Laboratoire Matière et Systèmes Complexes

Considering the architecture of the most widespread information processing machine -the computer- one is led in a first step to tackle neuronal computation as a set of fundamental computing units assembled in a convenient way. The basic unit we deal with is a population of some hundreds of neurons grown on a two dimensional chip.

Designing neuronal devices : Microfluidics techniques allow to design a very broad variety of devices, each population being seeded in a well, the wells being linked by two-ways or diode micro-channels allowing neurites to grow. Neuronal activity can propagate through axons grown in micro-channels.

Controlling neuronal activity : The field of optogenetics has developed methods to control neuronal activity by transducing genes coding for light-gated ion channels into neurons.

Recording neuronal activity : Since action potentials in neurons go along with a calcium influx, it is possible to indirectly measure the activity of neurons by calcium imaging using fluorescence microscopy.

We recently demonstrated that the combination of optogenetics, calcium imaging and microfluidics provides a very powerful and flexible fully optical tool to investigate neuronal activity and functional connectivity of neuronal devices 3. Besides the fundamental goal of understanding and manipulating neuronal computation, it should be a milestone towards technological applications of great importance, such as neuronal implants able to repair some neuronal damages in vivo.

Spike-Timing-Dependent Plasticity Since the signaling between neurons involves the release of neuro-transmitters through synapses the (actual) functional connectivity must be distinguished from the topological one : The use of drugs which block neuro-receptors, thus able to weaken the connections, is a powerful tool in controlling the functional connectivity in vitro ; such an effect is usually interpreted as a decrease of the so-called synaptic strength. Moreover, variations in the synaptic strength lie at the heart of one of the most fundamental capabilities of the brain : the ability to learn (and also to forget) over a wide range of time scales. The mechanism we are interested in is Spike Time Dependent Plasticity (STDP) : Basically, the relative synaptic strength between two neurons is increased if the presynaptic spike fired before the post-synaptic one and is decreased otherwise ; the convenient repetition of time-correlated pre and post synaptic spikes will thus ensure learning. Such a mechanism is known to work in the case of two neurons, but has not been thoroughly investigated between two random populations in vitro connected by micro-channels, and is of fundamental interest in designing optical devices. Optogenetics allows to control the temporal correlation of the excitation between two chambers while calcium imaging enables to record their response. Preliminary simulations carried out in the framework of the Izhikevich model between two random populations were conclusive. The experiments carried out by R. Renault at the Weizmann Institute under confocal microscopy revealed a very great variability : A huge panel of parameters and learning protocols has to be explored in order to conclude in a clear way. For that purpose, R. Renault conceived a low cost and scalable neuronal chip reader including a custom optical system and a fast detection system based upon photodiodes which will enable to parallelize experiments.