The Computational Microscope Images Biomolecular Machines and Nanodevices - Klaus Schulten

The Computational Microscope Images
Biomolecular Machines and Nanodevices
shadow is Accuracy • Speed-up • Unprecedented Scale
crystal structure

protein folding:
villin headpiece
3 months on 329 CPUs
Klaus Schulten
Department of Physics and
Theoretical and Computational Biophysics Group
University of Illinois at Urbana-Champaign
Lecture at CPLC Summer School, Urbana, July, 2012
Wednesday, August 1, 2012

Our Microscope is Made of...

Chemistry NAMD Software 100

ns/day
10

Virus 1

128
256
512
1024
2048
4096
8192
16384
32768
51,000 registered users cores
Physics

Math
..and Supercomputers
(repeat one billion times = microsecond)


3
Outside researchers choose NAMD and succeed
Corringer, et al., Nature, 2011 2100 external citations since 2007 Voth, et al., PNAS, 2010

180K-atom 30 ns study of anesthetic binding to bacterial
ligand-gated ion channel provided “complementary
interpretations…that could not have been deduced from
the static structure alone.”

Bound Propofol Anesthetic
500K-atom 500 ns investigation of effect of
actin depolymerization factor/cofilin on
mechanical properties and conformational
dynamics of actin filament.

Bare actin Cofilactin
Recent NAMD Simulations in Nature
• M. Koeksal, et al., Taxadiene synthase structure and evolution of modular architecture in terpene biosynthesis. (2011)
• C.-C. Su, et al., Crystal structure of the CusBA heavy-metal efflux complex of Escherichia coli. (2011)
• D. Slade, et al., The structure and catalytic mechanism of a poly(ADP-ribose) glycohydrolase. (2011)
• F. Rose, et al., Mechanism of copper(II)-induced misfolding of Parkinson’s disease protein. (2011)
• L. G. Cuello, et al., Structural basis for the coupling between activation and inactivation gates in K(+) channels. (2010)
• S. Dang, et al.,, Structure of a fucose transporter in an outward-open conformation. (2010)
• F. Long, et al., Crystal structures of the CusA efflux pump suggest methionine-mediated metal transport. (2010)
• R. H. P. Law, et al., The structural basis for membrane binding and pore formation by lymphocyte perforin. (2010)
• P. Dalhaimer and T. D. Pollard, Molecular Dynamics Simulations of Arp2/3 Complex Activation. (2010)
• J. A. Tainer, et al., Recognition of the Ring-Opened State of Proliferating Cell Nuclear Antigen by Replication Factor C Promotes
Eukaryotic Clamp-Loading. (2010)
• D. Krepkiy, et al.,, Structure and hydration of membranes embedded with voltage-sensing domains. (2009)
• N. Yeung, et al.,, Rational design of a structural and functional nitric oxide reductase. (2009)
• Z. Xia, et al., Recognition Mechanism of siRNA by Viral p19 Suppressor ofBioinformatics
BTRC for Macromolecular Modeling and RNA Silencing: A Molecular Dynamics Study. (2009) UIUC
Beckman Institute,
http://www.ks.uiuc.edu/


Successful folding of largest fast-folding protein:
λ-repressor / Collaboration Gruebele-Schulten
System:
• 5-helix bundle protein
• Largest fast-folding protein
• Experimental folding time < 20 µs
Recent progress:
• Successful folding in < 5 µs using enhanced sampling
• 100-µs folding trajectory using both NAMD and Anton

0 15
RMSD (Å)

Liu et al, J. Chem. Phys. Lett. (2012)
Alpha helix Pi-helix 3-10 helix
Turn None (coil)
BTRC for Macromolecular Modeling and Bioinformatics Beckman Institute, UIUC
Lamda repressor folding, 24.1 µs. Per-residue secondary structure and RMSD.


