Experimental and Computational Protocols for Studies of Cross-Seeding Amyloid Assemblies

Abstract

Alzheimer’s disease (AD) and type 2 diabetes (T2D) are two common protein aggregation diseases. Compelling evidence has shown a link between AD and T2D, which may derive from interspecies cross-sequence interactions between amyloid-β peptide (Aβ), associated with AD, and human islet amyloid polypeptide (hIAPP), associated with T2D. Herein, we present experimental and computational protocols and tools to study the aggregate structures and kinetics, conformational conversion, and molecular interactions of Aβ-hIAPP mixtures. These protocols could be generally applied to other cross-seeding behaviors of amyloid peptides.

1. Introduction

Amyloid peptides can misfold and aggregate into conformationally similar amyloid fibrils consisting of characteristic cross-β-sheet structures. Fibrils formed by different amyloid peptides are associated with more than 20 neurodegenerative and other protein aggregation diseases [1], Among them, Alzheimer disease (AD) and type 2 diabetes (T2D) are the two common age-related chronic disorders [2, 3], both affecting millions of people globally. Accumulating evidence has shown that ~81% of AD patients often have a higher risk of developing type II diabetes (T2D), and vice versa [4, 5]. Although the precise mechanisms linking these two disorders are still unclear, the evidence suggests a potential pathological link between AD and T2D.

Considering that amyloid peptides share a common aggregation process and their aggregates share similar structural hallmarks, the AD-T2D link could arise from cross-species interactions between the amyloid-β peptide (Aβ) [6, 7] and the human islet amyloid polypeptide (hIAPP or amylin) [8–11], which are associated with AD and T2D, respectively. Several lines of evidences appear to support the hypothesis of the cross-species interaction between Aβ and hIAPP: (1) Aβ and hIAPP coexist in the serum and cerebrospinal fluids with similar concentrations [6]; (2) Aβ co-localizes with hIAPP in pancreatic islet amyloid deposits of T2D patients [7]; and (3) hIAPP and Aβ show high degrees of sequence identity (25%) and similarity (50%), making it possible for them to interact with each other to form hybrid amyloid fibrils [6, 12]. Similar cross-seeding behavior between different amyloid species (Aβ and α-synuclein [13], Aβ and tau [14], and hIAPP and insulin [15]) and even between bacterial curli and amyloid peptides of SEVI, Aβ, and hIAPP have been reported [16]. This fact at least suggests that amyloid cross-seeding sets the scene to reveal some general principles governing heterogeneous protein misfolding and aggregation.

Only a few studies have shown that Aβ and hIAPP may cross-seed fibrillization, but with different cross-seeding efficiencies, strongly depending on experimental conditions (seeding concentrations, sequence specifity, and even agitation) [17–20]. Different seeding abilities of Aβ vs. hIAPP aggregates suggest the existence of different cross-species barriers for different amyloid species to interact with the other and form hybrid amyloids [21]. Given the current state of knowledge of the cross-seeding of amyloid peptides, it still remains unclear how Aβ and hIAPP interact with each other to induce cross-seeding behaviors. Herein, we detail computational and experimental protocols, obtained from our recent works, to probe the aggregation kinetics of Aβ/hIAPP mixtures by fluorescent dye thioflavin-T (ThT), the fibril morphologies by atomic force microscopy (AFM), the secondary structure transformation by circular dichroism (CD), and atomic-scale structures and association by molecular dynamics (MD) simulations. This chapter will hopefully provide robust and general strategies to study the cross-species amyloid aggregation by different amyloid peptides, which may help to reveal a possible link between different neurodegenerative diseases.

Understanding the cross-species aggregation mechanisms between Aβ and hIAPP peptides is critical for revealing a potential pathological link between AD and T2D. In this chapter, we report the details of computational and experimental tools, protocols, and conditions to study how Aβ and hIAPP peptides interact with each other and how such cross-sequence interactions affect the structures of their aggregates and pathways. Collective data demonstrate the existence of cross-seeding interaction and aggregation between Aβ and hIAPP, yielding hybrid Aβ – hIAPP amyloid fibrils morphologically similar to pure Aβ or hIAPP fibrils. Further molecular simulations reveal different atomic structures of hybrid Aβ – hIAPP fibrils, indicating that the complex assembly of Aβ and hIAPP produces polymorphic amyloid cross-seeding species through different aggregation pathways. The cross-seeding effects in this work may provide a molecular explanation for a potential link between AD and T2D and hopefully lead to therapeutic advances that could treat different neurodegenerative diseases. Technically, the tools, protocols, and conditions we used for Aβ-hIAPP cross-seeding could be readily applied to other amyloid peptides.

