Molecular dynamics simulations of sodium channel protein

Molecular dynamics simulations of sodium channel
protein
i - drug
d
i
interactions
i
Ben Corry
Overview
Introduction to:
• Voltage-gated sodium channels
• Molecular
o ecu a dy
dynamics
a cs modelling
ode g
Understanding the binding site and route of entry of:
• Pore blockingg compounds
p
• Voltage sensor inhibitors
Voltage gated sodium channels
Essential for initiating electrical activity in cells
p in response
p
to changes
g in membrane p
potential
Open
to raise voltage in cell
Able to discriminate between Na+ and K+, 20:1
Blockage of channels reduces nerve impulses
local anaesthetics
anti arrhythmics / anti
anti-arrhythmics
anti-epileptics
epileptics
Na+
Three topics:
Na+
K+
Na+
Na+
K+
1. How do sodium channels distinguish
between Na+ , K+ , Ca2+ ?
2 How do the channels open
2.
open, inactivate and
close?
3. Where do inhibitory drugs bind to sodium
channels? How do they get there? Can they
be made subtype selective?
Sodium channel structure
Voltage sensors
Selectivity filter
Pore
Structure of a bacterial voltage gated sodium channel
Payandeh et al, Nature July 2011
Eukaryotic and bacterial sodium channels
Available structures:
NavAb (Catterall et al) - ‘pre-open’?
- inactivated?
NavRh (Yan et al) - inactivated?
NavMs (Wallace et al) - open?
- with blocker bound
NavAe1 (Minor et al) - closed
NavAb / Nav1.7 Chimera (Payandeh et al)
Sodium channels as drug targets
Most local anaesthetics are sodium channel blockers
as are anti-epileptic and anti-arrhythmic drugs
Desire for subtype selectivity
9 types of human channel ( subunit)
CNS neurons
CNS neurons
CNS neurons
PNS neurons
Skeletal Muscle
CNS neurons
Heart
DRG neurons
DRG neurons
Anaesthetic block of bacterial channels has been reported. (Lee et al 2012, Bagneris 2014)
Can we use the bacterial structures as a first step to understanding binding in
eukaryotic channels?
Modelling drug-channel interactions
Molecular docking
compound
library
protein
•
Uses simplified interaction score
•
Assumes a static receptor structure.
•
Cannot account for access routes to
binding sites.
computer
•
Docking score dependent on
receptor structure, and including or
missing parts of the structure.
structure
•
Cannot predict explicit solvent
effects entropy …
effects,
Docked structure
Modelling drug-channel interactions
•
Accurate protein-drug interactions
•
Include
l d protein
i fl
flexibility,
ibili solvation,
l i
entropy
•
Can examine entry / exit
•
Requires
q
accurate force field
parameterisation
•
Time consuming
•
Methods required to overcome time
requirement.
requirement
Molecular dynamics
Local anaesthetics are
known to block central pore
Types of channel block:
Tonic block – Drug can enter and block
channel in the closed state
Use dependent block – Drug can only
enter / bind channel when it is in open or
inactivated state
closed
open
Three channel states
inactivated
Measuring channel block
Types of channel block:
Tonic block – Drug can enter and block
channel in the closed state
Use dependent block – Drug can only
enter channel when it is in open (or
inactivated) state
Example
p of tonic block
Use dependent
d
d
tonic
Lee, Goodchild & Ahern, J. Gen. Physiol, 2012
Tonic block of bacterial sodium channels
How do drugs enter closed channel?
Where do they bind to the channel?
M
Membrane
e
Pore
Recent structure (2011)
shows possible route of
entry
‘Fenestrations’
route of drug entry?
A the
Are
th fenestrations
f
t ti
bi
big enough?
h?
Do they differ due to functional state?
