NUMERICAL EXPERIMENT IN MOLECULAR BIOPHYSICS OF PROTEINS AND MEMBRANES:

TASKS AND SOLUTIONS FOR PROTEOMICS

Laboratory ofBiomolecularModeling

Roman EFREMOV

efremov@nmr.rumodel.nmr.ru

Russian Academy of SciencesM.M. Shemyakin & Yu.A. Ovchinnikov Institute of Bioorganic Chemistry

www.ibch.ru

The Team:Dmitry E. NOLDE Ph.D.Anton O. CHUGUNOV Ph.D.Anton A. POLYANSKY Ph.D.Pavel E. VOLYNSKY Ph.D.Darya V. PYRKOVA Ph.D. studentNikolai A. KRYLOV Ph.D. studentNatalya K. TARASOVA student

Grant sponsors:Russian Academy of Sciences;Russian Foundation for Basic Research.

Computations:Joint Supercomputer Center, Russian Academy of Sciences;i-SCALARE: INTEL-MIPT Laboratory of supercomputer systems.

Thanks:

Department of Structural Biology

Laboratoryof Biomolecular NMR

Head: Prof. Alexander Arseniev

Laboratoryof Optical Microscopy &

Spectroscopy of Biomolecules

Head: Dr. Alexei Feofanov

Laboratoryof Biomolecular Modeling

Head: Prof. Roman Efremov

NMR spectrometers with CRYOPROBE:UNITY 600 (Varian)Aliance III 800, II 700, II 600 (Bruker)

TIRF-confocal microscope (Carl Zeiss)4-pi microscope (Leica), etc.

Access to high-performance computational clusters (up to 140 TFlops)

LABORATORY OF BIOMOLECULAR MODELINGCURRENT RESEARCH PROJECTS:

OUTLINE

• Why membranes? Why computational experiment?

• Computational “kitchen”: models, methods, realization;

• Examples of problems under study:1. “mosaic” nature of biomembrane surface and its role in binding of

amphiphilic peptides to cell membranes;2. peptide-peptide interactions in membranes – accent on the peptide

surface;

• Problems, conclusions, perspectives;

Biomembranes as perspectivepharmacological targets

Up to 70% of currently marketed drugs act either on membrane proteins or on membrane itself

Examples of potential targets:

- G-protein coupled receptors (GPCRs);

- Transmembrane ion channels and transporters;

- Integral MPs involved in oligomerization upon their functioning (receptor tyrosine kinases, apoptotic proteins, etc.);

- Lipid bilayer of biomembranes (direct and indirect modificationsof its properties can be vitally important for cell)

MODERN APPROACHES IN MODELING OF CELL MEMBRANES

Implicit (“hydrophobic slab”) models;

All-atom hydrated lipid bilayers and micelles;

Simplified continuous models (no microscopic details);

Simplified corpuscular - coarse grain (CG) -models.

Potential Energy Function:Intraprotein: (Némethy et al, 1983)

Protein/Solvent : ∑ ×=i

iisolv ASPE σ

≥⋅−⋅+<⋅−⋅−

= λ−−

λ−

0/)(

0/)(

,)(5.0,)(5.0

)(0

0

zzeASPASPASPzzeASPASPASPzASP zzwat

imem

iwat

i

zzwati

memi

memi

i(Efremov et al., 2000,

Nolde et al., 2000)

Implicit membrane models: How do they look like?

Method:Monte Carlo simulation with energy minimization

Full-atom membrane models

DOPS bilayer SDS micelle

Parameters of the systems:60-500 lipid (detergent) molecules

103 – 104 waters

Size of the box: 10·10·10 nm3.

3D periodic boundary conditions

Molecular dynamics in NPT ensemble.

Polyansky A. et al. (2005) J. Phys. Chem. B 109:15052

Models of lipid bilayers mimicking different types of biomembranes

Membrane “ERYTH” (288 lipids) Membrane “GRAMN” (288 lipids)The mimic of the human erythrocyte membrane The mimic of the cytoplasmic membrane of gram

negative bacteria (E. coli)

POPC – 40%, POPE – 40%, CHOL – 20% POPE – 70%, POPG – 30%

Coarse-grain (CG) modelsMain characteristics:

(Lopez et al, 2006)

Interactions:

1. LJ-potential

2. Electrostatic

3. Harmonic potentials for bonds and valence angles between CG-particles

Some parameters of MD protocol:

Cutoff/smoothing functions (~12Å )

Integration time ~10-50 fs

Gain in the computational time: 3-4 orders of magnitude!

Self-assembling of DPPC bilayer

128 DPPC / 2000 x4 H2O

Т = 323 К

Trajectory: 5 x 4 ns

CPU: AMD Athlon 64 x2 3.8 ГГц

Performance: 16 x 4 ns/hour (on 1 core)

Gromacs 3.31

L-J: 12 Å (switch) / El. 12 Å (shift)

CG-model: water, choline, phosphate, glycerol,aliphatics

Resulting area per lipid molecule - 62 Å

CG-models: equilibrated lipid bilayer – for 1 day!