2. Materials

Prepare ultrapure water by purifying deionized water to obtain a sensitivity of 18 MΩ cm at 25 °C; use this ultrapure water for all aqueous solutions. All chemical reagents were purchased at the analytical grade and stored at room temperature (unless indicated otherwise).

2.1. Experimental Materials

1. Beta-amyloid (Aβ_1–42, ≥95.0%) and human islet amyloid polypeptide (hIAPP_1–37, ≥95.0%) (American Peptide Inc., Sunnyvale, CA).

2. l,l,l,3,3,3-Hexafluoro-2-propanol (HFIP, ≥99.9%).

3. NaOH solution: 10 mM NaOH in water (see Note 1).

4. PBS buffer: 10 mM, pH 7.4. Phosphate buffered saline (pH 7.4) dry powder was purchased in foil pouches. Dissolve each pouch in 1 L of water to yield 10 mM phosphate buffered saline, NaCl 0.138 M, KCl 0.0027 M. Store at 25 °C (see Note 2).

5. Thioflavin T (ThT) Tris-buffer: 10 μM ThT, pH 7.4. Add about 100 mL water to a 1 L amber bottle. Weigh 1.22 g Tris and transfer it to the bottle. Add water to a volume of 900 mL. Mix and adjust pH to 7.4 by carefully adding concentrated HCl. Make up to 1 L with water to produce a 1 L 10 mM Tris-buffer.

6. ThT-water solution (2 mM): add 33 mg of ThT powder into 50 mL of water. Dilute 250 μL of this ThT-water solution into 50 mL of Tris-buffer resulting in a final concentration of 10 μM ThT Tris-buffer (see Note 3).

3. Methods

3.1. Purification and Preparation of Peptides

1. Obtained both Abeta and hIAPP as a lyophilized form and each peptide of 1 mg was stored separately in a small amber glass bottle in a −20 °C refrigerator.

2. Add 1 mL HFIP into each glass bottle with 1.0 mg Aβ or hIAPP peptide inside and seal the bottle cap with parafilm. Allow the peptide to fully dissolve in HFIP solution for 2 h, followed by 30 min sonication to make peptides in a monomeric state in the solution (see Note 4).

3. Transfer the 1 mL peptide-HFIP mixture to a 1.5 mL centrifuge tube, and centrifuge the mixture solution at 13,148 × g for 30 min at 4 °C.

4. Extract 80% (0.8 mL) of the top solution, save it in four 500-μL micro tubes, each containing 0.2 mL of the top solution.

5. Freeze the extracted solutions in a −80 °C refrigerator and freeze-dry at −110 °C and 10 mTorr.

6. Extract 0.2 mg dry peptide powder from the micro tube and lyophilize in a −80 °C refrigerator [22] (see Note 5) (Fig. 1).

Fig. 1.

Preparation procedure of amyloid peptide purification

3.2. Peptide Incubation

1. Add 25 μL of 10 mM NaOH solution to 0.2 mg purified Aβ powder in a microfuge tube, and sonicate the mixture for 1 min to obtain a homogeneous Aβ-NaOH solution.

2. Add 25 μL of 10 mM NaOH solution into 0.2 mg purified hIAPP powder in the micro tube and sonicate the mixture for 1 min at 13,148 × g to obtain a homogeneous hIAPP-NaOH solution (see Note 6).

3. Add the 25 μL Aβ-NaOH solution into 2 mL of 10 mM PBS buffer solution in 2 mL centrifuge tube at 4 °C, and mix using vortex. Right after mixing, add 25 μL hIAPP-NaOH solution to the previous Aβ-NaOH-PBS solution, and mix using vortex.

4. Centrifuge the Aβ-hIAPP mixture solution at 13,148 × g for 30 min at 4 °C to remove any potential oligomers, followed by extracting 90% of the top Aβ-hIAPP solution (1.8 mL) in the monomeric state of both peptides for cross-seeding tests [23] (Fig 2).

Fig. 2.

Procedure of amyloid peptide incubation

3.3. Monitor Amyloid Formation Using Thioflavin T Fluorescence Assay (ThT)

1. Add 3 mL ThT Tris-buffer to a 1 cm × 1 cm quartz cuvette and install the cuvette in the sample chamber. Measure the emission curve between 470 and 500 nm with an excitation wavelength of 450 nm. The emission curve is used as a background control (see Note 7).

