Oscillatory and stochastic dynamics in thalamo-cortical networks
Dir. Alain Destexhe


The laboratory of Alain Destexhe is specialized in computational neuroscience and theoretical neuroscience. The research themes are the modeling of the integrative properties of neurons and networks of cerebral cortex during "active" states (such as wakefulness), the role of noise on information processing of single neurons or networks of neurons, macroscopic aspects such as local field potentials and the collective dynamics at large scales, as well as the rhythmic properties of thalamic and cortical networks during sleep and epilepsy. The laboratory also participates to dynamic-clamp experiments (in collaboration with Bal's lab at UNIC), where computational models directly interact with living neurons. The laboratory also participates to the conception of methods to analyze experimental data in vivo (intracellular, optical imaging), or to record in high-resolution (AEC metthod), in collaboration with several laboratories across Europe and North-America.  This research is supported by grants from different institutions, such as the ANR (Agence Nationale de la Recherche), the EC (European Commission), the MRC (Medical Research Council of Canada), the NIH (National Institutes of Health, USA), and the HFSP (Human Frontier Science Program). 


Techniques used

Computational modeling and numerical simulations of neurons and neural networks, Theoretical calculus, Dynamic-clamp experiments, Conductance analysis from intracellular recordings, Morphological reconstruction with optical microscopy (Neurolucida)

For more details

For further information, see the laboratory's web site http://cns.iaf.cnrs-gif.fr, or click on the icons below:




The research themes developed in the "Computational Neuroscience" laboratory, animated by Alain Destexhe, stand at the interface between several disciplines, such as biophysics, physics (dynamical systems) and neurosciences. The subjects investigated range from the microscopic level (ion channels, single neurons) to the macroscopic and population levels (networks of neurons).

1. Biophysics of elementary neuronal processes

At the channel level, we study the activation properties of ion channels important for neuronal function, such as the low-threshold calcium current IT or the cationic current Ih ("anomalous rectifier"). These properties are modeled using Hodgkin-Huxley type models or models based on state-transition (Markovian) kinetics. The same type of approach can be used for modeling synaptic transmission via the principal types of receptors (AMPA and NMDA for glutamate, GABAA and GABAB for GABA). Presently, we study the biophysical bases of synaptic plasticity, both short-term (facilitation, depression) and long-term (Spike Timing Dependent Plasticity or STDP), in collaboration with several experimental groups in USA and UK.

Another direction is presently to model local field potential (LFP or EEG) signals based on first principles of electromagnetism (Maxwell equations). The problem of the filtering of signals by extracellular media is central for interpreting extracellular potentials such as LFPs. We have shown that one of the physical bases for such a filtering is the inhomogeneity of extracellular media. The finite velocity of charge migration results in delayed polarization phenomena, which is at the origin of a strong frequency filtering by extracellular media. A precise model of LFPs that takes this contribution into account is in construction and should allow a better characterization of neuronal activity by analyzing LFPs. This type of consideration can also be applied to the re-equilibration of charges following membrane potential changes. Integrating this effect into neuronal cable equations is in progress (in collaboration with Claude Bédard).

2. Integrative properties of single neurons

At the single-neuron level, our modeling effort concerns the integrative properties of neurons of the thalamocortical system. A first theme consists of building computational models to understand the genesis of complex intrinsic neuronal properties, such as bursting, adaptation or endogenous oscillations in thalamic and cortical neurons. In the case of thalamic neurons, the rebound-bursting properties depend on the IT current, which is localized in dendrites. This dendritic localization was shown by a combination of electrophysiological recordings and computational models of thalamic neurons. The dendritic localization confers very particular properties to thalamic neurons, in particular concerning the integration of excitatory and inhibitory signals. A theme pursued presently is the genesis of intrinsic persistent activity by prefrontal cortex neurons, based on the presence of the Ih current.

Presently, one of our principal effort is to model stochastic states of cortical neurons. The presence of synaptic "noise" in vivo confers particular integrative properties to the cell. Synaptic noise sets the membrane in a "high-conductance" state, and the properties of cortical neurons seem fundamentally different in such states. A combination of electrophysiological recordings and computational models were used to characterize such high-conductance states, and deduce their computational consequences. It was shown that neurons have an enhanced responsiveness; they also follow a stochastic dynamics of dendritic integration, in which synaptic inputs have equivalent effects ("synaptic weight"), independent of their position in dendrites. Neurons also have finer temporal discrimination abilities in high-conductance states. Some of these predictions were tested by the dynamic-clamp technique, which consists of re-creating high-conductance dynamics in neurons maintained in vitro by injecting computer-generated synaptic noise. This approach is presently pursued intensely in collaboration with Thierry Bal's group.

