[PPT] - Neuronal Cell Death and Synaptic Pruning driven by Spike-Timing PowerPoint Presentation

SLIDE 1

Neuronal Cell Death and Synaptic Pruning driven by Spike-Timing Dependent Plasticity

Javier Iglesias1,2 and Alessandro E.P . Villa2 ICANN 2006 Athens, 2006-09-11

1 Information Systems Department, University of Lausanne, Switzerland 2 Laboratory of Neurobiophysics, University Joseph-Fourier, France

<javier.iglesias@unil.ch>

SLIDE 2

introduction: synaptogenesis and synaptic pruning

1 NB 0.5 1 2 5 10 adult aged (74-90) years 10 6 8 4 12 2 neurons / mm x10

3 4

NB 0.5 1 5 10 15 20 40 60 80 100 years 15 10 20 5 synapses / mm x10

3 8

modified from Huttenlocher, Synaptic density in human frontal cortex – developmental changes and effects of aging, Brain Research, 163:195–205, 1979

SLIDE 3

introduction: existing work

2

The memory performance of a network is optimally maximized if,

under limited metabolic energy resources restricting their number and strength, synapses are first overgrown and then pruned.

Chechik et al., Synaptic pruning in development: A computational account, Neural Computation, 10(7):1759–77, 1998

Neuronal regulation might maintain the memory performance of

networks undergoing synaptic degradation.

Horn et al., Memory maintenance via neuronal regulation Neural Computation, 10(1):1–18, 1998

STDP has been shown to maintain the postsynaptic input field.

Abbott et al., Synaptic plasticity: taming the beast Nature Neuroscience, 3:1178–83, 2000

SLIDE 4

model: network

3

50

50

x

50

0.0 0.2 0.4 0.6

probability

e

50

50

x

50

50

y

0.0 0.2 0.4 0.6

probability

+50

50

x

+50

50

y

f

+50

50

x

+50

50

y

+50

50

+50

50

y x

+50

50

+50

50

y x

g

250 500 200 400

cell count connection count

250 500 200 400

cell count connection count

a b c d h

50 0 y

i e i i

600

e e e i

600

Iglesias et al., Dynamics of Pruning in Simulated Large-Scale Spiking Neural Networks, Biosystems, 79(1-3):11-20, 2005

SLIDE 5

introduction: leaky integrate and fire neuromimetic model

4

V(t) S(t) B(t) w(t) ~190 excitations ~115 inhibitions

Type I = excitatory 80% Type II = inhibitory 20% Vrest =

76

[mV] θi =

40

[mV] τmem = 15 [ms] trefract = exc:3 inh:2 [ms] λi = 5 [spikes/s] n = 50

Vi(t+1) = Vrest[q]+(1−Si(t))·((Vi(t)−Vrest[q])·kmem[q])+

j

wji(t)+Bi(t) Si(t) = H(Vi(t) − θqi) wji(t + 1) = Sj(t) · Aji(t) · P[qj,qi] Bi(t + 1) = Preject(λqi) · n · P[q1,qi]

SLIDE 6

model: STDP and pruning

5

Lji(t + 1) = Lji(t) · kact[qj,qi] +(Si(t) · Mj(t)) −(Sj(t) · Mi(t))

post pre S (t)

i

S (t)

j

time time

SLIDE 7

model: STDP and pruning

5

Lji(t + 1) = Lji(t) · kact[qj,qi] +(Si(t) · Mj(t)) −(Sj(t) · Mi(t))

post pre S (t)

i

S (t)

j

time time post pre S (t)

i

M (t)

i

S (t)

j

M (t)

j

time time

SLIDE 8

model: STDP and pruning

5

Lji(t + 1) = Lji(t) · kact[qj,qi] +(Si(t) · Mj(t)) −(Sj(t) · Mi(t))

post pre S (t)

i

S (t)

j

time time post pre S (t)

i

M (t)

i

S (t)

j

M (t)

