Modelling protein Modelling protein-surface interactions: a surface ...
Transcript of Modelling protein Modelling protein-surface interactions: a surface ...
Modelling proteinModelling protein--surface interactions: asurface interactions: achallenge for computationschallenge for computations
A. Calzolari, R. Di Felice, F. Iori, S.Corni
INFM-CNR S3 National Research Center on nanoStructures and bioSystems atSurfaces, Modena, Italy
G. Cicero - Politecnico di Torino, Italy
A. Catellani – CNR-IMEM, Parma, Italy
C. Cavazzoni – CINECA, Bologna, Italy
ProteinsProteins
• Long linear chains of a-amino acids
• 20 natural a-amino acids
• Have complex 3D structures• Have complex 3D structures
• Perform different functions, e.g.:– Catalyze chemical reactions (enzymes)
– Bind external molecules (immunoproteins)
– Transport electrons between redox partners
Can specifically recognize other molecules/proteins
Exploiting the intrinsic capabilities of proteinsExploiting the intrinsic capabilities of proteins• Proteins can recognize other proteins/molecules
Exploiting the intrinsic capabilities of proteinsExploiting the intrinsic capabilities of proteins• Proteins can recognize other proteins/molecules
Can proteins recognize surfaces?Can we design a protein specific for a given surface?
ora surface that maximally/minimally binds a given protein?
Proteins specific for a surfaceProteins specific for a surfaceExamples of possible applications
•Use proteins to guide theself-assembly of inorganicnanodevices (proteins as asmart glue)
•Use protein to control the growth of inorganic materials, e.g., to obtain nanocrystalswith new shapes
Examples of possible applicationsExamples of possible applicationsprotein specific for a surface
•Use proteins to guide theself-assembly of inorganicnanodevices (proteins as asmart glue)
large gold sphere
•Use protein to control the growth of inorganic materials, e.g., to obtain nanocrystalswith new shapes
small polystirene sphere
SurfaceSurface--recognizing peptides viarecognizing peptides viacombinatorial biotechnologiescombinatorial biotechnologies
• In recent years, combinatorial biochemical methods have beenused to select the best binder to given surfaces, or to findselective binders
iterateiterate
recoverrecoverthe bindersthe binders
amplifyamplify
iterateiterate
S. Brown, PNAS 89, 8651 (1992); M. Sarikaya et al. Nature Mat. 3, 577 (2003)
Proteins can recognize surfacesProteins can recognize surfaces
• Some specific proteins indeed found in this way, e.g.:
GaP InP Ge
•However:
– Still no understanding of the basic physical mechanisms thatgovern the specific interaction No concepts to enable rationaldesign
–Combinatorial selects proteins for a given surface, not viceversa
K.Goede et al., NanoLett 2004C. Tamerler et al., Small 2006 A.Artzy-Schnirman etal., NanoLett 2006
Our (longOur (long--term) goalterm) goal
Understanding the basic physicalprinciples that govern specific protein-principles that govern specific protein-surface interactions by computational
methods
OutlineOutline
• Motivations
• General computational strategies– Multiscale modelling– Ab initio on case studies
• Ab initio molecular dynamics for poly-serine on Au(111)– Method– Simulated system– Results
• Summary
Complexity of the systemsComplexity of the systems
• Very large systems(103-104 atoms)• Protein + Inorganic Surface + Solvent• Multiple length scales
Computational challenge:
• Multiple length scales
• Require statistical sampling and/orsimulation of time evolution
• Multiple time scales
• Involve interactions of different origins• Chemical bonds, Coulomb, dispersion...• Different methods most suitable fordifferent portions
Choice of prototypical systemsChoice of prototypical systems
• The uncertainties on real surfaces (e.g., presence of defects,steps, amorphous oxide layers) and on proteins/peptides (e.g.,unknown structure, too flexible peptides) introduce furthercomplexity in modelling.
• To minimize uncontrolled assumptions, initially focus on well-defined and well-characterized surfaces and polypeptides.
• The knowledge generated from these studies on a range ofprotoypical systems fosters understanding of more complexexamples.