Closed
activation gate
Do theyy differ amongg subtypes?
yp
Fenestration size in different bacterial channels
160ns MD simulations
NavAb
Closed
NavAb
‘Inactivated’
Inactivated
NavMs
‘O
‘Open’
’
NavRh
‘I ti t d’
‘Inactivated’
Large enough for small molecule blockers to pass
Potential dependence
p
on channel type
yp
No obvious dependence on channel state
Corry, Lee, Ahern. Hand. Exp. Pharmacol. 2014
Kaczmarski & Corry, Channels, 2014
Eukaryotic channels – a model of Nav1.4
Large
Small
Little sequence differences among these residues in channel subtypes
How do current drugs work on bacterial channels?
Benzocaine
Phenytoin
Display tonic block in bacterial channels similar to eukaryotic channels
Perhaps less ideal as models of use dependent block
Parameterisation: Check via computed water/octanol partition coefficients
and lipid partitioning
Martin, Chao, Corry. Biophys Chem. 2014
A
Lipid partitioning
3
Density (atoms
s/Å )
10
8
6
4
Water
P
CH2
2
O
0
Benzocaine CGFF
Benzocaine HF
Phenytoin CGFF
Phenytoin HF
Free Energy (k
kcal/mol)
6
4
2
0
-2
-4
-6
-30
30
-20
20
-10
10
0
Position (Å)
10
20
30
B
Finding the binding sites of benzocaine and phenytoin in NavAb
3 approaches:
• Equilibrium simulations
• Metadynamics
• Umbrella sampling
p g
All show similar binding positions
Cluster analysis from
3 x 125ns MD simulations
Where do the drugs bind?
Benzocaine
activation gate
Phenytoin
fenestration
Martin & Corry. PLoS Comp Biol. 2014
Where do the drugs bind?
Fenestration
Phen
nytoin
Ben
nzocaine
Activation gate
Non-specific hydrophobic interactions
Martin & Corry. PLoS Comp Biol. 2014
Where do the drugs bind? Metadynamics
Drugs can
exit closed
gate!
Binding Affinity – Free Energy Perturbation
Martin & Corry. PLoS Comp Biol. 2014
Benzocaine can move easily through fenestration
Phenytoin experiences larger barriers but passage is plausible
How good are bacterial channels as models of eukaryotic channels?
bNav lack the key aromatic residues responsible for binding in eNav
What if we add the aromatics back in?
WT
1 subunit
4 subunit
What if we add the aromatics back in?
No exit through
closed gate
Smith & Corry. Channels. 2016
Mutant bacterial channels as better models of human channels
1. Activation gate binding site disappears
2. Closed gate entry no longer possible
3. Binding affinity closer to eukaryotic results
G (kcal/mol) Kd (M)
eNav Exp
Wild Type
yp
5.6 ± 1.0
78 ± 30
300 1200 m
300-1200
Mutant
4.4
644
Wild Type
6.1 ± 0.1
30 ± 10
Mutant
8.2
1.0
Compound
Benzocaine
Phenytoin
4-10
4
10 m
Mutant bacterial channels may be better models of eukaryotic block
Some channels do not easily tolerate these mutations
Smith & Corry. Channels. 2016
Binding of brominated drug
Making subtype selective pore blockers
Sodi m channel DIV seq
Sodium
sequence
ence alignment
Pore domain
Filter
S6
Known binding
b
residues
New residues from this study
S5
Making subtype selective pore blockers
Sodium channel DIV sequence alignment
Voltage sensor domain
S1
S2
S3
S4
Voltage sensor inhibitors
Selective for Nav1.7 IC50 = 28nm
c.f. IC50 for Nav1.3, Nav1,.5 > 10m
Binds to DIV extracellular side of voltage
sensor
PF-04856264
Residue on S2, S3 responsible for selectivity
Can we use the bacterial structures as a first step to understanding binding and mode
of action in eukaryotic channels?
Where do they bind?
‘Flooding’ 3.9 s simulation
Where do they bind?
How do the VS inhibitors work?
Drug locks voltage sensor in activated state
Holds channel in inactivated state
Equivalent to pore blockage
Can we improve subtype selectivity?