Computational Microscope at Work:
Inspecting the Mechanical Strength of a Blood Clot

A Blood Clot
20ns SMD Simulation of fibrinogen, 1.06 million atoms, 1.2 ns/day
Red blood cells within a
with pencil decomposition, 15 days on PSC XT3 Cray (1024
network of fibrin fibers,
processors) composed of polyme-
B. Lim, E. Lee, M. Sotomayor, and K. Schulten. Molecular basis of fibrin clot rized fibrinogen mole-
elasticity. Structure, 16:449-459, 2008. cules.
12


(d)
Design of Tyrosine Kinase Sensor
•Peptides are attached to a gold surface. The end of oligopeptide contains rhodamine.
•Peptides are exposed to ATP and appropriate kinase.
•Electric field is applied; change in conformation depends on phosphorylation state.
•Distance from rhodamine determines magnitude of fluorescence signal.

initial sequence: {1 GLU} {2 GLY} {3 ILE} {4 TYR} {5 GLY} {6 VAL} {7 LEU} {8 PHE} {9 LYS} {10 LYS} {11 LYS} {12 CYS}
Rhodamine
(R6g) Tyrosine residue interacts with kinases Experiment: Logan Liu and
al image of sensor: A
(a) by phosphorylation
(b) (c)
Fluorescence image
(d) Chen, Nano Lab, U. Illinois
Yi
yer of 5 nm titanium showing peptide probes
were successfully

o activate the sensor (b) immobilized on the
(a) nanochip surface
(e)
than smooth surface. 300 nm

ensor. Peptide probes
kinase

sensor. Peptides are pattern are created by a by experiments and simulations. A thin layerimage nmsensor: A
Figure 2: Proposed kinase sensor studied
5×5 square array SEPERISE method (see text).
(a) Optical
of 5
of
titanium
representation. Blue of 80 nm gold is depositedwithin theofsurface, so they structures tothan smooth surface.
followed by a layer
surface. The nanostructures trap incident light
on top (a)
the nanocone
look darker
(b)
activate the sensor
Fluorescence image
showing peptide probes
(c) (d)

ectively. oligopeptide with 12Nanoscopic structures of the activated areas of the kinase sensor. Peptide probes
Au surface, (b) MISSING FIGURE. (c) AA,
were successfully
immobilized on the
are immobilized on the gold covered surface of nanocones. (c) Atomic model of kinase sensor. Peptides are
nanochip surface

bound via sulﬁde bond
colored in green, metallic surface in yellow. A single peptide is highlighted in licorice representation. Blue
300 nm

(c) (d)
and red arrows point to the rhodamine cap and phosphorylated tyrosine residue, respectively.
Figure 2: Proposed kinase sensor studied by experiments and simulations. (a) Optical image of sensor: A
5×5 square array pattern are created by a SEPERISE method (see text). A thin layer of 5 nm titanium
followed by a layer of 80 nm gold is deposited on top of the nanocone structures to activate the sensor
kin. surface. The nanostructures trap incident light within the surface, so they look darker than smooth surface.
(b) MISSING FIGURE. (c) Nanoscopic structures of the activated areas of the kinase sensor. Peptide probes
are immobilized on the gold covered surface of nanocones. (c) Atomic model of kinase sensor. Peptides are
V V colored in green, metallic surface in yellow. A single peptide is highlighted in licorice representation. Blue
and red arrows point to the rhodamine cap and phosphorylated tyrosine residue, respectively.
Spectrometer


MD Simulations Revealed Problems in Design
initial sequence: {1 GLU} {2 GLY} {3 ILE} {4 TYR} {5 GLY} {6 VAL} {7 LEU} {8 PHE} {9 LYS} {10 LYS} {11 LYS} {12 CYS}

Initial sequence did not work due to surface adhesion and rhodamine aggregation

•Improved sequence.
Avoid residues that
strongly bind to gold
surface.

•Rhodamines
molecules tend to
aggregate.
Aggregation affects
peptide bending.

•MD systems include gold surface, water,
ions and 5x5 peptide grid, ~ 100,000 atoms.
•Different sequences tested under positive
and negative volt biases.