j

time time post pre S (t)

i

M (t)

i

S (t)

j

M (t)

j

time time L (t)

ji

SLIDE 9

model: STDP and pruning

5

Lji(t + 1) = Lji(t) · kact[qj,qi] +(Si(t) · Mj(t)) −(Sj(t) · Mi(t))

post pre S (t)

i

S (t)

j

time time post pre S (t)

i

M (t)

i

S (t)

j

M (t)

j

time time post pre S (t)

i

M (t)

i

S (t)

j

M (t)

j

time time L (t)

ji

Lji(t) Aji(t) time [s] 50 100 150 4 2 1

wji(t + 1) = Sj(t) · Aji(t) · P[qj,qi]

SLIDE 10

model: synaptic adaptation examples

6

time L0 L1 L2 L3 L4 time L0 L1 L2 L3 L4 time L0 L1 L2 L3 L4

a b c

A4 A3 A2 A1

SLIDE 11

model: graph considerations

7

Strongly Interconnected (SI) units at the end of the simulation set of cells (discarding input units) maintaining kout ≥ 3 and kin ≥ 3 with strongest activation level (Aji(t) = 4) with units with the same properties. Neighbourhood all excitatory units (including input units) with at least one projection to or from SI-units.

SLIDE 12

results: circuit

8 Iglesias et al., Emergence of Oriented Cell Assemblies Associated with Spike-Timing-Dependent Plasticity, LNCS 3696:127-132, 2005

layer 1 layer 2 layer 3 layer 4 layer 5

5326 5697 2800 5550 7199 7297 9661 992 4695 1539 1202 8402 1608 8889 931 931 8120 8120 8022 2267 3027 2007 1152 2666 8467 1151 7928 5724 5724 490 490 490 4547 8863 4761 2279 804 8267 8267 8267 1234 1048 1435 8300 7794 1049 2532 1842 4622 2427 2427 9983 1845 2848 492 492 1472 1472 1472 7235

time = 500 s

SLIDE 13

results: circuit

8 Iglesias et al., Emergence of Oriented Cell Assemblies Associated with Spike-Timing-Dependent Plasticity, LNCS 3696:127-132, 2005

layer 1 layer 2 layer 3 layer 4 layer 5

5326 5697 2800 5550 7199 7297 9661 992 4695 1539 1202 8402 1608 8889 931 931 8120 8120 8022 2267 3027 2007 1152 2666 8467 1151 7928 5724 5724 490 490 490 4547 8863 4761 2279 804 8267 8267 8267 1234 1048 1435 8300 7794 1049 2532 1842 4622 2427 2427 9983 1845 2848 492 492 1472 1472 1472 7235

time = 500 s

layer 1 layer 2 layer 3 layer 4 layer 5

5326 5697 2800 5550 7199 7297 9661 992 4695 1539 1202 8402 1608 8889 931 931 8120 8120 8022 2267 3027 2007 1152 2666 8467 1151 7928 5724 5724 490 490 490 4547 8863 4761 2279 804 8267 8267 8267 1234 1048 1435 8300 7794 1049 2532 1842 4622 2427 2427 9983 2848 1845 492 492 1472 1472 1472 7235

time = 0 s

SLIDE 14

model: stimulus

9

t = 1 stimulation

nset

t = 2 t = 3 t = 4 t = 5 t = 6 t=duration (50, 100

r 200 ms)

[...]

A B

[...]

every 2 seconds 10 groups of 40 units activated in sequence (10% input units) during 100 time steps (5×A + 5×B or 5×B + 5×A)

Animated sequences are available for both stimuli A and B.

SLIDE 15

model: stimulus (cont.)

