Skip to main content
European Commission logo print header

Optogenetic decomposition of inhibitory micro-circuits in the mouse V1 and S1

Final Report Summary - INTERNEURONS (Optogenetic decomposition of inhibitory micro-circuits in the mouse V1 and S1)

Executive summary

The brain is capable of complex computations that are fundamental to our everyday lives. These computations are composed of fundamental operations that have been the focus of many research in the past few years. In particular, two basic arithmetic operations, namely division and subtraction have been reported in a wide range of experimental conditions. Neuronal response division takes place when the function performed by a neuron is scaled down by a constant factor without affecting its overall shape. This form of gain modulation has been reported during directed visual attention, contrast-invariant selectivity, multisensory integration and value estimation. On the other hand, subtraction acts on neuron responses by homogeneously reducing the firing pattern of a cell regardless of the sensory stimulus. As a result, the representation that this neuron holds from the world becomes more selective to a specific feature. Because it influences so dramatically the selectivity of individual neurons, response subtraction can play a key role in improving the perceptual discriminability of sensory systems. It has been suggested that different cortical inhibitory cell classes provide distinct combinations of divisive or subtractive inhibition during sensory stimulation of cortical networks.

Although glutamatergic neurons compose a majority of cortical neurons, GABAergic interneurons play a key role in shaping sensory processing and plasticity of the mature and developing brain. Interneurons can be divided into classes that differ in molecular, morphological and functional properties. Recent development of new mouse lines have allowed us to systematically study genetically defined cell-types. In particular, considerable attention has been drawn to parvalbumin-expressing (PV) and somatostatin-expressing (SOM) interneurons, the two key cell-types that target pyramidal neurons. Several studies have been performed in the past few years trying the dissect out the contribution of these neurons to the local computation performed by principal cells in the mouse visual cortex. Based on the design and the outcome of these experiments, they attributed immutable functions to these interneurons by showing that they inhibit the response of their targets through either of the two fundamental arithmetic operations: subtraction or division. However opposite conclusions were reached leaving the problem partly unresolved. More importantly no mechanism was suggested that would explain how these neurons inherited these particular properties.

During his work in Mriganka Sur’s laboratory the researcher addressed this issue by using a combination of genetically engineered mouse lines designed to manupilate PV or SOM interneurons and highly advanced imaging and optogenetic tools to manipulate precisely the activity of these neurons while recording the response of their targets. By using precisely timed optogenetic manipulation in combination with two-photon calcium imaging or targeted electrophysiology during presentation of various stimulus ensemble, the researcher has been able to demonstrate that these interneurons display very contrasting response modes depending on the sensory context. In particular, SOM neurons are very sensitive to large visual stimuli whereas PV neurons tend to be very responsive to small stimuli comparable to their target neurons. We hypothesize that these different response modes would confer them with the ability of influence principal neurons in different ways. We showed experimentally and with computational models that the function of these interneurons adapts to the stimulus ensemble used to characterize them. In particular, SOM interneurons effect on target cells was found to be divisive when large stimuli were displayed and became subtractive for small stimulus sizes. This more general description of interneuron functions was used to reconcile divergent previous results and demonstrate how the function of parvalbumin-expressing and somatostatin-expressing neurons in mice in vivo is governed by the overlap of response timing between these neurons and their targets.

