We first became interested in how amyloid beta (Abeta) peptides cause Alzheimers disease in the early 90s when Nelson Arispe and Harvey Pollard, who were then at NIH, approached us with their finding that Abeta peptide forms ion channels in bilayers. In 94 we coauthored highly tentative models of the Abeta channels. Back then the consensus was that Alzheimers was caused by fibrils formed by Abeta. Much has been learned since then about the structures of Abeta assemblies and ion channels, and how they relate to Alzheimers disease. Several other groups also found that Abeta forms channels and have microscopically observed Abeta assemblies in synthetic membranes. Heavy metals and peptides that inhibit the Abeta channels also inhibit Abeta-induced apoptosis of neurons. Microscopy studies have also revealed superstructures of numerous protofibril assemblies that form in solution. Solution NMR studies of Abeta monomers in apolar solvents and solid state NMR studies of fibrils have helped elucidate their structures. Given these new data, some of which are inconsistent with our original models, and the biomedical importance of amyoid-associated neurodegenerative diseases such as Alzheimers, we decided to revive our modeling efforts. We became interested in the prions when others noted sequence similarities in the hydrophobic segments of Abeta and the human prion precursor, Prion Protein, and found that the toxic form, PrPSc, could self-assemble into hexagonal lattices similar to those we had developed in modeling Abeta assemblies. We have constructed the following types of soluble Abeta assemblies: dimers, trimers, tetramers, hexamers, strings of hexamers, dodecamers, AbetaOs (18 subunits), beaded and smooth annular protofibril (36 subunits), protofibrils with mass-per-lengths of 18 and 27kDa/nm that correspond to experimentally determined values, and two-dimensional hexagonal lattices that can extend indefinitely. We have also developed numerous models in which up to 36 Abeta peptides form large assemblies on the membrane surface and then insert through the membrane to form channels, and how heavy metals and synthetic peptides can block these channels. These models were constrained by results of microscopy and biochemical studies of the dimensions, masses, and secondary structures of both soluble and membrane-bound assemblies. Most of our models of assemblies composed of six or more subunits involve hexamers in which the C-termini segments (residues 29-42) form a six-stranded antiparallel beta barrel. However, in our final 36-peptide models of annular protofibrils and channels, six of these barrels merge to form a large 36-stranded beta-barrel). Our collaborators have found that Abeta channels can be blocked by low concentrations of two small drugs (MRS2481 and MRS2485synthesized by Ken Jacobson at NIH) and histadine-containing peptides (e.g., His4). We have modeled how these drugs and peptides may bind in the central region of our favored model of the channels. This is an exciting finding because so far all agents that block the channels also inhibit Abeta-induced neurotoxicity. Thus, the models may be useful for structure-based design of drugs to treat Alzheimer's disease. We have also simulated how some of these assemblies could morph or grow into models of fibrils that are based on solid state NMR studies. We are using both atomically complete and coarse graining molecular dynamic simulations to better understand the structures of these Abeta oligomers, how they assemble, and how larger assemblies can form from smaller assemblies. The atomically complete simulations using GROMACS, CHARMM, or NAMD programs are better for analyzing precise interactions within relatively ordered assemblies;however only short time scales can be analyzed. We have been using coarse-grained methods with Discrete Molecular Dynamics (DMD) simulations to analyze longer time scales and less ordered assemblies. We have also used steered dynamics simulations to analyze transitions between different types of assemblies. Prion Protein (PrP) can exist in two forms, the normal PrPC form and a toxic PrPSc form that causes Creutzfeldt-Jakob disease, bovine spongiform encephalopathy, scrapie, and other spongiform encephalopathies. PrPSc can form different types of assemblies;i.e., two-dimensional hexagonal lattices, fibrils, and membrane-bound. Aspects of the PrP and PrP-associated diseases resemble those of Abeta and Alzheimers. The most hydrophobic portion of the 230 residue-long PrP protein (residues 111-135) has a sequence quite similar to the hydrophobic segment of Abeta that we propose to form a six-stranded antiparallel beta-barrel. Experimental studies indicate that PrP protein can span membranes and that sub-peptides of the hydrophobic region are neurotoxic and can form ion channels in bilayers. PrPSc can assemble in water into a P3 hexagonal lattice that has three parallel axes of three-fold symmetry. We have constructed PrPSc lattice models with the axis of the putative hydrophobic beta-barrels on one of the three-fold axes of a P3 hexagonal lattice. In these models, residues 111-135 form a beta-hairpin and three of theses hairpins associate about an axis of three-fold symmetry to form a six-stranded beta barrel. In addition, a triple-stranded beta helix formed by residues 50-91 is proposed to surround the hydrophobic beta-barrel and shield it from water. The segment forming the beta-helix includes an octapeptide sequence (HGGGWGQP) that is repeated five times. Membrane-penetrating tails of some viral proteins have a repetitive octapeptide sequence that has been shown to form a similar triple-stranded beta helix. These models are consistent with experimental findings that the beta content of N-terminus region increases when PrPC converts to PrPSc. The structure of the C-terminus helical domain of PrPC has been determined in numerous NMR studies and one X-ray crystallography study. In our lattice models, three C-terminus helical domains, based on the NMR structures of PrPC, surround another three-fold axis, with the C-termini alpha-helices forming a triple-stranded coiled-coil, which is a commonly observed structural motif of fibrous proteins. The models were constrained to have the dimensions and lattice type of the experimentally observed lattice. They were constrained further to be consistent with other experimental results, such as formation of the lattice when substantial portions of the PrP protein are deleted. Our models of a membrane bound hexagonal lattice assemble have similar properties. In these models, the hydrophobic beta-barrel spans the bilayer, and a more collapsed beta-helix formed by the octapeptide repeats extends into the aqueous phase. In our fibril models, residues 111-135 form a continuous beta-strand, and six of these strands assembles into a hexameric antiparallel beta barrel. An overlapping dimeric structure of the helical domain observed in the crystal structures was used to model an elongated trimeric helical domain of fibrils. This domain is in series with the elongated beta domains. An alternative model in which the C-termini domains form a triple-stranded beta-helix was also developed. We plan to publish these preliminary models this year, and to perform additional simulations of refine the models.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Investigator-Initiated Intramural Research Projects (ZIA)
Project #
Application #
Study Section
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
National Cancer Institute Division of Basic Sciences
Zip Code
Shafrir, Yinon; Durell, Stewart; Arispe, Nelson et al. (2010) Models of membrane-bound Alzheimer's Abeta peptide assemblies. Proteins 78:3473-87
Shafrir, Yinon; Durell, Stewart R; Anishkin, Andriy et al. (2010) Beta-barrel models of soluble amyloid beta oligomers and annular protofibrils. Proteins 78:3458-72