Simulating the first steps of amyloid peptides aggregation Jessica Nasica Département de physique - Université de Montréal GEPROM – 2ième réunion scientifique May 27th, 2009

Overview of the presentation Medical interest Problem of amyloid fibril formation Our simulations & results

Protein misfolding and aggregation Many protein-misfolding diseases include conditions where a protein forms insoluble aggregates, amyloid fibrils, that deposit toxically Amyloid diseases e.g. : Alzheimer’s Parkinson’s Huntington type II diabetes Dobson et al. (2003), Nature 426

Amyloid fibril formation Amyloid fibrils insoluble fibrous aggregates with a highly organized Macrostructure made of β-sheets. Mechanism of formation  3 steps: Alignment of the molecules to form β-sheets  fastest stage  involves H-bonds Formation of the cross-β structure  slower than step 1  involves Van-der-Waals forces  interdigitation of residues side chains  “steric zipper” structure Fibril formation  involves non-covalent bonds Amyloid fibril formation is a nucleated-growth process stabilized by the protein concentration and by the formation of steric zippers Nelson et al. (2005), Nature 435

Amyloid fibril polymorphism For the Aβ40 sequence, they construct a full molecular model showing 2 distinct possible morphologies for the fibril structure. Paravastu et al. (2008), PNAS 105 & Petkova et al. (2006), Biochemistry 45

Amyloid oligomers intermediate states during the formation of amyloid fibrils thought to be more toxic than the fibrils themselves Different types of soluble amyloid oligomers share a common structure  suggests they share a common toxicity mechanism Lashuel et al. (2002), Nature 418 Kayed et al. (2003), Science 300

Theories on intermediate oligomer states Oligomers as protofibrils Oligomers as pores  may form annular pore-like structures to go through the cell membranes : β-barrel Esposito et al. (2006), PNAS 103 Amyloid oligomer pore structure observed experimentally (left) & numerically (right) Lashuel et al. (2008), Nature 418 Irbäck et al. (2007), Proteins 71

The “short peptides” approach Nelson et al. (2006), Current opinion in structural biology 16 The gain-of-interaction model for amyloid structure suggests that only a small portion of a native protein is responsible for amyloid fibrils. Under conformational changes this small portion is exposed & binds to an identical portion on another molecule builds up a fibril or Cross-β spine (no domain swapping) Cross-β spine with domain swapping For the budding yeast Sup35p fibril-forming protein GNNQQNY GNNQQNY The essential element involved in the fibril structure is the small portion

Experiments done on GNNQQNY Atomic structure of cross-β spine constructed from X-ray diffraction analysis Fomation of a double β-sheet with parallel β-stands Side chains form a self-complementing steric zipper Interdigitation of the side-chains would stabilize the sheets Nelson et al. (2005), Nature 435

Our short-term goals To understand the aggregation process of short peptides Kinetics of aggregation Final structures (morphologies accessible) To study different sizes of systems Trimer GNNQQNY Pentamer GNNQQNY 20-mer GNNQQNY 50-mer GNNQQNY

Our simulation methods Replica exchange MD launch n molecular dynamical simulations in parallel, at n different temperatures at regular intervals, try an exchange of configurations between two adjacent temperatures using a Metropolis accept-reject criterion REMD accelerates sampling (in some cases) for the cost of losing dynamical information. REMD still provides thermodynamical information Final structures of simulations tested with an all-atom potential (GROMACS)

Our simulations : small aggregates We observe a strong tendency to form planar β-sheets GNNQQNY trimer GNNQQNY pentamer

Our simulations : bigger aggregates Formation of twisted pair-of-sheet “protofibril-like” structures Formation of β-barrel-like structures GNNQQNY 20-mer – 300 ns simulation Similar to the atomic structure described by Nelson et al. (2005)

GNNQQNY 20-mer results We observe a nucleated-growth aggregation process & a clear loss of entropy as the energy of the system drops rapidly

GNNQQNY 20-mer results β-sheets favor a parallel orientation of the β-strands Consistent with the atomic description given by Nelson et al. (2005) Statistics over 300 ns

Conclusions We have described the first steps of aggregation of GNNQQNY. The results are consistent with experimental results on that sequence. Polymorphism exists and we already see a clear separation between 2 different semi-organized structures. The obtained β-barrel-like structures might not be on the fibril formation pathway. It seems it is a separate possible morphology. For small aggregates (trimers, pentamers), the study of 2 other sequences (SSTSAA,SNQNNF) shows similar features compatible with the GNNQQNY results.

Future work Lots of statistics obtained from our 20-mer simulations  a lot of the analysis is still being done to understand The formation process The triggering of the nucleated-growth process Simulation of the GNNQQNY 50-mer Simulation on other short sequences (SSTSAA, SNQNNF,…)

Acknowledgments Collaborators MONTREAL Normand Mousseau PARIS Philippe Derreumaux MILAN Giorgio Colombo Massimiliano Meli (Ph.D.) Funding GEPROM CRSNG FESP Resources Réseau québécois de calcul de haute performance Special thanks to: Rozita Laghaei (post-doc), Lilianne Dupuis (Ph.D.) and Jean-François St- Pierre (Ph.D.)