3. Neuronal networks

At the network level, we aim at understanding collective phenomena in neuronal populations, which in many cases, cannot be described simply from the properties of individual neurons. In cortex and thalamus, neurons are characterized by complex intrinsic properties and responsiveness (see above) and they interact via different types of interactions in which different types of synaptic receptors are implicated. Such networks have a high structural complexity and their dynamics is highly irregular. Computational models are needed to understand the genesis of such states, and the properties of such complex states in information processing. This theme is studied either with numerical simulations, or using the technology of neuronal networks on integrated circuits (in collaboration with S. Renaud, Bordeaux and K. Meier, Heidelberg).

The simplest dynamical state is the periodic oscillation, and the numerous types of neuronal or EEG oscillations have been thoroughfully characterized since the beginning of XXth century. In collaboration with the groups of Steriade (Canada) and Sejnowski (USA), we have studied the genesis of several types of oscillations in the thalamocortical system, such as the spindle oscillations (during sleep) and the "spike-and-wave" oscillation during absence epileptic seizures (petit-mal). In particular, we have built a coherent theoretical framework which accounts for in vivo and in vitro experiments (often contradictory!) about these oscillations. This theoretical framework can also account for the large-scale synchrony of spindle oscillations in cortex, and for their transformation into pathological states such as absence seizures. A strong point of this research has been the experimental confirmation (by Thierry Bal's group) of a model prediction by using the dynamic-clamp technique.

At the other side of the complexity scale are the complex and irregular dynamical states which constitute cortical activity during aroused states (awake animals). The approach followed by our group was first to model the genesis of such "active" states using networks of excitatory and inhibitory cortical neurons. Next, with Michelle Rudolph-Lilith, we investigate the link between structural complexity (connectivity) and dynamical complexity (activity). Finally, we attempt to understand how sensory information is processed in such active states, despite their highly complex and apparently random character. This research theme is pursued presently with Sami ElBoustani by using techniques derived from fluid dynamics and information theory, in collaboration with the laboratory of Diego Contreras (USA).

4. Analysis methods

Finally, an important aspect of computational neuroscience is to provide methods to analyze experimental data based on biophysical representations of neurons. A first approach has been to conceive methods to analyze multi-electrode recording data, which often show very complex signals difficult to analyze. The spatiotemporal analysis of such recordings allowed us to demonstrate that the dynamics of "UP states" during slow-wave sleep is quasi-identical to that during wakefulness, suggesting that wake activity is "replayed" during sleep.

More recently, with Michelle Rudolph-Lilith and Martin Pospischil, we have conceived different techniques to analyze another type of complex signal: the synaptic "noise" present in recordings performed in vivo. From a stochastic description of synaptic noise (using the Fokker-Planck formalism), we have obtained a series of analytic expressions for experimentally measurable quantities, such as the membrane potential distribution or the reverse correlation between spike and membrane potential. These expressions allow us to directly estimate, from intracellular data, the synaptic conductances received by the neuron. This type of application is presently in progress in collaboration with different laboratories at the UNIC and USA.

Finally, in collaboration with Romain Brette and Thierry Bal, we have conceived a high-resolution recording method, which principle is to compensate "on-line" for electrode artefacts (Active Electrode Compensation, AEC). Using a linear model of the electrode, it is possible to determine the contribution of the electrode and subtract it from the recording, which leads to recordings with a resolution much higher compared to traditional recordings. Numerous applications are in progress, such as the conception of a high-resolution single-electrode voltage-clamp, high-resolution dynamic-clamp, and also the direct estimation of conductances in vivo (in collaboration with different groups at UNIC).

Selected Publications

Jakob Wolfart, Damien Debay, Gwendal Le Masson, Alain Destexhe and Thierry Bal, Synaptic background activity controls spike transfer from thalamus to cortex, Nat Neurosci 8: 1760-7, (2005) [pdf] [abstract]

Luc Estebanez, Sami El Boustani, Alain Destexhe and Daniel Shulz, Correlated input reveals coexisting coding schemes in a sensory cortex, Nature Neuroscience 15 No12: 1691-1699, (2012) [pdf]

Alain Destexhe and Diego Contreras, Neuronal computations with stochastic network states, Science 314: 85-90, (2006) [pdf] [abstract]

Lyle Muller, Alexandre Reynaud, Frédéric Chavane and Alain Destexhe, The stimulus-evoked population response in visual cortex of awake monkey is a propagating wave, Nature Communications 5: 3675, (2014) [pdf] [abstract] [PubMed]

Alain Destexhe, Stuart W. Hughes, Michelle Rudolph and Vincenzo Crunelli, Are corticothalamic 'up' states fragments of wakefulness, Trends Neurosci 30: 334-42, (2007) [pdf] [abstract]