Gurney

advertisement
Spiking neuron models of the basal
ganglia: dopaminergic modulation of
selection and oscillatory properties
Kevin Gurney, Mark Humphries, Rob Stewart
Adaptive Behaviour Research Group
University of Sheffield, UK
1
Rationale: basal ganglia and action
selection

Aim: to understand underlying function of
basal ganglia.

While learning is crucial – what is being learned?

Hypothesis: Main computational role of basal
ganglia is to perform action selection

Supported by high (systems) level model

2
Simple leaky integrators to represent population
dynamics
BUT…..
Beyond the systems level

Do more realistic models support the selection hypothesis?
Constraints provided by:



Specific neuronal properties
Physiological phenomena displayed by BG in toto….
If the price of a model performing selection is its failure exhibit these
phenomena, the selection hypothesis is in question

In particular, can models display oscillatory phenomena in
BG?

If so, then we can use the model to explore possible
function of these oscillations

3
Function or artifact?!
Systems level – the model architecture
Cortex (‘salience’ input)
input
-
+
striatum
Striatum
STN
STN
output
3 ‘channels’
output nuclei
4
Diffuse
projection
• Assumes relatively diffuse projection from STN
• Emphasises STN’s role as input nucleus
cf, Hazrati and Parent, 1992, Mink and Thach 1993, Nambu et al 2000, Sato et al 2000
New functional architecture:
selection and control pathways
Cortex/thalamus
Striatum (D1)
STN
EP/SNr
Gurney et al, 2001
Striatum (D2)
GP
Interpret GP
efferents as
control signals for
modulating
selection pathway
Diffuse
projection
5
Selection pathway
Control pathway
Oscillations in basal ganglia: matching
mechanisms to phenomena

Basal ganglia display a wide range of
oscillatory phenomena – from <1Hz to >100Hz

These are probably associated with a
correspondingly wide range of underlying
mechanisms

We focus on four BG features.



6

Intrinsic nature of STN-GP coupling
Dopaminergic modulation of this coupling
Rebound bursting in STN
Synaptic patterning
Constructing the model
7
Rebound bursting in STN
Beurrier et al 1999
Current
IT
IL
IK > IL burst ends
IK
Time
inhibition
8
Importance of synaptic patterning
Inhibition at soma or proximal dendrites acts divisively
(rather than ‘subtractively’)
70% of GPe input is proximal or somatic (Bevan et al
1997)
cortex
STN
Distal dendrites
Proximal dendrites
soma
GP
Captured phenomenlogically: use inhibition in proximal
9
dendrites/soma to explicitly ‘gate’ more distal input
Dopaminergic action in striatum
Ctx
Ctx
Glu
Glu
DA
D1
Striatum
D2
Striatum
decreased PSP
Increased PSP
W = W0(1 - λ)
W = W0(1 + λ)
10
DA
λ<1
Dopaminergic action in STN
Similar story in GP…
GP
Ctx
D2
DA
Glu
DA
GABA
STN
STN
W = W0(1 – k1 λ)
11
D2
K1, k2 < 1
W = W0(1 – k2 λ)
Dopamine: hypotheses

Low levels of dopamine serve to couple STN
and GP more tightly and to make STN more
sensitive to its input

Dopamine in striatum will make channel
selection easier to achieve
12
Model neurons: summary

Leaky Integrate and Fire with







13
AMPA NMDA, GABA, synaptic currents
Shunting inhibition at proximal dendrites and soma
Spontaneous currents
Rebound bursting in STN,
Dopamine in striatum, STN and GP.
Inter-neuronal delays
All of the above parametrised by best
estimates from the literature
Network

Based on systems level model

3 discrete channels

64 neurons per channel, per nucleus

Probabilistic connection scheme within
channels (only 25% of all possible
connections made)
14
Constraining phenomena 1: Low frequency
oscillations in STN-GP
(Magill et al, Neuroscience,106, 2001)

Low frequency oscillations (LFOs) in STN
are driven by cortical slow wave under
urethane anaesthesia.

GP does not oscillate in control (normal DA)
conditions. Only shows oscillation under
dopamine depletion (6-OHDA lesion)

