Exploring the free energy landscapes of
carbohydrates using microsecond molecular
dynamics simulations in water
Vision: rationalise glycomics
Understand carbohydrate structure-activity & inform design
Turnbull & Field, NCB (2007)
OPPORTUNITIES
Poorly understood (structure-activity)
New 3D-structural insight required to harness potential
1
Structure and function
Challenges in 3D-characterisation:
• X-ray not tractable: heterogeneous, flexible, charged
• NMR equilibrium observables: difficult to measure, ambiguous
• Short and long timescale events, interaction with water essential
Glycosidic linkage key determinant of 3D structure?
2KQO
Studies on heparin – a sulphated sugar
Sattelle et. al. (2010) Carb. Res.; Sattelle & Almond (2010) J. Comp. Chem.
Sulphation has little effect on linkage, is there something else?
2
Is puckering the missing link?
Carbohydrate ring can pucker “flex”
2KQO
Why has this not been proposed previously?
Relatively unexplored degree of freedom
Microsecond blindspot:
• Timescales are long for simulation
• Short for experiment
Solution:
Combined high performance computing / experimental solution
3
Interdisciplinary strategy: NMR & MD
Simulations are useful (e.g. atomic resolution of any sequence)
Limited to 100’s of ns with CPU-based cluster (much biology > 100ns)
MD slow-exchange
NMR fast-exchange (> ms)
GPUs?
Dror et al, JGP(2010) 135(6), 555-562
4
Can we equilibrate pucker using GPUs?
Equilibrium: theoretical prerequisite for prediction
free energies (DG), 3D-structure, molecular properties
Adopted GPUs to converge pucker of L-iduronic acid
IdoA: controversial heparin monosaccharide
Antithrombin-heparin binding: IdoA 2SO conformer initiates anticoagulation
4C
1
2S
O
1C
4
C2
C1
O5
C4
5
Setup: hardware & software configurations
Hardware
IdoA
OMe
OH
1 x Linux box (E5520 CPU)
2 x nvidia GPUs (GTX260)
-
O
C5
C4
C3
OH
OH
IdoA2S
OMe
OH
-
O
C5
C1
O
C3
C4
NVT Simulations (~250 ns/day)
C2
O
Software
ACEMD (parallel)
C1
O
C2
O
OSO3-
OH
SULFATION
IdoA, IdoA2S & GlcA
32 Å
GlcA
Initiated in 1C4- and 4C1-puckers
EPIMERISATION
6 trajectories (5.3 μs each)
Potential: GLYCAM06/TIP3P
32 Å
6
R1: ~3μs to equilibrate fast exchanging IdoA
1C -chair
4
skew-boat
4C -chair
1
θ describes pucker (JACS 1975)
IdoA2S
IdoA
GlcA
7
R2: fine-tuning in biology
8
R3: long timescales reveal force-field errors
9
RIGID
Protein-binding
S-domain
1C -chair
4
FLEXIBLE
Linker
NA-domain
skew-boat
4C -chair
1
Mobli et. al. (2007)
GlcNS6S
IdoA2S
GlcNAc
GlcA
Un-sulphated (NA-)
Sulphated (S-)
A new biological hypothesis
GlcNAc (mono)
GlcA (mono)
μs puckering: maps dynamic 3D-structure to function (missing link)
10
Proof of concept
DG library > 250 μs
All-atom (months)
Mesoscale (minutes)
Biologically-relevant spatiotemporal simulations
11
Conclusions
Puckering, the missing link in structure-activity
GPUs are enabling progress toward goals
Opportunities
Rationalise glycomics (cf. genomics, proteomics)
Inform design (e.g. QSAR, virtual screening)
Acknowledgements
BBSRC, UOB (900 MHz NMR)
AA
BS
JS