First principle computations X-ray spectra & diffraction patterns A synergic approach to the structural study of metallo-protein complexes Francesco Stellato I.N.F.N. Roma ‘Tor Vergata’ Milan, Jan 12th 2017
Summary Intro: < 5 minutes for 5 W’s Metal ions, protein misfolding & neurodegenerative diseases In vitro: the Experimental side Serial crystallography at synchrotrons & FELs X-ray Absorption at synchrotrons & FELs In silico: the Computational side Classical molecular dynamics & ab initio calculations X-ray Absorption & Diffraction simulations Budget & Timeline
What?
In words… Protein mis-folding aggregation fibrils *** Experiments XRD & XAS structure and dynamics *** Theory Interpret data in terms of molecular structures & aggregates A- (proteins + metals + water ) classical MD B- (metal binding) QM calculations C- Compute XRD and XAS from QM electron density calculations Aβ, PrP Metal ions (Cu, Zn)
Comprehensive picture of complex phenomena Why? High-Performance Computers & soft-matter theoretical models Brilliant X-rays (Synchrotrons & Free Electorn Lasers - FELs) Connecting two worlds Comprehensive picture of complex phenomena
Who? 10 years experience in -XAS and XRD Experiments at synchrotrons and FELs in Europe and in the US -Developing novel experimental techniques: serial crystallography High performance computing: MD & in silico simulation of in vitro experiments
When? The time is now ! High-performance computing resources Novel FEL and synchrotron X-ray sources Novel FEL and synchrotron X-ray techniques 2000s – Growth of amyloyd nanocrystals 2010 – First X-rays Free Electron Lasers 2011 – First XRD experiments @ FELs 2012 – First X-ray Spectroscopies @ FELs 2014 – First serial XRD @ synchrotrons 2015 – Ab initio calculation of XAS data 2015 – Revival of quantum XRD 2017 – Availability of “cheap” HPC time
Where? INFN Rome “Tor Vergata” Synchrotron (ESRF, Petra III, SLS) Two young scientists directly involved in the project: 1 PhD student working 70% on the project (E. de Santis) 1 Researcher working 30% on the project (V. Minicozzi) Expertise to build, host and run a mid-sized HPC cluster is available on site & environment with strong expertise in X-ray measurements and computational methods (and in combining them): 2 INFN staff (R. Ammendola & G. Salina) 2 Full Professors (S. Morante & G.C. Rossi) > 20 years activity CSN4 and/or CSN5 Synchrotron (ESRF, Petra III, SLS) & Free Electron Laser (FEL) (European XFEL, LCLS) sources around the world
Protein Folding Protein folding is strictly connected to protein functioning. Mis-folding is responsible of Protein Conformational Disorders Even unrelated mis-folded proteins (with homologies neither in sequence nor in structure) build up fibrils with remarkably similar morphology and cross-beta structure
Metals, Mis-Folding & Neurodegeneration Transmissible Encephalopaties Alzheimer’s Disease A peptide plaques in a mis-folded, fibrillar form Abnormally high metal concentration Transmissible Encephalopaties The cellular prion protein folding moves from a cellular form (PrPC) into an aberrant conformer (PrPSc) Cu and Zn are thought to have different effects on peptide oligomerization, aggregation and fibril formation The process is influenced by the presence of metal ions (mainly Cu and Zn) Minicozzi, Stellato et al., JBC 2008 Stellato et al., EBJ 2006 De Santis, JPCB 2015 Morante et al. JBC 2004; Stellato et al., EBJ 2011; Stellato et al., EBJ 2014
Experimental Techniques X-Ray Diffraction (XRD) Atomic resolution structures of whole macromolecule, BUT needs crystals Radiation-damage limited resolution X-Ray Absorption Spectroscopy (XAS) Atomic resolution structures ONLY in the vicinity of metal ions, BUT does not need crystals
Serial Femtosecond X-ray Crystallography Serial: each crystal gives a diffraction pattern Femtosecond: diffraction-before-destruction to overrun radiation damge Crystallography: crystals as small as 300 nm TheX-raydoseneededtoachieveagivenresolutionforaparticularsamplecanbecalculated.Itcanbeshownthatthedoserequiredtoimageasinglebiologicalmoleculeismuchlargerthanthedoserequiredtocompletelydestroythemoleculethrghradiationdamageprocesses.X-raycrystallographersmitigatethisproblembyspreadingthedamageoverbillionsofmoleculesinasinglecrystal,greatlyenhancingthediffractionsignal.Sincethemoleculesareallidenticalandpreciselyalignedinthecrystal,theX-rayscatteringinformationispreservedandthestructurecanbedetermined. LCLSoffersanotherwayaroundthedamageproblem.SincetheFELX-raypulseisveryintenseandveryshort,itispossibleinprincipletodelivertherequireddosetoanano-scalesampleandrecordthescatteredX-rayinformationbeforethedamageprocesseshavetimetodestroythesample.Inotherwords,anLCLSX-raypulsecouldbefocusedontoasinglemolecule,whichwouldbedestroyed–butnotbeforethescatteredX-raysarealreadyontheirwaytothedetectorcarryingtheinformationneededtodeducetheimage.TheCoherentX-rayImaging(CXI)Instrumentwillofferthepossibilityofdeterminingstructuresatresolutionbeyondthedamagelimitforsampleswhichdonotformcrystals,includingimportantclassesofbiologicalmacromolecules Thousands of patterns are recorded and need to be analyzed an HPC challenge Free Electron Laser (FEL) 1013 photons/30 fs FEL beam Chapman et al., Nature 2011
Serial Crystallography at Synchrotrons Using a synchrotron for serial XRD allows spreading the dose over thousands of crystals Micro-crystals, too small for conventional technqiques, can provide useful diffraction Synchrotron 1013 photons/1 s Stellato et al., IUCrJ 2014
