Unité de Neurosciences, Information et Complexité (UNIC)
CNRSGif-sur-Yvette, France
http://cns.iaf.cnrs-gif.fr
How much stochastic is neuronal activity ?
Alain Destexhe
FACETS(EU IST)
Yayoyi Kusama, Fireflies on the Water
Contributors:Theory: Claude Bedard, Sami El Boustani, Olivier Marre,
Serafim Rodrigues, Michelle Rudolph (UNIC),Experiments: Diego Contreras (U Penn, USA), Igor Timofeev,
Mircea Steriade (Laval University, Canada)
WessbergCrist & Nicolelis
(2002)
Ensemble activityin the cortex of abehaving rhesus
monkey
Neuronal activity in awake monkeyComplex spatiotemporal patterns of neuronal discharges
Plan
1. Characterization of neuronal activity in theneocortex of awake animals
2. Characterization of LFPs
3. Modeling neuronal activity in awake cortex
Multisite bipolar LFP recordings
Destexhe et al., J. Neurosci.,1999
Awake
Multisite bipolar LFP recordings
Destexhe et al., J. Neurosci.,1999
VLC media file(.mp3)
VLC media file(.mp3)Data: Destexhe, Contreras & Steriade, J. Neurosci. 1999
Music: http://www.archive.org/details/NeuronalTones
Multiunit extracellular recordings in awake cats
Wake: Poisson:
Multiunit extracellular recordings in awake cats
Softky & Koch, J Neurosci. 1993Bedard, Kroger & Destexhe, Phys Rev Lett 2006
Apparent stochastic dynamics!
Multiunit extracellular recordings in awake cats
Bedard, Kroger & Destexhe, Phys Rev Lett 2006
Apparent stochastic dynamics!
Multiunit extracellular recordings in awake cats
Marre, El Boustrani, Fregnac & Destexhe(Phys Rev Lett, 2009)
Correlated
Statistics of spike patterns in cat parietal cortex
Uncorrelated
Intracellular recordings in awake and sleeping animals
(Courtesy of Igor Timofeev, Laval University, Canada)
Synaptic “noise” in vivo
Pare et al.J Neurophysiol. 1998
Steriade et al.J Neurophysiol. 2001
Destexhe et al.Nature ReviewsNeurosci. 2003
Conductance measurements in vivo
Paré et al., J. Neurophysiol. 1998Destexhe et al., Nature Reviews Neurosci. 2003
Characterization of up-states in vivo
Microperfusion of TTX in cat parietal cortexunder ketamine-xylazine anesthesia
Paré et al., J. Neurophysiol. 1998Destexhe et al., Nature Reviews Neurosci. 2003
Characterization of up-states in vivo
Vm distributionsin different network states
Destexhe & RudolphNeuronal Noise
Rudolph et al.J. Neurophysiol 2005
J. Neurosci. 2007
Characterization of up-states in vivo
Destexhe & RudolphNeuronal Noise
Rudolph et al.J. Neurophysiol 2005
J. Neurosci. 2007
Conductance measurements in different network states
Rudolph, Pospischil, Timofeev &Destexhe, J. Neurosci, 2007
Conductance measurementsin awake cats
Extracting conductances from in vivo activity
Spike-triggered averages of conductances
Rudolph et al.,J. Neurosci,2007
Characterization of up-states in vitro
Destexhe & RudolphNeuronal Noise
(data fromHasenstaub & McCormick)
Characterization of up-states in vitro
Destexhe & RudolphNeuronal Noise
(data fromHasenstaub & McCormick)
Characterization of up-states in vitro
Destexhe & RudolphNeuronal Noise
(data fromHasenstaub & McCormick)
Synaptic activity is intense and noisy,essentially Gaussian distributed (bothfor Vm and conductances)
Responsible for a “high-conductance state”(3 to 5-fold larger than resting conductance)
Statistics of neuronal activity is very closeto Poisson processes
Importance of inhibition (both for absoluteconductance and for the dynamics ofspike initiation)
Characterizing neuronal activity
Destexhe & Rudolph, Neuronal Noise, Springer 2010
Conclusions
Plan
1. Characterization of neuronal activity in theneocortex of awake animals
2. Characterization of LFPs
3. Modeling neuronal activity in awake cortex
PSD of Local Field Potentials
Bedard et al.,Phys Rev Lett 2006
Modeling LFPs
“Diffusive” LFP Model
Bedard & Destexhe, Biophysical Journal, 2009
Coulomb’s law:
Ionic diffusionin homogeneousmedium
Electrode
PSD of the LFP:
Modeling LFPs
Bedard & Destexhe, Biophysical Journal, 2009
Transfer function LFP - Vm activity
Bedard, Rodrigues,Roy, Contreras &DestexheSubmitted
Fitting differenttransferfunctions toexperimentaldata alsosuggestsWarburgimpedance
“Avalanche dynamics” from LFPs in vivo
Petermann et al., PNAS 2009
Avalanche analysis from LFP activity (awake cat)
Touboul & Destexhe,PLoS One, 2010
Avalanche analysis from LFP activity (awake cat)
Avalanche analysis from LFP activity (awake cat)
Avalanche analysis from LFP activity (awake cat)
Shuffled LFP peaks (random process!)
Touboul & Destexhe,PLoS One, 2010
Avalanche analysis from LFP activity (awake cat)
Shuffled LFP peaks (random process!)
Touboul & Destexhe,PLoS One, 2010
LFPs are broad-band with 1/f scaling at low freq.
1/f scaling can be explained by effect of diffusion
Power-law distributions from LFP peaks can alsobe explained by thresholding procedure
Similar to neuronal activity, a lot can be explainedby purely stochastic mechanisms...
Characterizing LFP activity
Conclusions
Plan
1. Characterization of neuronal activity in theneocortex of awake animals
2. Characterization of LFPs
3. Modeling neuronal activity in awake cortex
Network models of self-sustained irregular states
Network models of asynchronous irregular states
Brunel, J Physiol Paris, 2000
Self-sustained asynchronous irregular states
Vogels & Abbott,J Neurosci 2005
El Boustani & Destexhe,Neural Computation 2009
Analysis of AI states
El Boustani et al.,J Physiol Paris, 2007
Analysis of AI states
El Boustani et al.,J Physiol Paris, 2007
Analysis of AI states
El Boustani et al.,J Physiol Paris, 2007
Analysis of AI states
El Boustani et al.,J Physiol Paris, 2007
20 timestoo many!
Modulation of information transfer by network activity
How to obtain models consistentwith conductance measurements ?
El Boustani & Destexhe,Neural Computation 2009
Mean-field model of AI statesMacroscopic modeling of AI states in spiking networks
Optical imaging 1 pixel = network ofrandomly-connected neurons
Mean-field model of AI states
Mean-field model of AI states
Model predictionNumerical simulation Difference
Conductancemaps
Mean-field model of AI states
Network models with realistic conductance patternsBest model: N=16000, 320 synapses/neuron
Vogels & Abbott, J Neurosci, 2005
Comparison
Network models with realistic conductance patterns
Conclusions
Randomly connected networks of IF neuronscan generate dynamics which reproduceexperimental observations in the awake brain...
... except for conductances measurements!
Mean-field models can be used to identifynetwork configurations with correctconductance state (work in progress...)
Modeling the awake neocortex
Thanks to the team...Michelle Rudolph
MartinPospischilSami
El BoustaniClaudeBedard
OlivierMarre
JonathanTouboul
SerafimRodrigues