Neuronal networks and physiopathological rhythms

Brain function is believed to arise from an extensive interconnectivity between the different regions of the central nervous system. In this complex process of information exchange and integration, two structures in the diencephalon, the thalamus and the epithalamus, together act as relay and integration hubs for virtually the entire nervous system. Our team is seeking to understand the functioning of these structures from the molecular level (ion channel, neurotransmitter receptor, ...) to behavior (control of states of alertness, sensory perception, fear, ...) and associated pathologies (absence epilepsy, neuropathy, anxiety, aversion, ...).

We use different animal models and viral transduction strategies to dissect the neural subnetworks specifically involved in thalamus and epithalamus function, determine their functional properties using in vitro and in vivo electrophysiological techniques, manipulate these subnetworks using optogenetic approaches and evaluate the behavioral consequences.

Selected publications ...

Otsu Y. et al. Control of aversion by glycine-gated GluN1/GluN3A NMDA receptors in the adult medial habenula. Science. 2019

Quiquempoix, M. et al.Layer 2/3 pyramidal neurons control the gain of the cortical output. Cell Report. 2018

Grand T, et al. Unmasking GluN1/GluN3A excitatory glycine NMDA receptors. Nat Commun. 2018

Crunelli V. et al. Dual function of thalamic low-vigilance state oscillations: rhythm-regulation and plasticity. Nat Rev Neurosci. 2018

McCafferty C., et al. Cortical drive and thalamic feed-forward inhibition control thalamic output synchrony during absence seizures. Nature Neurosc. 2018

Otsu Y, et al.,Functional Principles of Posterior Septal Inputs to the Medial Habenula. Cell Rep. 2018

Recent achievements

Control of aversion by glycine-gated GluN1/GluN3A NMDA receptors in the adult medial habenula

This project aimed to examine the physiological role of an atypical NMDA receptor (NMDARs) subunit, the GluN3A subunit.

NMDARs generally consist of two distinct groups of subunits, the GluN1 subunits, which bind glycine, and the GluN2 subunits, which bind glutamate. Binding between glycine and GluN1 subunits is a necessary, but not sufficient, condition for activation of these receptors. However, glycine also binds to an unconventional and poorly characterized NMDA subunit, GluN3A, whose expression had been observed only transiently during the early stages of development. The association of GluN1 with GluN3A gives rise to excitatory heterodimeric receptors (Figure a.) which have the remarkable feature of being exclusively activated by glycine. Until now, these heterodimers have only been detected in cultured cells, and thus have been considered as artefacts of expression systems.

Using multiple experimental approaches including morphological, ex vivo electrophysiological and in vivo behavioral analyses, we were able to demonstrate that GluN3A subunits are expressed in specific thalamic and epithalamic regions in adult mice (figure b.). Furthermore, we have established that glycine activation of GluN1/GluN3A heterodimers in the Medial Habenula (MHb) increases neuronal excitability (figure c. & d.), confirming the excitatory role of this atypical glycinergic receptor, and that their function in MHb is crucial for complex behaviors, such as place aversion (figure e.).

Otsu Y, et al. Science. 2019;366(6462):250-254. doi: 10.1126/science.aax1522.

Layer 2/3 Pyramidal Neurons Control the Gain of Cortical Output

This project aimed to determine the role of the laminar organization of the primary cortex in the processing of somatosensory information. Indeed, the canonical model suggests that sensory information is processed serially by the different cortical layers. It would first be pre-processed in layer 4 (L4), receiving information from the thalamus, then sent to layers 2/3 (L2/3) and finally propagated to the neurons of layers 5 (L5) which innervate the sub-cortical structures (basal ganglia, collicles, spinal cord, ...). However, information from the thalamus also directly enervates the L5 neurons, calling into doubt the relevance of the canonical model of serial processing and raising the question of the respective roles of the different thalamic entries and the cortical layers (figure a.).

Using an in utero electroporation approach in mice we selectively expressed opsins in the L2/3 excitatory neurons of the barrel cortex (figure b.) in order to specifically and reversibly manipulate these neurons during vibrational stimulation. In vivo recordings of anaesthetized and awake mice (figure c.) revealed that serial processing of thalamic information, L4->L2/3->L5, modulated the gain of the response of L5 neurons to somatosensory excitations and thus of the cortical output (figure d. & e.).

Quiquempoix M et al. Cell Rep. 2018; 24(11):2799-2807.e4. doi:10.1016/j.celrep.2018.08.038.