2. Add 60 μL of incubated peptide solution, obtained from “peptide incubation,” to the ThT Tris-buffer in the cuvette and mix well. Measure the emission curve and subtract the background curve to obtain a true-fluorescence intensity at this time point [24]. Each peptide sample is tested in triplicate (see Note 8).

3. Repeat steps 1–2 at different time points, collect all the data to do the analysis of variance, use graphics software (Excel, Origin et al.) to plot a time-dependent ThT curve to quantify the aggregation kinetics of the cross-seeding of Aβ and hIAPP (Fig. 3a-c).

Fig. 3.

Time-dependent ThT fluorescence curves for pure Aβ, pure hIAPP, and mixed Aβ/hIAPP at different concentration of 12.5 mM, 25 mM, and 50 mM. Reprinted with permission from ref. [23]. (Copyright 2015 ACS Publication)

3.4. Circular Dichroism Spectroscopy (CD)

1. Obtain CD spectra by scanning between 250 and 190 nm at a 0.5 nm resolution and 50 nm/min scan rate. Correct all spectra by subtracting the baseline and report the data as the average of three successive scans for each sample.

2. Add 160 μL incubated peptide solution into a rectangular quartz cuvette with a 1 mm path length without dilution (see Note 9).

3. Repeat steps 1–2 at different time points to obtain a time-dependent CD curve to quantify the conformational change of the cross-seeding of Aβ and hIAPP.

4. The secondary structure contents from the CD spectra can be estimated using the self-consistent method (CDSSTR program) in the CD proanalysis software [25, 26] (Fig. 4a–f).

Fig. 4.

Time-dependent far-UV CD spectra for (a) pure Aβ (25 μm), (b) pure Aβ (50 μm), (c) pure hIAPP (25 μm), (d) pure hIAPP (50 μm), (e) mixed Aβ/hIAPP (25/25 μM), and (f) mixed Aβ/hIAPP (12.5/12.5 μm). Reprinted with permission from ref. [23]. (Copyright 2015 ACS Publication)

3.5. Tapping-Mode Atomic Force Microscopy (AFM)

1. All images are acquired in tapping mode as 512 × 512 pixel images at a typical scan rate of 1.0–2.0 Hz with a vertical tip oscillation frequency of 250–350 kHz [27].

2. Take 20 μL incubated peptide solution and deposit the solution onto a freshly cleaved mica substrate for 1 min.

3. Gently rinse the surface with 10 mL of ultrapure water to remove salts and loosely bound peptides, dry the surface with compressed air for 1 min, and leave the sample to air-dry before AFM imaging.

4. Representative AFM images are collected by scanning at least six different locations of different samples (Fig. 5).

Fig. 5.

Representative AFM images for pure Aβ, pure hIAPP, and mixed Aβ/hIAPP aggregates after incubation for 0, 4, 8,12, and 24 h. Reprinted with permission from ref. [23]. (Copyright 2015 ACS Publication)

3.6. Simulation Protocols

3.6.1. Aβ Pentamer

1. Obtain the atomic structure of Aβ pentamer from the Protein Data Bank (PDB) database (ID:2EBG) [28].

2. Add the N-terminal missing residues in an in-register manner in the β-sheet conformation using the GaussView program (see Notes 10 and 11, Fig. 6).

3. Cap the N- and C-termini with NH₃⁺ and COO⁻ groups. Assign the tautomeric state of His residues based on local environment.

4. Use a 500-step steepest descent (SD) minimization with backbone atoms fixed to relax the initial structure of the Aβ pentamer (see Note 12).

Fig. 6.

All-atom and CG-based structures and sequences of hIAPP and Aβ peptides

3.6.2. hIAPP Pentamer

1. The initial coordinates of hIAPP monomer were kindly provided by the Tycko lab [29].

2. Five hIAPP monomers are parallel-packed on top of each other in an in-register manner with inter-peptide distance of 4.7 Å (see Note 11).

3. Construct the intramolecular disulfide bond between Cys2 and Cys7 at the N-termini of hIAPP peptides. The tautomeric state of His residues is assigned based on local environment in the hIAPP monomer.

4. Perform a 500-step SD minimization with backbone atoms fixed to refine the overall structure (see Note 12, Fig. 6).

3.6.3. Coarse-Grained Model of Aβ Pentamer

1. First, perform a 100-ns all-atom explicit-solvent simulation of Aβ to obtain the equilibrium structure of Aβ pentamer in bulk water.