Brewster et al. Acta Cryst D 2015 Research Plan / 1 1- Grow amyloid nano- & micro-crystals in different conditions (pH, metal ions) Brewster et al. Acta Cryst D 2015 Nelson et al., Nature 2005 2- Perform (parallel and simultaneous) XAS & XRD measurements (both at synchrotrons & FELs)
Research Plan / 2 serial XRD: high-resolution structures from nano- and micro-crystals Serial XAS: high-resolution structures of the metal binding site Comparison between metal binding sites in crystalline (XRD) and non-crystalline (XAS) form Sample Experimental Technique Deliverables A & prion microcrystals Synchr. serial crystallography High-res electron density A & prion nanocrystals FEL serial crystallography A & prion fibrils w/o metal ions Microscopy, AFM Fibril conformation Non-crystalline A & prion XAS Metal binding site structure Crystalline A & prion
Classical and ab initio calculations Classical Molecular Dynamics Simulations ideal tool to simulate large molecules, but unable to give information on bond formation/breaking Ab initio Calculations (Quantum Mechanics) computationally expensive, but allow to reliably calculate electron densities and analyze bond formation and breaking
Ab initio XAS calculations Initial State Final state with a core-hole Initial state: no core-hole Incident-photon energy Final State La Penna, …, Stellato, J Phys Chem 2015
Ab initio XRD calculations Standard X-ray crystallography methods use free-atom models to calculate mean unit-cell charge densities. Real molecules have charge densities not accurately captured by the routinely used free-atom models Charge density of crystalline urea - Wall , IUCrJ 2016 Charge density of crystalline molecules calculated using Density Functional Theory and compared with experimental data
Research Plan / 3 Comparison of structural data from XAS & XRD experiments and calculations Sample Methodology Deliverables Amyloid multimers (>10 monomers) Classical MD Peptide aggregation Amyloid monomers and dimers Ab initio methods Metal binding site structure XAS spectra and XRD patterns ab initio calculations Computational and experimental data comparison Amyloid with and without metal ions X-ray damage (Coulomb explosion dynamics) Simulation of damage process (Coulomb explosion) and comparison with XAS & XRD experimental data
XAS spectra and XRD pattern calculations: Computational needs Classical MD simulations of 10 A peptides or prions in complex with metal ions in a water environment ( 105 atoms): 0.5x106 core hours + Ab initio calculations for fragments of A peptides or prions in complex with metal ions ( 103 atoms 3x103 electrons ): 106 core hours XAS spectra and XRD pattern calculations: = 2x106 core hours
INFN Rome ’Tor Vergata’ HPC cluster A suitable machine 128 cores 128x24x365 = 1 MCh / year Gives flexibility in HPC time usage and in software optimization dual CPU Intel E5-2697AV4 (4 nodes, each 16x2 cores @ 2.6 GHz) 256 GB RAM DDR4 2400 - 50 TB HD storage Infiniband FDR Frontend node, RAID controller Switch Infiniband: FDR (56 Gb/s) To be hosted in the APE-lab INFN Rome ’Tor Vergata’
Timeline The computational part will start immediately using already available (although limited) computational resources Proposal writing for beamtime access will start immediately. The experiments will start as soon as the samples will be available
Budget Computational side 50.000 € 128 cores parallel computing = 1 MCh/year Experimental side 25.000 € 20.000 € amyloid proteins ( 400 €/mg) 5.000 € wetlab consumables Travel expenses 20.000 € access to light source facilities 5.000 € participation to conferences and workshops Other costs 10.000 € 4.000 € open access publications 6.000 € running costs of the cluster 110.000 € = 60.000 € year 1 + 50.000 € year 2
Conclusion Novel computational methods, high computing power Novel X-ray sources, novel techniques People, Expertise, Ideas Ideas need to be powered… €€€ Pure science: getting closer to understand neurodegeneration Applied science: drug design - Methods: techniques development - Cross-fertilization: experimental & computational scientists interaction
Thank you for your attention
Research Plan / 4 Classical MD & ab initio calculations of Aβ and prions with and without metal ions Calculation of XAS spectra and XRD pattern coming from structures along the MD trajectories taking into account the structural disorder Atomic configuration C1 XAS1 XRD1 Atomic configuration C2 XAS2 XRD2 Average XAS spectrum & XRD pattern … … Atomic configuration Cn XASn XRDn
Computations & Experiments Molecular dynamics (classical and ab initio): strucuture of Aβ & prions in the absence and in the presence of metal ions Several experimental techniques have been used to assess the role of metal ions in A & prions mis-folding and aggregation This project will exploit and combine in vitro & in silico X-ray Absorption Spectroscopy and X-ray Diffraction with Molecular Dynamics
Diffract-and-destroy experiments One pulse, one measure Particle injection A detectable signal must be recorded before the sample is destroyed FEL 1013 photons/30 fs vs Synchrotron 1013 photons/s The diffract-and-destroy idea can work for in principle all synchrotron X-ray techniques FEL puse Diffraction pattern R. Neutze et al, Nature 406 (2000)