Cortical drive and thalamic feed-forward inhibition control thalamic output synchrony during absence seizures

The dynamic interaction of the different structures of the thalamocortical loop is at the basis of the pathological rhythmic activities of absence epilepsy in children. This non-convulsive epilepsy is characterized by the frequent repetition of brief episodes of unconsciousness associated with the appearance on the EEG of a specific oscillation, the peak-wave discharges (figure a.).

The aim of this project was to evaluate the relevance of the mechanisms suggested by previous in vitro or anaesthetized animal studies. These hypotheses, based on the thalamocortical dynamics demonstrated for the genesis of sleep spindles, proposed that the appearance of peak-wave discharges is based on the genesis of bursts of action potentials by thalamocortical excitatory neurons (TC) and inhibitory nucleus reticularis

thalami (NRT) neurons, subtended by the recruitment of low-threshold calcium channels (T-type).

We analyzed the dynamics of the thalamocortical network during peak-wave discharges in free-moving animals. The activity of cortical and thalamic neurons was recorded using silicone probe electrodes in a rodent model of epilepsy absence, Genetic Absence Epilepsy Rats from Strasbourg (GAERS). The involvement of T-type calcium channels was evaluated by dialysis of a specific antagonist, TTA-P2, locally.

Analysis of the activity of neurons recorded simultaneously in the different regions of the thalamocortical loop indicates that, during peak-wave discharge, cortical neurons excite the thalamocortical neurons - which then discharge preferentially in a tonic manner and not by bursts of action potentials - and the NRT neurons which discharge by bursts of high-frequency action potentials (figure a.). These bursts of action potentials in the NRT will then sculpt and synchronize the tonic discharge of TC neurons by a feed-forward type inhibition mechanism (figure b.). Local microdialysis of TTA-P2 shows that T-type calcium channels play a primary role in the NRT and cortex but not in TC neurons (figure c.).

This work shows that the cellular mechanism of absence seizures differs from that of sleep spindles, and that the intrinsic excitability properties of TC neurons contribute little to the hyperexcitability of the thalamocortical loop at the origin of peak-wave discharges.

McCafferty et al. Nat Neurosci. 2018; 21(5):744-756. doi: 10.1038/s41593-018-0130-4.

Reconstructing the functional connectivity of multiple spike trains using Hawkes models

This project was initiated with the goal of developing new analytical tools and computational models that will allow us to better understand the activity neuronal we observe.

The simultaneous recording of the unit activity of many neurons in the anesthetized and awake animals theoretically opens the possibility of determining the functional interactions between these neurons and how they evolve during a task. Commonly used approaches are based on histograms treating neurons in pairs and can easily lead to imprecise functional connectivity graphs due to common inputs to several neurons or chains of connections in registered networks. In order to avoid these errors, more recent methods aim to model all recorded activity simultaneously.

In this project we have set up and tested on an Integrate and Fire neural network model, the limits of a new statistical method developed by our mathematician collaborators that can be routinely used on the experimenters' laptops. This method, which models neuronal activities in the form of Hawkes' processes (figure a.), is fast and does not require prior knowledge or hypothesis about the recorded network. The procedure reconstructs a connectivity graph (figure b.) which is provided with a complete estimation process that can be simulated to reproduce realistic data sets.

Lambert RC et al. J Neurosci Methods. 2018; 297:9-21. doi: 10.1016/j.jneumeth.2017.12.026


  • V. Crunelli, Cardiff School of Biosciences UK.
  • E. Bourinet, Institut de Génomique Fonctionnelle, CNRS UMR5203 - INSERM U661 - Univ. Montpellier I & II.
  • L. Ascady, Department of Cellular and Network Neurobiology, Institute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, Hungary.
  • P. Reynaud-Bourret, Laboratoire JA Dieudonné – CNRS UMR7351 – Univ. Nice-Sophia Antipolis
  • B. Kieffer, Institute of Genetics and Molecular and Cellular Biology (IGBMC), University of Strasbourg.
  • V. Rivoirard, CEREMADE- CNRS UMR7534 – Univ. Paris Dauphine
  • M. Mameli, Department of Fundamental Neuroscience, University of Lausanne, Switzerland.
  • F. Matyas, Institute of Cognitive Neuroscience and Psychology, Research Center for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary.
  • P. Paoletti, Institute of Biology - Ecole Normale Superieure (IBENS), Paris.
  • M. Soiza-Reilly, CONICET Institute of Physiology, Molecular Biology and Neurosciences (IFIBYNE) Buenos Aires, Argentina.