Current Design Improved
New sequence: rhodamin removed and measurement replaced by Raman spectroscopy

Rhodamine removed,
40
Fluorescence signal discarded
35
MD Simulations 30

dist Tyr-Gold / A
show that +60V
25
phosphorylated -60V
tyrosine bends the 20 0V
peptide depending on 15
voltage polarity
10

5

0
0 1 2 3 4 5 6

time / ns

800

+1.2V
Experiments show 600 -1.2V

intensity / AU
that peptide bends 0V
towards the surface at
400
positive voltages,
increasing the
intensity of Raman 200
Phosphorylated tyrosine shifts.
0

Peptide bending is now measured with Raman Spectroscopy. 700 1100 1500
The detection relies on careful comparison of peak points
Raman shift / cm-1

Unphosphorylated Phosphorylated
(a) 4
4
rho-EGIYGVLFKKKC-gold (i)
(a) rho-EGIYGVLFKKKC-gold (i)

dist Y-gold / nm
EGIYGVLFKKKC-AU EGI(pY)GVLFKKKC-AU
3

Simulation

dist Y-gold / nm
3
2
40 40 2
35 35 1
1
30 30
dist Tyr-Gold / A

0

dist Tyr-Gold / A
25 0 1 2 3 4 +60V
5 6
25
1 2 3
time / 4
ns -60V
5 6
20 20 time / ns 0V
15 15 4
+60V 4 (b) rho-EGI(pY)GVLFKKKC-gold (ii)
10 -60V 10 (b) rho-EGI(pY)GVLFKKKC-gold (ii)

dist Y-gold / nm
3

dist Y-gold / nm
5 0V 5 3 (i)

2 (i)
0 0
0 1 2 3 4 5 6 0 12 2 3 4 5 6
1
time / ns time / ns (ii)
1 (ii)
0
1 2 3 4 5 6
0
1 2 time
3 / ns
4 5 6
time / ns
Figure 4: Peptide phosphorylation revealed by molecular dynamics simulations. Panels
Figure 4: Peptide phosphorylation revealed by surface for dynamicsvoltage polarities. Er
the distance from the tyrosine residue to the gold molecular diﬀerent simulations. Panels
thestandard deviation. (a) Non-phosphorylated peptide sensor under +60 V (blue), -60 V (
± distance from the tyrosine residue to the gold surface for diﬀerent voltage polarities. Er
± standard deviation. (a) Non-phosphorylated peptide sensor under +60 (red) and -60 V (
biases. (b) Phosphorylated peptide sensor under +60 V (blue), -60 V V (blue), 0 volta
biases. (b) Phosphorylated peptide peptides after+60 Vfor +60 V (i) and -60 V 0 volta
(i) and (ii) show snapshots of two sensor under 5 ns (blue), -60 V (red) and (ii) vo
(i) and (ii) show snapshots gray tubes. Rhodamine and for +60 V (i) and -60 V (ii) vo
peptides are represented as of two peptides after 5 ns tyrosine residues are shown in bl
peptides are represented as gray tubes. Rhodamine and tyrosine residues are shown in bl
respectively.
respectively.


Unphosphorylated Phosphorylated
(a) 4
4
rho-EGIYGVLFKKKC-gold (i)
(a) rho-EGIYGVLFKKKC-gold (i)

dist Y-gold / nm
EGIYGVLFKKKC-AU EGI(pY)GVLFKKKC-AU
3

Simulation

dist Y-gold / nm
3
2
40 40 2
35 35 1
1
30 30
dist Tyr-Gold / A

0

dist Tyr-Gold / A
25 0 1 2 3 4 +60V
5 6
25
1 2 3
time / 4
ns -60V
5 6
20 20 time / ns 0V
15 15 4
+60V 4 (b) rho-EGI(pY)GVLFKKKC-gold (ii)
10 -60V 10 (b) rho-EGI(pY)GVLFKKKC-gold (ii)

dist Y-gold / nm
3

dist Y-gold / nm
5 0V 5 3 (i)

2 (i)
0 0
0 1 2 3 4 5 6 0 12 2 3 4 5 6
1
time / ns time / ns (ii)
1 (ii)
0
1 2 3 4 5 6