10 1 10 50 11 51 60 61 100

A1 A10 A1 A10 B1 B10 B1 B10

1 ordered sequence of set A

stimulus

nset

stimulus

ffset

2001

A1

1 ordered sequence of set B

time steps [ms]

stimulus

nset

SLIDE 16

model: cell death mechanisms

11

apoptosis induced by excessive firing rate

50 ms running window firing rate threshold: θexc

ν

= 245 sp/s for excitatory units θinh

ν

= 250 sp/s for inhibitory units death probability function: Papopt(t) =

0.5·t2−4.5·10−6·t3 44·(2.5·106+6·10−3·t2)

Papopt(t = 100) = 4.5 · 10−5 Papopt(t = 700) = 2.2 · 10−3 Papopt(t = 800) = 2.9 · 10−3

apoptosis induced by lack of excitatory afferents

loss of all excitatory inputs due to: apoptosis of pre-synaptic unit STDP driving to Aji(t) = 0

SLIDE 17

model: simulation layout

12

0 ≤ t < {700, 800} ms (initial phase)

apoptosis induced by excessive firing rate

{700, 800} ≤ t < 105 ms

Spike-Timing Dependent Plasticity ⇒ synaptic pruning ⇒ apoptosis induced by lack of excitatory inputs

t = {1000, 3000, 5000, . . .} ms

100 ms lasting stimuli 50 presentations (random mix: 50% AB and 50% BA)

SLIDE 18

results: excitatory vs. inhibitory cell death

13

100 95 90 85 80 75 1000 750 500 250 time [ms] surviving units [%] apoptosis phase STDP phase excitatory units inhibitory units Lag ≈ 120 ms

inh

Lag ≈ 190 ms

exc

800

SLIDE 19

introduction: detection of spatiotemporal patterns of activity

14

A B C simultaneous recording of spike trains

a

time (ms)

SLIDE 20

introduction: detection of spatiotemporal patterns of activity

14

A B C simultaneous recording of spike trains

a

time (ms)

detection of statistically significant spatiotemporal firing patterns <A,C,B; ∆t1,∆t2>

cell # B cell # C

time (ms)

cell # A

Δt1 Δt2

b

patterns found n=3 expected count N=0.02 significance of this pattern pr ( 3, 0.02) ≈ 1.3·10

6 < 0.001

SLIDE 21

introduction: detection of spatiotemporal patterns of activity

14

A B C simultaneous recording of spike trains

a

time (ms)

detection of statistically significant spatiotemporal firing patterns <A,C,B; ∆t1,∆t2>

cell # B cell # C

time (ms)

cell # A

Δt1 Δt2

b

patterns found n=3 expected count N=0.02 significance of this pattern pr ( 3, 0.02) ≈ 1.3·10

6 < 0.001

for methods, see Villa et al., 1999; Tetko and Villa, 2001

SLIDE 22

introduction: representation of spatiotemporal patterns of activity

15

A B C

a

time฀(ms)

A B C rasters฀of฀spikes฀aligned฀on฀pattern฀start

A C B

time฀(ms)

simultaneous฀recording฀of฀spike฀trains

c

SLIDE 23

results: spatio-temporal pattern of activity

16

stopping firing rate-induced apoptosis at t=700ms

<79, 79, 79; 453±3.5, 542±2.5>

+453 +542

b a c

25 50 100 75 time฀[s]

400

+1600 lag฀[ms]

400

+1600 lag฀[ms]

SLIDE 24

results: spatio-temporal pattern of activity

17

stopping firing rate-induced apoptosis at t=800ms

<13, 13, 13; 234±3.5, 466±4.5>

+234 +466

400

+1600 25 50 100 75 time฀[s] lag฀[ms]

b a c

400

+1600 lag฀[ms]

SLIDE 25

summary

18

inhibitory units enter apoptosis

about 70 ms before excitatory units

death dynamics of both populations followed

the probability function to die Papopt(t)

addition of cell death to the model

improved stability of the network stopping apoptosis at 700 ms lead to larger number of surviving units with lower mean firing rate stopping apoptosis at 800 ms lead to smaller number of surviving units with greater mean firing rate

maintained ability to let emerge cell assemblies