Problems with MD approach:
Timescale of channel inhibition
Strength of interaction (me mM vs M)
Intracellular ‘S6’ binding site: clearly defined,
defined but largely due to hydrophobic interactions
Characterising the VS binding site in NAvAb
NavAb
Nav1.7
View toward the membrane
S2
S4
R2
T64
R3
S2 sequence alignment
N67
80
83
69
68
89
87
68
R4
View from the membrane
V61
S2
S4
L65
Q68
I71
S105
S3
Ahuja et al 2015
Route of binding to NavAb
Protein route
Lipid route
p
protein
lipid
p
protein
lipid
Blue: early
Red: late
Route of binding to NavAb
Free energy (kc
F
cal/mol)
10
5
0
Kd = 4 m
-5
10
15
20
Distance of warhead from R4 backbone
25
Route of binding to Nav1.7
Summary
Great desire for more selective sodium channel inhibitors
Pore Blockers:
Feasibility of fenestration entry
differences among structures
barriers to entryy
Clarify binding positions and poses
Difficulty in generating subtype selectivity
Voltage sensor Inhibitors
Clarify binding site in bacterial Nav
Indentifying barriers to entry
How good are bacterial Nav as models of eukaryotic Nav?
MD simulations have to be carefully applied to overcome timescale,
sampling limitations
Thank You:
Lewis Martin
Joe Kaczmarski
Rebecca Chao
Natalie Smith
Delin Sun
Michael
Mi
h l Thomas
Th
Zhongjin He
Kasia Walczewska-Szewc
Siri Sondergaard
Nishank Shah
karri.anu.edu.au
University of Iowa
Chris Ahern
Sora Lee
State dependence
Does binding differ between – closed, inactivated and open states?
Expanding the range of drugs studied
Lidocaine
QX314
Ranozaline
Brominated blockers
PI1
PI3
Carbamazepine
(Bagneris… Wallace, PNAS, 2014)
PI7
PI5
Ion Selectivity
To enter a narrow channel,
H2O molecules must be
removed from the ion
K+ is larger than Na+
K+
Na+
So.... potassium channels will be wider than sodium channels, wont they?
K+ channel
Na+ channel
Cost to remove water balanced by
interactions with protein
Single ion potential of mean force
Sex
Scen
10
Sin
Free enerrgy (kcal/mo
ol)
8
6
Scav
4
Na
+
2
0
-2
-10
Scav
0
Sin
Scen
Sex
10
20
Position in channel (A)
Corry & Thomas, J. Am. Chem. Soc. 2012
Two ion potential of mean force
1
2
kcal/m
mol
3
4
Corry & Thomas, J. Am. Chem. Soc. 2012
Na+ / K+
Na+ / K+ selectivity
Na+ / Na+
Na+ / K+ selectivity
Not simply due to:
Size of pore
To bigg to discriminate on bare ion size
Both ions can fit with full hydration shell
g
Number of coordinatingg ligands
(both ions have favourable numbers)
Simple presence of Glu residues
(site by Glu not selective)
Geometry of ions in plane with Glu177
5.9 Å
B
6.1 Å
A
49Å
4.9
4.7 Å
5.0 Å
2.3 Å
4.8 Å
2.6 Å
59Å
5.9
62Å
6.2
uncharged pore
Both ions could fit through
with solvation shell
K+
charged
h
d pore
Presence of Glu creates a
favoured geometry
K+
-
-
charged pore
Only Na+ can fit through with
water in this geometry
-
Na+
-
Corry & Thomas, J. Am. Chem. Soc. 2012
Dependence on pore size
16
14
+
+
10
K Fit
Na Fit
G ((kcal/mol)
12
8
6
4
2
0
5.0
5.5
6.0
6.5
7.0
Glu-Glu separation (Å)
Corry & Thomas, J. Am. Chem. Soc. 2012
Selection between Na+ and Ca2+
Ca2+ binds
C
bi d strongly
t
l in
i filter
filt
Ca2+ not stabilised in centre of filter
Na+ can bypass resident Ca2+
How to make a Ca2+selective channel?
Corry, Peer J. 2013
Understanding selectivity in simple pores
He et al, ACS Nano. 2013
Na+ selection is harder than K + selection
w/o knock on
with knock on
He et al, ACS Nano. 2013