Experiment
0
1 2 time
3 / ns
4 5 6
time / ns
800 Figure 4: Peptide phosphorylation revealed by molecular dynamics simulations. Panels
800

+1.2V Figure 4: Peptide phosphorylation revealed by surface for dynamicsvoltage polarities. Er
the distance from the tyrosine residue to the gold molecular diﬀerent simulations. Panels
600 -1.2V thestandard deviation. (a)+1.2V
± distance from the tyrosine residue to the gold surface for diﬀerent +60 V (blue), -60 V (
Non-phosphorylated peptide sensor under voltage polarities. Er
± standard deviation. (a) -1.2V
600
biases. (b) Phosphorylated peptide sensor under +60 sensor under +60 (red) and -60 V (
Non-phosphorylated peptide V (blue), -60 V V (blue), 0 volta
intensity / AU

intensity / AU

0V
biases. (b) Phosphorylated0V two peptides after+60 Vfor +60 V (i) and -60 V 0 volta
(i) and (ii) show snapshots peptide sensor under 5 ns (blue), -60 V (red) and (ii) vo
of
400 (i) and (ii) show snapshots gray tubes. Rhodamine and for +60 V (i) and -60 V (ii) vo
peptides are represented as of two peptides after 5 ns tyrosine residues are shown in bl
400
peptides are represented as gray tubes. Rhodamine and tyrosine residues are shown in bl
respectively.
200 respectively. 200

0 0

700 1100 1500 700 1100 1500

Raman shift / cm-1 Raman shift / cm-1

dist
Simulation-Optimized 1

Sequence 0
1 2 3 4 5
time / ns

4 4
(a) EGIYGVLAAAAC-gold
(b)
dist Y-gold / nm

dist Y-gold / nm
3 3

2 2 EGI(pY)GVLAAAAC-gold

1 1

0 0
1 2 3 4 5 1 2 3 4 5
time / ns time / ns
A new sequence is proposed from molecular dynamics simulations. There are dynamics simulations
Figure 5: Optimized sequence. A new sequence is proposed from molecular no
rhodamine caps, avoiding aggregation. Unnecessary lysineUnnecessary lysine residues (a), whileby alanine residu
4
is no rhodamine caps; avoiding aggregation.
residueselectric field changed the phosphoryla
non-phosphorylated peptide sensor is still insensitive to an
are changed to
alanine residues. The resulting non-phosphorylated peptide sensor is still distinctive raman signat
(b) is responsive to an external electric field (b); therefore it should produce
insensitive to an electric field, while the and 0 voltage Error bars represent is standard deviation. to an red an green r
opposite voltage polarities.
phosphorylated one ± responsive Color blue,
t Y-gold / nm

+60 V, -60 V biases.
3
external electric field; therefore the sensor should produce distinctive Raman
signatures for opposite voltage polarities. Error bars represent standard deviation.
Colors blue, red, and green represent +60 V, -60 V and 0 voltage biases.
2 EGI(pY)GVLAAAAC-gold
Wednesday, August 1, 2012 16

Imaging Cellular Machines


14

Larger machines
enable larger
simulations



Petascale Gateway to the Nation’s Top Computers

Petascale computers (e.g., Blue Waters) will create
~100x larger datasets. The BTRC Petascale Gateway
provides the necessary storage and analysis facilities
all located in the BTRC area.

1-10 Gigabit Network

External Visitor,
Resources, 90% Researcher &
of our Computer Developer
Power Workstations
Beckman Institute, UIUC
RecentMacromolecular Modeling and Bioinformatics
BTRC for


How NAMD Runs Efficiently on Large Parallel Computers
Before Load Balancing: Much Idle CPU Time!
23,000 atoms on BlueGene/P 4K processor cores
WHITE = 32 ns/day
IDLE CPU TIME!
2.75 ms/step
Percentage CPU Utilization

Time



After Load Balancing: Efficient Computing!
23,000 atoms on BlueGene/P 4K processor cores
IDLE TIME REDUCED ~ 40% 63 ns/day
1.38 ms/step
Percentage CPU Utilization

