Nano-assembly of amyloid β peptide: role of the hairpin fold

Structural investigations have revealed that β hairpin structures are common features in amyloid fibrils, suggesting that these motifs play an important role in amyloid assembly. To test this hypothesis, we characterized the effect of the hairpin fold on the aggregation process using a model β hairpin structure, consisting of two Aβ(14–23) monomers connected by a turn forming YNGK peptide. AFM studies of the assembled aggregates revealed that the hairpin forms spherical structures whereas linear Aβ(14–23) monomers form fibrils. Additionally, an equimolar mixture of the monomer and the hairpin assembles into non-fibrillar aggregates, demonstrating that the hairpin fold dramatically changes the morphology of assembled amyloid aggregates. To understand the molecular mechanism underlying the role of the hairpin fold on amyloid assembly, we performed single-molecule probing experiments to measure interactions between hairpin and monomer and two hairpin complexes. The studies reveal that the stability of hairpin-monomer complexes is much higher than hairpin-hairpin complexes. Molecular dynamics simulations revealed a novel intercalated complex for the hairpin and monomer and Monte Carlo modeling further demonstrated that such nano-assemblies have elevated stability compared with stability of the dimer formed by Aβ(14–23) hairpin. The role of such folding on the amyloid assembly is also discussed.

S3 mixture was too crowded, while the surface with a 1:100 ratio showed only a very few bright spots. However, the surface with a MAL-PEG/mPEG mixture of 1:60 had a reasonable amount of coverage; therefore, this ratio was chosen for further experiments.

Photo-bleaching and blinking tests
Photo-bleaching and blinking experiments were performed to test the photo-physical properties of the Cy3 dye by covalently attaching the Cy3-labeled Aβ(14-23) monomer to the surface and performing TIRF imaging for an extended time. The glass coverslip was cleaned with chromic acid and water, followed by treatment with a 167 µM APS solution.
MAL-PEG-SVA/mPEG-SVA mixture (167 µM) with a molar ratio of 1:60 was added to the surface and incubated for 1 h, followed by washing with water. A TCEP treated 50 pM Cy3-labeled Aβ(14-23) monomer solution was added to the surface, which allowed the maleimide-thiol reaction to occur. The surfaces were washed, assembled into the instrument, and 100 µL of 10 mM sodium phosphate buffer was added to the surface. TIRF movies were recorded from the surface for 3 min. The data suggested that no photobleaching or blinking occurred during the acquisition (Movie S3). A time trajectory is shown in Fig. S4, which indicates no bleaching or blinking events.

Specific vs. nonspecific binding of the Cy3-labeled hairpin on the modified surface
Specific binding of the fluorophore-labeled hairpin for H-M complex formation was compared with the nonspecific binding of the fluorophore-labeled hairpin with the surface, which does not have any covalently bound monomers. The nonspecific binding of the Cy3labeled hairpin was tested by placing the fluorophore-labeled hairpin solution on the mPEG-coated surface and recording a TIRF video over time. The glass coverslip was cleaned, treated with APS, and treated with only mPEG-SVA. Note that the coverslip was S4 not treated with MAL-PEG-SVA. The mPEG-treated surface was washed, placed on the TIRF stage and 1 nM of the Cy3-labeled hairpin solution was added. A TIRF video was then recorded. Only a very few bright spots were observed on the surface, indicative of nonspecific absorption. Fig. S5 indicates a large number of specific interactions on the surface modified with the monomer (Fig. S5a), but only a very few spots were observed on the mPEG-coated surface (Fig. S5b). These results clearly demonstrate that the complexes identified during probing experiments were from specific interactions.

MD simulation of H-M and H-H complex assembly.
We applied all-atom MD simulations with the explicit TIP3P water model to both systems using the approach described in our recent publications 1 . First, we performed MD simulations of the fully stretched hairpin for 1.2 µs to obtain the most probable conformations of the hairpin ( Fig.   S10 and S11). Next, the most representative hairpin structure was used to assemble the H-M complex with the monomeric structure obtained in reference 1 . In parallel, two copies of the hairpin were used to assemble the H-H complex. Each of the systems were simulated for 2.4 µs using conventional MD (cMD) simulations, followed by 500 ns of accelerated MD (aMD) simulation to extend the conformational sampling efficiency by several orders of magnitude 2 . This simulation scheme was recently used to successfully characterize the interaction of full-size Aβ42 3 .
We performed cluster analysis based on the dihedral Principle Component Analysis