Residual LFOs (with 6-OHDA lesion) in STN
and GP under cortical ablation
15
Data – STN control
16
Model - STN control
20Hz
Pseudo-eeg
STN unit
actvity
100
Power
Multi-taper spectrum
20
30
mean firing rate (Hz)
Bin count
1s
10
17
0
0
0
1
2
3
Frequency (Hz)
4
5
-1.5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
Spike Trig Wav av
20
10
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Data – GP control
18
Model - GP control (1)
20Hz
GP unit
actvity
50
Power
Multi-taper spectrum
50
20
Spike Trig Wav av
mean firing rate (Hz)
Bin count
1s
25
019
0
0
10
20
30
Frequency (Hz)
40
50
-1.5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
10
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Model GP control (2)
20Hz
GP unit
actvity
1s
20
Spike Trig Wav av
Power
Multi-taper spectrum
50
mean firing rate (Hz)
Bin count
100
10
20
0
0
0
25
Frequency (Hz)
50
-1.5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Data – STN DA-depleted
21
Model STN DA-depleted
20Hz
STN unit
actvity
Bin count
1s
Auto corr.
Power
1000
22
0
0
0
1
2
3
Frequency (Hz)
4
5
-1.5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
Spike Trig Wav av
30
mean firing rate (Hz)
Multi-taper spectrum
250
20
10
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Data – GP DA-depleted (in phase)
23
Model - GP DA-depleted (in-phase)
20Hz
GP unit
actvity
1s
2D Graph 7
Power
Multi-taper spectrum
24
50
20
mean firing rate (Hz)
Bin count
200
100
0
0
0
1
2
3
Frequency (Hz)
4
5
-1.5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
Spike Trig Wav av
10
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Data – GP DA-depleted (anti-phase)
25
Model - GP DA-depleted (anti-phase)
20Hz
GP unit
actvity
Multi-taper spectrum
20
Power
500
50
mean firing rate (Hz)
Bin count
1s
26
0
0
-1.5
0
1
2
3
Frequency (Hz)
4
5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
Spike Trig Wav av
10
0
-1.0
-0.5
0.0
Time(s)
0.5
1.0
Data – cortical ablation and DA-depleted
Most neurons do not show LFOs but residual LFO activity…
27
100
100
Multi-taper spectrum
Power
Bin count
Model - no cortex (DA-depleted)
50
1s
STN
0
-1.5
-1.0
-0.5
0.0
0.5
1.0
0
1.5
0
Time(s)
1
2
3
4
5
Frequency (Hz)
GP
1s
Bin count
50
Multi-taper spectrum
Power
Power
Multi-taper spectrum
20
25
10
28
0
0
0
0
25
Frequency (Hz)
50
-1.5
0
1
2
3
Frequency (Hz)
4
5
-1.0
-0.5
0.0
Time(s)
0.5
1.0
1.5
LFO counts
Neuron is LFO if significant peak in power spectrum below 1.5Hz
In DA control
conditions, no GP
LFOs, STN driven
by cortex
100
LFO in GP
promoted by DA
depletion
50
25
29
GP
-ct
x
Nctx
ST
GP
+c
tx
tx
+c
ST
N
GP
-ct
x
Nctx
ST
GP
+c
tx
+c
tx
0
ST
N
LFOs (%)
75
Control Data
DA dep. data
Control Model
DA dep. model
Residual LFO in
STN & GP under
cortical ablation
Mean firing rates (Hz)
Mean firing rates
20
10
30
GP
-ct
x
Nctx
ST
GP
+c
tx
N+
ctx
ST
GP
-ct
x
Nctx
ST
GP
+c
tx
ST
N+
ctx
0
Control Data
DA dep. data
Control Model
DA dep. model
LFO – mechanistic explanation

Low frequency oscillations associated with
rebound bursting will be ‘unmasked’ at low
levels of dopamine….

GP more likely to generate pre-conditioning
hyperpolarisation
31
Constraining phenomena 2: gamma
oscillations in STN
(Brown et al., Exp Neuro. 177, 2002)
There is gamma oscillation (4080Hz) in alert rats
This is increased (86% mean)
by systemic D2 agonist
(quinpirole)
Local field potential spectrum (control)
32
Model simulated D2 agonist
60
60
128% power
DA=0.2
control
→
increase
40
Power
Power
40
DA=0.8
‘D2 agonist’
20
20
Mean power spectra
(192 neurons)
0
0
0
20
40
60
80
Frequency
100
0
20
40
Frequency
60
33
No. of sinificant peaks
Control
DA agonist
Peaks in power
spectrum
60
40
20
0
0
20
40
60
Frequency (Hz)
80
100
80
100
Gamma oscillations: explanation

Gamma oscillations are associated with the
natural frequency of oscillation of the GP-STN
circuit


At control levels of dopamine, the presence
of some LFO masks gamma


34
determined by circuit delays
Can’t be doing gamma during quiet phase of LFO
period.
At higher levels of dopamine, gamma is unmasked
Selection experiments
Cortical input
(Mean firing rate)
ch2
ch1
1
35
2.5
time
Selection and switching
ch2
ch3
Mean firing rate SNr
ch1
Firing rate
Time
ch2
ctx
36
ch1
Ctx Ch1: 20 Hz
Ctx Ch2: 40 Hz
1
time
2.5
DA depletion prevents selection
ch3
ch2
Mean firing rate SNr
ch1
Firing rate
LFOs?
Time
ch2
ctx
37
Ctx Ch1: 12 Hz
Ctx Ch2: 20 Hz
ch1
1
time
2.5
Effects of DA depletion overcome by
highly salient action
ch2
ch3
Mean firing rate SNr
ch1
Firing rate
Time
ch2
ctx
38
ch1
Ctx Ch1: 20 Hz
Ctx Ch2: 40 Hz
1
time
2.5
DA increase results in simultaneous
selection
ch2
ch3
Mean firing rate SNr
ch1
Firing rate
Time
ch2
ctx
39
ch1
Ctx Ch1: 20 Hz
Ctx Ch2: 40 Hz
1
time
2.5
Summary

A spiking model of BG constrained by known
physiology is able to account for a range oscillatory
phenomena

Oscillations are modulated under Dopaminergic
control of STN and GP

The same model displays selection and switching
properties, thereby supporting the selection
hypothesis for BG function

Currently exploring computational role of LFOs

40
Perturb BG to selection in otherwise unresolved selection
competition?
The adaptive behaviour research group
Peter Redgrave
Paul Overton
Kevin Gurney
Tony Prescott
Mark Humphries
Ben Mitchinson
Rob Stewart
Ric Wood
Jonathan Chambers
Tom Stafford
41
Ψ
Download