Time



18
Computational Microscope on New Supercomputers
100M-atom performance on Jaguar and Blue Waters
Jaguar
BlueWaters Cray
Jaguar XT5 (ORNL)
4.000
224,000 cores
ns/day

3.000
2.000 BlueWaters XE6 (UIUC)
1.000 380,000 cores, 3000 GPUs

0
0 75000 150000 225000 300000

Number of Cores
1.50 CPU 12 cores
CPU 12 cores + 1 GPU
1.13 CPU 12 cores + 2 GPU
Tsubame CPU 12 cores + 3 GPU
0.75
ns/day

Tokyo Institute of Tech.
4224 GPUs 0.38

0
50,000 registered users 64 128 256 512 700
Number of Nodes

19

1M Atom VirusGPU and CPU performanceGPU
1M-atom stmv on TitanDev

GPU
CPU
5
Single STMV
PME every 4 steps
ns/day

1

0.2

1 2 4 8 16 32 64 128 256 512
number of nodes
BTRC for Macromolecular Modeling and Bioinformatics
Beckman Institute, UIUC


20

100M Atoms on TitanDev



Advanced Computer Facilities
and the
Computational Microscope
Illinois Petascale
Computing Facility



2
Molecular Dynamics Flexible Fitting (MDFF) Method
Electron microscopy X-ray crystallography

MDFF
APS at Argonne

FEI microscope

Acetyl – CoA Synthase

L. Trabuco, E. Villa, K. Mitra, J. Frank, and K. Schulten. Flexible fitting of atomic structures into electron
microscopy maps using molecular dynamics. Structure, 16:673-683, 2008.


4
MDFF is young (2008), yet already successful
Over 30 reports of MDFF applications:
• By Group Researchers:
Villa et al. PNAS (2009): EF-Tu ribosome complex
Seidelt et al. Science (2009): TnaC ribosome translation stalling
Becker et al. Science (2009): Sec61 ribosome complex
Frauenfeld et al. Nat. Struct. Mol. Biol. (2011): SecYE ribosome complex

Agirrezabala et al. PNAS (2012): Ribosome translocation intermediates

• By Outside Researchers:
Lorenz et al. PNAS (2010): actin-myosin interface
Bhushan et al. Nat. Struct. Mol. Biol. (2010): α-helical nascent chains
Armache et al. PNAS (2011): Eukaryotic ribosomal proteins
Armache et al. PNAS (2011): Translating eukaryotic ribosome
Guo et al. PNAS (2011): RsgA GTPase on ribosomal subunit
Becker et al. Nat. Struct. Mol. Biol. (2011) Dom34–Hbs1 stalled ribosome
complex
Wollmann et al. Nature (2011): Mot1–TBP complex
Strunk et al. Science (2011): Ribosome assembly factors
Lasker et al. PNAS (2012): Proteasome
Becker et al. Nature (2012): Ribosome recycling complex
MDFF derived ribosome structure
Villa et al. PNAS 2009

Tightly integrated into NAMD and VMD
• Capable of quickly fitting very large structures
• Adaptable to a wide range of applications


Computation and Experiment in Virology 10

Collaboration with A. Gronenborn et al, U. Pittsburgh
Molecular Dynamics Flexible Fitting of Tubular HIV Capsid:
In the tubular geometry the capsid contains only hexameric subunits.

EM fitted
density structure

view of hexameric
subunit during MDFF
final cross-correlation 0.96
Relaxation of capsid segment requires
13 million atom MD simulation on
Blue Waters, 50 ns ns done

ed within a cone-shaped capsid assembled from ≈1500 copies of the viral capsid protein (CA)
From Cylinder to HIV Capsid
ous attempts have been made to model the HIV capsid as a fullerene cone composed of ≈250
examers and 12 CA-pentamers, either by self-assembling of coarse-grained models [28] or
Replace hexameric units by pentameric units
o-atomic models [29]; however, these models lack the atomic-level interactions that hold the
d together.
Collaboration with A. Gronenborn et al, U. Pittsburgh

hexamers pentamer whole capsid made of
e 3: million atoms of the hexameric form of thehexamers found in cylindrical assemblies
10 (A) All-atom model unit CA protein as and 12 pentamers
V capsids in vitro. The model accurately captures the curvature of the tube. (B) CA pentame
60 million atoms
nded by ﬁve CA hexamers. (C) Schlegel diagram for a fullerene end apex connected to a cylindrica
of hexagonally linked carbon atoms.