2. Extract 500 all-atomic structures of the Aβ pentamer from the last 20-ns MD trajectory excluding water molecules. Each Aβ structure is energy minimized with 500-step SD minimization to remove any structural clashes.

3. Obtain an average structure of Aβ pentamer by averaging the 500 structures above, followed by 500-step SD minimization to refine the average structure.

4. Use the elastic network method with the Martini force fields (version 2.4) to convert the all-atom average structures of Aβ pentamer into the CG structure of the Aβ pentamer (see Note 13, Fig. 6).

3.6.4. Coarse-Grained Model of hIAPP Pentamer

1. First, perform 100-ns all-atom explicit-solvent simulation of hIAPP to obtain the equilibrium structure of the hIAPP pen tamer in bulk water.

2. Extract 500 all-atomic structures of the hIAPP pentamer from the last 20-ns MD trajectory excluding water molecules. Energy minimize each hIAPP structure with 500-step SD minimization to remove any structural derivations.

3. Obtain an average structure of the hIAPP pentamer by averaging the 500 structures above, followed by a 500-step SD minimization to refine the average structure.

4. Use the elastic network method with the Martini force fields (version 2.4) to convert the all-atom average structures of the hIAPP pentamer into the CG structure of the hIAPP pentamer (see Note 14, Fig. 6).

3.7. Simulation Methods

3 7.1. Coarse-Grained Replica Exchange Molecular Dynamics (CG-REMD) Simulations of Hybrid Aβ-hIAPP Assemblies

1. Separate the CG Aβ and hIAPP pentamers from each other by random intermolecular distances and orientations.

2. Place the Aβ-hIAPP system in a simulation box of 100 × 100 × 100 Ǻ³ with the three-dimensional periodical boundary condition (3D-PBC).

3. Fully solvate the simulation system with coarse-grained water molecules with minimal margins of 15 Å from any protein atom to any edge of water box.

4. Add Cl⁻ and Na⁺ counterions into the systems to mimic an ionic strength of ~150 mM.

5. Structurally optimize and relax the resulting system using 5000 steps of the steepest decent minimization.

6. Heat the energy minimized system gradually from 0 to 310 K by 1-ns MD simulations, constraining the peptide backbones using the NPT ensemble (T = 310 K and P = 1 atm) and under periodic boundary conditions.

7. For the production run, treat the Aβ-hIAPP system under NPT ensemble and 3D-PBC conditions using CG-REMD simulations with the Gromacs-4.6.5 software package. A total of 40 independent replicas with temperatures ranging from 305 to 410 K should be simulated simultaneously for each Aβ-hIAPP system (see Note 15).

8. The exchange rate between two replicas is attempted every 1000 integration steps and the acceptance ratio for exchanging replicas varied between 0.20 and 0.25.

9. Control the temperatures and pressure (1 atm) by the V-rescale and Parrinello-Rahman methods with coupling constants of 1.0 and 12.0 ps, respectively. Conduct the temperature coupling separately for protein and nonprotein atoms (i.e., ions and water molecules).

10. Leapfrog integrator was utilized to allow an integration time step of 4 fs (Fig. 7).

Fig. 7.

Schematic illustration of the CG-REMD simulations

3.7.2. CG Hybrid Aβ-hIAPP Assemblies

1. From the CG-REMD simulations, construct the free energy landscape that describes the interaction between hIAPP and Aβ using –RT log(H(x,y)), where H(x,y) is the histogram of two selected reaction coordinates: x is root-mean-square derivation (RMSD) and y is the radius of gyration (Rg) (see Notes 16 and 17).

2. Based on the free energy landscape, identify the four most populated CG Aβ-hIAPP hetero-assemblies (double-layer, elongation, tail-tail, and block modes) at different energy potential wells (Fig. 8).

Fig. 8.

CG-based free energy landscape along Aβ and hIAPP aggregation process as a function of two reaction coordinates of RMSD and Rg. Reprinted with permission from ref. [30]. (Copyright 2015 ACS Publication)

3.7.3. Conversion of CG Aβ-hIAPP Models to All-Atom Aβ-hIAPP Stmctures

1. Based on the most populated CG Aβ-hIAPP hybrids obtained from the CG free energy landscape, we developed a two-step strategy to convert the CG Aβ-hIAPP structures to all-atom Aβ-hIAPP ones.

