study of amyloidogenesis taking advantage of its high fecundity, well characterized developmental stages, transparency of embryos, and multicellular and multiorgan compositions.^{42–43, 83–85} Here an in vivo toxicity assay was performed by microinjecting IAPP with or without 19–29 S20G and 8–20 fragments into the chorionic fluids of 2 hpf zebrafish embryos. Embryonic survival, hatching and developmental phenotypic abnormalities (such as pericardial oedema and axial curvature) were quantified in Figure 5. Both fragments exhibited negligible toxicity to the embryos compared to embryos injected with distilled water as a negative control (Fig. 5A). In addition, hatching of larvae at the expected developmental stage of 3 days post-fertilisation (dpf) was timely for both of these treatment groups (Fig. 5B), and no statistically significant level of developmental abnormalities was detected (Fig. 5C). In the case of IAPP, embryonic survival was 67 ± 9 %, hatching was notably delayed (70 ± 11 %), and the frequency of phenotypic abnormalities observed was 25 ± 16 % higher than in the control group. Hetero-aggregation with 8–20 did not mitigate IAPP toxicity (survival: 66 ± 12 %), hatching (69 ± 6 %), and a greater proportion of embryos exhibited hallmark toxicity phenotypes than those exposed to IAPP alone (46 ± 16 %). In contrast, co-aggregation IAPP and 19–29 S20G increased the rate of survival of embryos relative to controls (77 ± 11 %), the hatching rate was 92 ± 14 %, and the rate of phenotypic abnormalities was comparable to the frequency observed in the control group. These findings with 19–29 S20G are reminiscent of the potency of peptide KLVFF in mitigating the aggregation and toxicity of Aβ, a major strategy developed against AD.^61–63

Figure 5.

Determination of in vivo toxicity of IAPP with or without the fragments by the zebrafish embryo toxicity assay. Rates of (A) survival, (B) hatching and (C) developmental abnormalities of zebrafish embryos were qualified until 72 h after administration of treatments (related to fully hatched, untreated embryos) by microinjection of 1 nL of 5 μM IAPP with or without 19–28 S20G and 8–20 fragments into the chorion of 2 h post fertilisation (hpf) embryos. (D) Representative images of embryonic development (0–72 hpf), dead, unhatched and phenotypically abnormal embryos. Assays were carried out in triplicate. Error bars show the standard deviations of the averaged data sets (*: P ≤ 0.05, **: P ≤ 0.01, ****: P ≤ 0.001).

Aberrant aggregation of human IAPP and amyloid deposition in the pancreas is a hallmark of T2D.⁸⁶ As the amyloid fibrils formed in vivo are highly heterogeneous in composition and structure,^{1, 28, 53} in this work we investigated the effects of two major amyloidogenic IAPP fragments on the fibrillization kinetics, dynamics and mesoscopic structure of full-length IAPP.⁴⁷ Specifically, 19–29 S20G elevated the flexibility, retarded the elongation (both elongation rate and dwell time) and altered the morphology of IAPP fibrils. In contrast, the secondary amyloidogenic IAPP domain 8–20 retarded the elongation (both elongation rate and dwell time), reduced fibrillar flexibility, but had little impact on the IAPP morphology.

Previous literature has shown that the mechanical properties of amyloid fibrils are related to their steric zipper patterns.⁸⁷ We propose that co-aggregation with the fragments induced defect sites within IAPP fibrils (Figs. 2M–O). Furthermore, it has been previously shown that fibrils formed from the 19–29 S20G fragment were structurally related to that of full-length IAPP and comprised the spine of the mature fibrils.¹⁹ Here we further revealed that hetero-amyloidosis altered not only the final mesoscopic structure and mechanical property of the amyloid fibrils but also the elongation process itself. Furthermore, we also demonstrated that hetero-aggregation with 19–29 S20G can alter the toxicity profile of IAPP in vivo. Together, these observations have provided valuable new insights into the heterogeneity, complexity and dynamics of protein aggregation on the single-molecule and sub-second levels. This new knowledge of hetero-amyloidosis may prove beneficial for developing alternative strategies in the mitigation and prevention of T2D and other amyloid diseases.⁸⁸