Recent MDFF application: Intra-ring cooperation in group II chaperonins
Collaboration with Fei Sun (Chinese Academy of Sciences)
Zhang et al. Submitted. Mar 2012 Symmetry-restrained
MDFF applied to 12 EM maps
revealed role of electrostatic interaction in chaperonin
hetero-oligomer formation

EM Maps of Various Chaperonin Conformations

5


Our Grand Ribosome Collaboration

Joachim Frank Roland Beckmann Taekjip Ha Kurt Fredrick Ruben Gonzalez
(Columbia U.) (U. Munich) (UIUC) (Ohio State U.) (Columbia U.)

Cryo-EM Cryo-EM Single-molecule Mutagenesis Single-molecule
FRET experiments FRET

Structure (2008), Science (2009) x2, J. Mol. Bio. (2010), New collaborator New collaborator
PNAS (2009), Nature Struct. 1 submitted
Proteins (2011), Mol. Bio (2011),
EMBO J. (2011), 1 submitted
PNAS (2012)


Science 3: How Proteins AreMade from Genetic Blueprint
Low-resolution data High-resolution structure

Close-up of nascent protein
Frauenfeld, Gumbart et al. Nat. Struct. Mol. Bio. 18, 614-621, 2011.

Molecular Dynamics Flexible Fitting (MDFF)
Study of the Localization of Nascent Proteins 9

Ribosome expressing
membrane protein

James Gumbart

2.7 million atom simulation

Ribosome-SecY-complex

Molecular Dynamics Flexible Fitting (MDFF)
Study of the Localization of Nascent Proteins 9

Ribosome expressing Ribosome expressing a
membrane protein cytosolic protein

QM description of
cation-π interaction
between nascent chain
and exit tunnel
L.Trabuco, C. Harrison, E. Schreiner, and
K. Schulten. Recognition of the
regulatory nascent chain TnaC by the
2.7 million atom simulation ribosome. Structure, 18:627-637, 2010.

Ribosome-Trigger Factor-
Ribosome-SecY-complex complex

Structural characterization of mRNA-tRNA translocation intermediates

X. Agirrezabala, H. Liao, E.
Schreiner, J. Fu, R. Ortiz-
Meoz, K. Schulten, R. Green,
and J. Frank. PNAS
109:6094-6099, 2012.


Lots to be done!



17

Achievements Built on People
5 faculty members (2 physics, 1 chemistry, 1 biochemistry, 1 computer science);
8 developers; 1 system admin.; 16 post docs; 24 graduate students; 3 administrative staff

L. Kale, J. Phillips: NAMD J. Frank, Columbia U.: MDFF, ribosome
J. Stone, K. Vandivort: VMD R. Beckmann, U. Munich: ribosome
I. Solov’yov, D. Chandler, M. W. Baumeister, MPI: whole cell imaging
Sener, J. Strumpfer, J. Perilla, L. Liu, Greg Timp, UIUC, H. Gaub, U.
Y. Liu, J. Gumbart, Y. Chan, X. Munich: biosensors; N. Hunter, Sheffield:
Zou, E. Cruz-Chu: Science Proj.photosynthesis;.



The Computational Microscope Images Biomolecular Machines and Nanodevices - Klaus Schulten

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to The Computational Microscope Images Biomolecular Machines and Nanodevices - Klaus Schulten

Similar to The Computational Microscope Images Biomolecular Machines and Nanodevices - Klaus Schulten (20)

Recently uploaded

Recently uploaded (20)

The Computational Microscope Images Biomolecular Machines and Nanodevices - Klaus Schulten