Computational strategiesComputational strategies
Multiscale modellingrelatively general but long term(needs to be developed)(needs to be developed)
Using the most accurate,already available methodsfeasible only for case studies(need enormous computer power)
Computational strategiesComputational strategies
Multiscale modellingrelatively general but long term(needs to be developed)(needs to be developed)
Using the most accurate,already available methodsfeasible only for case studies(need enormous computer power)
SettingSetting--up the multiscale schemeup the multiscale scheme
Parametrization of AA fragment-surface interaction energy: QM
From energy of fragments to free-energy of AAs in water Classical MD
Exper
validation
energy of AAs in water Classical MD
From free-energy of AAs tofree energy of proteins BD
riments
Computational multiscale toolbox
validation
validation
The Prosurf projectThe Prosurf project
Parametrization of AA fragment-surface interaction energy: QM
From energy of fragments to free-energy of AAs in water Classical MD
Exper
validation
“Computationaltoolbox for protein-surface docking”
energy of AAs in water Classical MD
From free-energy of AAs tofree energy of proteins BD
riments
Computational multiscale toolbox
validation
validation
The Prosurf projectThe Prosurf project
Parametrization of AA fragment-surface interaction energy: QM
From energy of fragments to free-energy of AAs in water Classical MD
Exper
validation
“Computationaltoolbox for protein-surface docking”
energy of AAs in water Classical MD
From free-energy of AAs tofree energy of proteins BD
riments
Computational multiscale toolbox
validation
validation
Target surfacesTarget surfaces
• Au(111)– stable in air and water
– important for nanobioelectronics (contacts)
– used in optical detection systems for protein materials– used in optical detection systems for protein materials
– target surface for well-characterized gold binding peptides
• Fully hydroxilated a-Al2O3(0001)– ceramic surfaces used in today's biomaterials
– transparent and thus used in optics application
– well-characterized surface, even when hydrated
Target surfacesTarget surfaces
• Au(111)– stable in air and water
– important for nanobioelectronics (contacts)
– used in optical detection systems for protein materials– used in optical detection systems for protein materials
– target surface for well-characterized gold binding peptides
• Fully hydroxilated a-Al2O3(0001)– ceramic surfaces used in today's biomaterials
– transparent and thus used in optics application
– well-characterized surface, even when hydrated
GolP: a force field for proteinGolP: a force field for protein--surfacesurfaceinteraction in waterinteraction in water
• Other FFs for gold exist (e.g., Zerbetto et al.), not tailored forproteins in water
• Derived from DFT calculations + experimental data + MP2calculations
acetone
mm
acetone
trans 2 butene
diethylsulfide
1-nonene
F. Iori et al. J. Comp. Chem. 30, 1465 (2009); J. Comp. Chem. 29, 1656 (2008)
GolP force field in actionGolP force field in action
• Studying liquid water on Au(111)
• Simulating b-sheets proteins/peptides adhesion on Au(111) (incollaboration with S.Monti, M. Hoefling, K.Gottschalk)
• Interpreting electron transfer measurements for cytochrome Cmutants on gold (see poster by M. Siwko)
• Studying potential of mean force for amino acids adsorption ongold (by M.Hoefling and K. Gottschalk)
• Developing free-energy models for Brownian dynamics (by
D.Kokh, B. Huang, R. Wade)
Computational strategiesComputational strategies
Multiscale modellingrelatively general but long term(needs to be developed)(needs to be developed)
Using the most accurate,already available methodsfeasible only for case studies(need enormous computer power)
OutlineOutline
• Motivations
• General computational strategies– Multiscale modelling– Ab initio on case studies
• Ab initio molecular dynamics for poly-serine on Au(111)– Method– Simulated system– Results
• Summary
Ab Initio Molecular Dynamics (AIMD)Ab Initio Molecular Dynamics (AIMD)
• Nuclei: classical particles moving on a energysurface determined by electron density
• Electrons: quantum particles that follow the slowernuclear motion
Car-Parrinello AIMD: a single equation of motion with fictitious electronicdegrees of freedom, to keep electrons in the ground state during thenuclear dynamics
Ab Initio Molecular Dynamics (AIMD)Ab Initio Molecular Dynamics (AIMD)
• Nuclei: classical particles moving on a energysurface determined by electron density
No empirical parameters
• Electrons: quantum particles that follow the slowernuclear motion
Car-Parrinello AIMD: a single equation of motion with fictitious electronicdegrees of freedom, to keep electrons in the ground state during thenuclear dynamics
Simulated systemSimulated system
Water layer
Protein: poly-Serine b-sheetProtein: poly-Serine b-sheet
Surface: Au(111)4 layers; 23x7 supercell; unreconstructed
Interstitial water layer
3D periodic boundary conditions
In total: # atoms = 587; # electrons = 2552
Simulated systemSimulated system
Why polySerine?• Serine (and other hydroxilated AA) appearsin experimental gold binding peptides (GBP)• Redundancy improves averaging
Why b-sheet?• Maximize contact with the surface• Allow easy implementation of periodicboundary condition
Details of the calculationsDetails of the calculations
ab initio Car-Parrinello molecular dynamics
xc: PBE, PW basis set (Ecut=25 Ry ), ultrasoft pseudopotential
CODE = cp.x (quantum espresso package)
Running machine = MareNostrumMareNostrum (Barcelona, Spain) ): DEISA project Running machine = MareNostrumMareNostrum (Barcelona, Spain) ): DEISA project
Total CPU time = 350 000 CPU hours.