<AL>=62.1 Å2

<Dpp>=39.1 Å2

<Scd >DOPE=0.140±0.059<Scd >DOPG=0.144±0.064

averaging: last 0.25 ns

Polyansky et al. J.Phys. Chem B, 2005

180PE/76PG/2500W/76Na+

Molecular hydrophobicity potential (MHP):

logP=log(Co/Cw)

logP1=f1+f2+f3logP2=f1+f2+f3+f4… … … … … … … … … … … …

Efremov R. et al., (1992). J. Prot. Chem. 11: 699Efremov R. et al., (2007). Curr. Med. Chem. 14: 393

МНР-calculations were done with the programPLATINUM http://model.nmr.ru/platinum

Pyrkov T. et al. (2009) Bioinformatics, 23: 2947

Systems / computational details

Mapping of hydrophobic properties of surfaces of helical peptides.

Hydrophobic properties of surfaces of hydrated lipid bilayers in atomistic

presentation

> 0 (Hydrophobic regions)< 0 (Hydrophilic regions)

Distribution of molecular hydrophobicity potential on themembrane surface Parameters of the systems:

60-500 lipids;

103 – 104 waters;

Cell dimensions: 10·10·10 nm3.

What properties of a molecular surface might be of interest?

• Distribution of hydrophobic/hydrophilic regions;

• Distribution of electrostatic potential;

• Landscape (“ridges”, “canyons”, and “plains”);

• Characteristics of conformational flexibility of a

molecule (e.g., obtained in molecular dynamics

simulations);

• Location of particular groups of atoms and residues, active sites, interfaces of intermolecular interactions, and so on;

Upon employment of common format of surface mapping, differential, averaged, etc. maps become very important.

Often, necessary information is obtained in the result of сomparative analysis of different types of surface maps (e.g., hydrophobicity/landscape, and so on)

Northern coast of the Black Sea, Таman, and Eastern Europe. Меrcator’s, “Atlas”, 1584. Re-edition of the map (dated 1513) based on ideas by Claudius Ptolemaeus.

Example:

“mosaic” nature of biomembrane surface and its role in

binding of amphiphilic peptides to cell membranes

En

gelm

an, N

atu

re, 4

38, 2

00

5

large complexes of proteins

lipid rafts

heterogeneous distribution of lipids of different chemical structure

Complicated character of interactions between the membrane and various peripheral agents !

Membranes are more mosaic than fluid

Mosaic nature of lipid bilayers

A «simple» system: 2-component lipid bilayer

Dioleoyl-phosphatidylcholine (DOPC)

Dipalmitoyl-phosphatidylcholine (DPPC)

DОPC DPPCDOPC 288 -DOPC90 258 30DOPC80 230 58DOPC70 200 88DOPC50 144 144DOPC30 86 202DOPC20 58 230DOPC10 30 258DPPC - 288

Lateral heterogeneity in lipid mixtures

Structural parameters of the lipid bilayers under study

AL (Å2) * DPP (Å) **

расчет эксп. расчет эксп.ДОФХ 71.7 +/- 0.1 72.5 35.8 +/- 0.1 36.9

ДОФХ90 66.4+/- 0.1 37.9+/- 0.1

ДОФХ80 66.1+/- 0.1 37.8+/- 0.1

ДОФХ70 65.6 +/- 0.1 37.8 +/- 0.1

ДОФХ50 64.2 +/- 0.1 37.7 +/- 0.1

ДОФХ30 63.9 +/- 0.1 37.2 +/- 0.1

ДОФХ20 62.2+/- 0.1 37.9+/- 0.1

ДОФХ10 61.9+/- 0.1 37.6 +/- 0.1

ДПФХ 61.4 +/- 0.1 63.3 37.6 +/- 0.1 38.3

* AL - area per lipid molecule;** DPP - bilayer thickness.

• DPP values in mixed bilayers are close to those in “pure” DPPC;• AL in mixed bilayers decreases with increasing of DPPC concentration.

• Small amounts of saturated lipid induce large changes in structural parameters of bilayer composed of unsaturated lipids.

linear approximation

MD data

exp. data

exp. data

Calc. Exp. Calc. Exp.

DOPC

DOPC90

DOPC80

DOPC70

DOPC50

DOPC30

DOPC20

DOPC10

DPPC

Mosaic nature of lipid bilayers

Pyrkova D. et al. (2011) Soft Matter, 7: 2569

Hydrophobic / hydrophilic organization of surfaces in model lipid blayers

DOPC

DOPS

hydrophilic hydrophobic

DOPC – dioleoyl-phosphatidylcholine (zwitterionic), DOPS – dioleoyl-phosphatidylserine (anionic).