2. The first step of backmapping is to generate an initial, all-atom structure. In the CG model, the center of mass of the groups of atoms are represented by a single CG bead as the center of mass location of backbone (Bcom), side chains (Scorn), and/or other peripheral groups (Pcom) for each residue. Then, the centers of mass of all-atom backbones, sidechains, and/or peripheral groups of each residue of the same hIAPP or Aβ peptides are superimposed into Bcom, Scorn, and/or Pcom, and all atoms are assigned coordinates according to standard geometries using GROMACS around these mapped atoms. This backmapping step produces an initial all-atom of A-β-hIAPP structure from the CG one. The resulting structure is then optimized by energy minimization to remove any bad contacts and to correct internal coordinates for all bonded atoms (steps 1 and 2 in Fig. 9, see Note 18).

3. The second step is to optimize interfacial contacts between Aβ and hIAPP. The resulting all-atom structure obtained from the first stage is further recast into the initial all-atom Aβ or hIAPP models to maximize residue-residue contacts by minor structural adjustments through structural superimposition and rotation, followed by energy minimization. In this way, the structural disorder of the resulting all-atom structures, caused by the high temperatures in CG-based REMD simulation, is minimized (steps 2 and 3 in Fig. 9).

Fig. 9.

Schematic illustration of the conversion of Aβ-hIAPP cross-seeding assemblies from coarse-grained to all-atom models

3.7.4. All-Atom MD Simulations of Aβ-hIAPP Models

1. The obtained all-atom Aβ-hIAPP models from the CG models are then subjected to all-atom explicit-solvent simulations to validate the models; perform these simulations using the Gromacs-4.6.5 program with the CHARMM27 force field with CMAP correction and SPCE solvent models [31, 32].

2. Solvate all-atom Aβ-hIAPP systems in a cubic water box with minimal margin of 15 Ǻ from any protein atom to any edge of the water box. Add Cl⁻ and Na⁺ ions into the systems to mimic ~150 mM ionic strength.

3. The resulting systems are structurally optimized and relaxed by 50,000 steps of the steepest decent minimization.

4. Gradually heat the energy minimized systems from 0 to 310 K by 1-ns MD simulations with constrained peptide backbones.

5. Conduct MD simulations for 60 ns using the NPT ensemble (T = 310 K and P = 1 atm) and under periodic boundary conditions. The Parrinello-Rahman and the V-rescale methods are used to maintain a constant pressure of 1 atm and a temperature of 310 K.

6. Constrain all covalent bonds involving hydrogen atoms by the LINCS method. Short-range Van der Waals (VDW) interactions should be described by the smoothly truncated method via potential shift at 14 Å. Long-range electrostatic interactions should be calculated by the particle mesh Ewald (PME) method with a grid spacing of 0.16 Å and a real-space cutoff of 12 Å.

7. Integrate the equations of motion using the Leapfrog integrator with a time step of 2 fs. Structures should be saved every 2 ps for analysis.

3.7.5. Structural and Population Analysis of All-Atom Aβ-hIAPP Models

1. Conduct in-house Monte Carlo simulations to calculate the relative populations for the Aβ-hIAPP models. First, 500 structures per each all-atom Aβ-hIAPP model are extracted from the last 10-ns explicit-water MD trajectories, excluding water molecules [30].

2. All extracted structures (i.e., all conformers) from all Aβ-hIAPP models are used to construct an energy landscape of Aβ-hIAPP assemblies for evaluating the probability of conformers using an in-house Monte Carlo program as reported in our previous studies [33–35].

3. Calculate the conformational energies of all conformers using the generalized Born method with molecular volume (GBMV) with the CHARMM force field, where the dielectric constant of 80 is used for water and the hydrophobic solvent accessible surfaces area (SAAS) term factor is set to 0.00592 kcal/mol Å [2], This GBMV free energy calculation generates a Markov chain of all conformers [E₁, E₂… E_i, …E_n].

4. For conformer population analysis, randomly select any two structures from conformers i and j and calculate their conformational energies using the GBMV model. Then, the metropolis algorithm is applied to determine the transition probability (i.e., acceptance probability) from conformer i to conformer j using $P_{\text{accept},\text{move}}=\min \left[1,\exp \left(\frac{E_i-E_j}{k_{B}T}\right)\right]$. After one million step samplings, a total number of “accepted” structures (N_i) for any conformer i is counted, and the relative probability of conformer i occurring in the energy landscape is evaluated as p_i = N_i/N_total (Fig. 10).

Fig. 10.

Schematic illustration of the calculations of the populations for different Aβ-hIAPP hybrid assemblies