Bloechl-Parrinello thermostat for electrons
20 ps of simulated evolution at T= 400K; time step dt=0.17 fs
OutlineOutline
• Motivations
• General computational strategies– Multiscale modelling– Ab initio on case studies
• Ab initio molecular dynamics for poly-serine on Au(111)– Method– Simulated system– Results
• Summary
Main resultsMain results
• Polyser and water has a weak but notnegligible interaction with on Au(111).
• Protein and water recognize the atomiccorrugation of the gold surface.
• Protein+hydration layer creates an object withwell-defined geometric properties
Main resultsMain results
• Polyser and water has a weak but notnegligible interaction with on Au(111).
• Protein and water recognize the atomiccorrugation of the gold surface.
• Protein+hydration layer creates an object withwell-defined geometric properties
Small charge transfer to AuSmall charge transfer to Au
e- population per Au atom(superpos. of all Au at all snapshots)
e- popul. per Ser side chain O(superpos. of all O at all snapshots)
öw
din
po
p.
öw
din
po
p.
External Au layers haveexcess negative charge
Lö
wd
inp
op
.
Lö
wd
inp
op
.
Small charge transfer to AuSmall charge transfer to Au
e- population per Au atom(superpos. of all Au at all snapshots)
e- popul. per water O atom(superpos. of all O at all snapshots)
öw
din
po
p.
öw
din
po
p.
External Au layers haveexcess negative charge
Lö
wd
inp
op
.
Lö
wd
inp
op
.
Orbital mixing?Orbital mixing?
Non-interacting
Projected density of states for serineP
DO
S(a
.u.)
Interacting
red Oblack tot
occupied virtualenergy (eV)
PD
OS
(a.u
.)
Simplifying the system: a singleSimplifying the system: a singlemethanol molecule on Au(111)methanol molecule on Au(111)
occupied virtual
side chain model
Eint= -12 kJ/mol
Simplifying the system: a singleSimplifying the system: a singlemethanol molecule on Au(111)methanol molecule on Au(111)
occupied virtual
side chain model
Eint= -12 kJ/mol
Simplifying the system: a singleSimplifying the system: a singlemethanol molecule on Au(111)methanol molecule on Au(111)
occupied virtual
side chain model
Eint= -12 kJ/mol
• Small but detectable O-to-Au charge transfer
• Small but detectable perturbation of PDOS.
• However, interaction is weak, as in experiments*
(small interaction energy; protein and water slide onthe surface)
*water: Kay et al., JCP 1989serine: Peelle et al. Langmuir 2005
Main resultsMain results
• Polyser and water has a weak but notnegligible interaction with on Au(111).
• Protein and water recognize the atomiccorrugation of the gold surface.
• Protein+hydration layer creates an object withwell-defined geometric properties
How to describe relative AuHow to describe relative Au--OOpositions?positions?