Mosaic nature of lipid bilayers

DOPC DPPC

Hydrophobic Hydrophilic Strongly hydrophilic

in clusters

outside clusters

Solvation, H-bonding with water

Tilt of lipid heads

Lipid ordering↑

↓

↑

↑ Hydrophilicity

Top-view of the water-membrane interface

Some conclusions about lateral organization of the binary DPPC/DOPC lipid membrane

(in a liquid crystalline state)

Mosaic nature of lipid bilayers

Hydrophobic organization of the surface

1

jkN

Rk

kMHP f e−

=

=∑

fk- constant of atomic hydrophobicity, Rjk-distance between atomk and point j ( Å), N – number of atoms

Mosaic nature of the membrane surface

Mosaic nature of lipid bilayers

WHAT KIND OF STRUCTURAL / DYNAMIC /

FUNCTIONAL DATA CAN BE DERIVED FROM

COMPUTATIONAL EXPERIMENTS?

MEMBRANES AND PEPTIDE-MEMBRANE COMPLEXES:

MD simulations of pAntp in all-atom lipid bilayers

In DOPS:-Deep penetration;-Helix destabilization;-No specific interactions;

-Interfacial location;-Helical conformation;-Specific interactions (“traps”);

In DOPC:-Always helical;-Interfacial location;-Few specific contacts;

Depth of insertion

Burial ofthe centerof mass

H-bonding

Tilt angle

Peptide-bilayerinteractions

MOLECULAR DYNAMICS DATA ON PEPTIDE-MEMBRANE BINDING

Bilayer’s properties affect peptide’s insertion

DMPC – “fluid” bilayer DPPC – “rigid” bilayer

Peptide-induced thinning of bilayer

Upper leaflet

Lower leaflet

Specific peptide-lipids contacts (“traps”)

Flexibility and geometry of lipid polar heads

- The peptide’s “traps” induce local destabilization of lipid bilayer:

- Immobilization of nearest lipids

- Changes in orientation of lipids’ headgroups;

- The effects are much prominent in DOPS than in DOPC bilayers;

Polyansky A.A. et al. (2009) J. Phys. Chem B, 113:1107Polyansky A.A. et al. (2009) J. Phys. Chem B, 113:1120

Peptide-lipid interactions may strongly depend on the local distribution of hydrophobic/hydrophilic properties of a bilayer in

the contact area

( MHP – molecular hydrophobicity potential )

“Mosaic” nature of the membrane surface

Polyansky A.A. et al. (2009) J. Phys. Chem B, 113:1107Polyansky A.A. et al. (2009) J. Phys. Chem B, 113:1120

Antimicrobial peptides (AMPs): Membrane-assisted action

A. Polyansky et al , J. Phys. Chem. Lett. (2010)

Martini v2.1 force-field, MD (1 μs),256 lipids (DOPE:DOPG 7:3 mixture)CG-water box+12 AMP (Ltc2a)

“Pure” PE/PG (top view) hydrophob./hydrophil. surface

+ 12 AMPs (1 μs MD) hydrophobic ”defects”

Example:

Interactions of transmembrane helical peptides in lipid

environment:

Prediction of 3D structure of helix-helix dimers taking into account

hydrophobic properties and landscape of their surfaces

PROBING PROTEIN-PROTEIN INTERACTIONS IN MEMBRANE.

WHY THIS IS IMPORTANT ?

• Most of membrane proteins contain in their TM-domains several polypeptide regions.

• Many membrane proteins are active only in oligomeric states:Examples: protein kinases (e.g., Erbb-, insulin-, and EpoR-receptors), mitochondrial apoptotic proteins, etc.

• Molecular mechanisms of such interactions are still poorly understood (even for two helices).

Normal cellligand

Cancer cell

ligand

Artificial“peptide-interceptor”

dimerization active dimer active dimer inactive dimer monomer

THE CONCEPT OF “PEPTIDES-INTERCEPTORS”

General scheme of the prediction approach

1 –building ideal helix from

TM-sequence

2 – calculation of the helical

surface (MHP + landscape)

3 – 2D interpolation of the

surface

4 – slicing of the surface to a

number of fragments (for each

tilt angle) and their pairwise

comparison (by scoring

function)

5 – reconstruction of 3D

structure of a dimer

PREDDIMER is capable of prediction reasonably correct structures of TM dimers

BNIP3 EPHA1

NMR and predicted (PREDDIMER) structures

Helix-helix association in membrane

Helix-helix interface

PERSPECTIVES:

• Development of hybrid models combining advantages of different approximations used to describe biomembranes and protein-membrane interactions (all-atom, coarse-grained, implicit models);

• Multi-level modeling of membranes with complex composition(3 and more components) in order to understand molecular basis of forming rafts and domains in cell membranes;

• Analysis of relationships between properties of mixed bilayers and behavior of membrane and membrane-active peptides and proteins;

• Application of structural, dynamic, and hydrophobic heterogeneity of native membranes to rational design of principally new membrane materials with predefined properties and to design efficient membrane-active compounds (including drugs).

CONCLUSIONS:

Modeling is capable of providing reasonable guesses about important pharmacological targets - membranes and their complexes with peptides and proteins;

The most reliable results are obtained viacombined using of experimental and several self-consistent in silico methods;

Computer simulations represent a powerful tool in design of new molecules targeting biological membranes and membrane proteins.

Laboratory of Biomolecular Modeling

Shemyakin-Ovchinnikov Institute of Bioorganic ChemistryRussian Academy of Scienceswww.ibch.ru model.nmr.ru