1
( )AuO O AuAu OAu O
gN N
In-plane pair distribution function:
Au OAu O
Meaning: where atoms O sit w.r.t. an atom Au, averaged onall the Au atom and projected on the surface plane.
gAuO averages over the exposed Au and O atoms -> takesadvantage of the redundancy in our system
color scale: gAuOser(x,y):where Oser sits w.r.t. Au atom,
averaged on Au atoms andprojected on the surface
max
The protein recognizes the atomicThe protein recognizes the atomicstructure of the gold surfacestructure of the gold surface
Oser prefers one adsorption site over the others
0
The same for water...The same for water...
max
Owat prefer one adsorption site over the others
0
• Polyser and water “are aware” of the atomicstructure of the surface.
• Oser localize in bridge sites; Owat localize ontop sites
O watO ser
Effects of missing longEffects of missing long--rangerangedispersion?dispersion?
• GGA xc functionals do not correctly describe long-range dispersion
• No direct estimate of effects of long-range dispersion• No direct estimate of effects of long-range dispersionin our system
• However: for related systems (rare gases on metals):– on-top is the preferred adsorption site
– for tested cases, adsorption site preferences are notmodified by adding dispersion
Effects of dispersion on adsorptionEffects of dispersion on adsorptiongeometries of rare gases on metalsgeometries of rare gases on metals
•Calc
•Exp
top and bridge
Bruch et al.Rev. Mod. Phys. 2007
Main resultsMain results
• Polyser and water has a weak but notnegligible interaction with on Au(111).
• Protein and water recognize the atomiccorrugation of the gold surface.
• Protein+hydration layer creates an object withwell-defined geometric properties
Profiles of atomic densitiesProfiles of atomic densities
Profiles of atomic densitiesProfiles of atomic densities
The solvent andgold exposedprotein surfacesprotein surfacesinclude a solventlayer
Hydration layersHydration layers
•Hydration layer exchanges with the solution•Hydration water is structured
The surfaces of the hydrated proteinThe surfaces of the hydrated protein
density map of WAT Oxygen in the hydration layers
max
0
The surfaces of the hydrated proteinThe surfaces of the hydrated protein
SER Oxygens not distinguishable by Water Oxygens
max
0
• Protein and water recognize the surface
• Hydration water is (dynamically) bound to theprotein in well-defined positions (similar toantifreeze proteins )antifreeze proteins Nuts et al. JACS 2008)
• In spatial matching between the protein andthe surface, hydration water may play a role
SummarySummary
• Specific protein-inorganic surface interactions arepromising for several nanobiotechnology applications
• Basic principles regulating such interactions are notunderstood; High level calculations can shed light onthemthem
• Main results of ab initio simulation:1. Polyser and water do not chemisorb on Au(111). However, a
weak interaction is indeed present.
2. Protein and water recognize the atomic corrugation of the goldsurface.
3. Protein+hydration layer creates an object with well-definedgeometric properties
AcknowledgmentsAcknowledgments
Exp
validation
G. Schreiber
E. Molinari, F. Iori, M. Siwko, F. DeRienzo, A. Calzolari, A. Catellani, R. DiFelice@MODENA: QM
K. Gottschalk, M. Hoefling @LMU:Classical MD
Computational multiscale toolbox
validation
validation
G. SchreiberO. CohaviD. Reichman
A. VaskevitchI. RubinsteinA. B. TeslerO. Kedem
T. Karakouz
@WEIZMANN
R. Wade, B.Huang, D.Kokh,P. Winn@EML: BD
Classical MD
SIXTH FRAMEWORKPROGRAMME
AcknowledgmentsAcknowledgments
• Computational time and assistance with calculations
• Funding: SIXTH FRAMEWORKPROGRAMME
AcknowledgmentsAcknowledgments
• Computational time and assistance with calculations
• Funding: SIXTH FRAMEWORKPROGRAMME
Thank you for your attention!
The same for water...The same for water...
max
Owat prefer one adsorption site over the others
0
on-top site
hollow site
bridge site
AuAu--O distribution functions:O distribution functions:Classical system setClassical system set--up vs ab initio runup vs ab initio run
Duringsystemset up
Ab initiorun
SER Oxygen WAT Oxygen
Effects of missing van der Waals onEffects of missing van der Waals onadsorption geometryadsorption geometry
•Calc
•Exp
top and bridge
Bruch et al.Rev. Mod. Phys. 2007
The trajectoryThe trajectory
Only water molecules between the protein and the surface are shown
A single water molecule on Au(111)A single water molecule on Au(111)
HOMO at Goccupied virtual
Dr