Summary description of project context and objectives

The project entitled “INTERNEURONS” (n° 274920) has started on February the 1st 2012 at MIT (Cambridge, USA) under the supervision of Prof. Mriganka Sur (Picower Institute for Learning and Memory, MIT). During the first two years of the outgoing phase (2012-2014), the fellow researcher, Dr. Sami El-Boustani, has been performing experiments in order to achieve the goals established in the grant agreement. After this time period the researcher requested an additional year (2014-2015) during which funding would be suspended in order to finish the remaining experiments and publish his result in a high-profile journal. The return phase of the project started on May the 1st 2015 in the laboratory of Carl Petersen at l’Ecole Polytechnique Fédérale de Lausanne and ended in May the 1st 2016. During the whole period of the project – outgoing and return phases – the researcher has had access to state-of-the art techniques and excellent work conditions to execute the project. According to the list of objectives established in the grant agreement, the fellow had to be trained to in vivo experiments in the mouse primary visual and somatosensory cortices involving 2photon calcium imaging, targeted electrophysiological recordings as well as the use of optogenetics tools such as light-gated activation of specific interneurons cell-types through channelrhodopsin. Since the techniques and results have evolved dramatically over the course of the past few years, the researcher has also had the opportunity to develop advanced skills beyond the scope of the initial objectives. These include the development of sensory-based behaviour paradigms for awake mice, the use of gene delivery techniques such as single-cell electroporation to study the precise molecular events underlying single neurons in vivo and the systematic exploration of neural dynamics across layers of cortical areas involved in somatosensory discrimination. The training as well as the experiments that have been performed subsequently were done under protocols approved by MIT’s Animal Care and Use Committee and conforming to NIH guidelines as well as done in accordance with the Swiss Federal Veterinary Office. The different objectives as they have been redefined during the course of the project are described below and their achievements are discussed in detail in the next section:

Research Objective 1: Characterization of V1 interneurons functional properties
During the first part of the project, we wanted to gain a better understanding of the various functional properties that specific interneuron subsets display in different sensory contexts. In particular, our attention was focused on parvalbumin-expressing and somatostatin-expressing interneurons, two cell-types that could not be distinguished before the recent development on Cre mice lines. These two cell types differ in the morphology, electrophysiological properties and gene expression profile. Because of these differences we have hypothesized that they should be involved in distinct neural computations. More precisely, we posit that soma- and axon-targeting PV+ interneurons should display functional properties comparable to nearby principal cells that could potentially confer them with ability to divide the response of their target. On the other hand, the radially-organized, dendrite-targeting SOM+ interneurons would display large receptive field with weak responses significantly delayed compared to nearby pyramidal cells. These functional properties will be explored using two contrasting stimulus ensemble, namely a sparse noise sequence of small square randomly displayed in various location of the visual field or full-field flashes covering the whole visual field with different intensity contrast. Computational models of recurrent neuron networks would then be designed to reproduce the variety of sensory responses found in different cell-types.

Research Objective 2: Decomposition of V1 interneuron micro-circuits
Once the functional properties of the various cell-types have been described, we will use state-of-the-art genetically-encoded tools to disentangle the precise role the these interneurons. By using the light-gated ion channel channelrhodopsin-2 (ChR2) delivered through adeno-associated viral vectors, we will manipulate specific interneurons populations. In order to express these genes specifically in these interneurons sub-types, we will obtain conditional gene expression by transfecting neurons with a Cre-dependent plasmid that will act through the use of CreloxP system in specific Cre driver lines. Experiments will then be performed to characterize the response of principal cells during precise and reversible alteration of PV+ or SOM+ activity using blue light stimulation of ChR2. According to our main hypothesis, the nature of the operation performed by PV+ or SOM+ interneurons will depend on their response mode evoked by the stimulus ensemble used to characterize their effect on target cells. In particular, if a given cell-type display different response mode for different stimulus ensemble, we expect them to display context-dependent neural computation. These results will be used to expand our network model and try to further our understanding of the key mechanism that underlie these functions. Our model should then predict new network behaviour that could be tested with further experiments. In particular, we hypothesize that PV+ interneurons and pyramidal cells properties are more alike indicating that both cell-types receive similar direct input from the thalamo-recipient layer IV. On the contrary, SOM+ interneurons would lack direct feed-forward input but instead provide strong feedback inhibition on the local circuit to account for their delay and weak responses. Together with Research Objective 1, these results would provide a rich description of the function of specific interneurons in V1 micro-circuits.

Research Objective 3: Origin of divisive and subtractive inhibition in mouse V1
Through the experimental and computational work done for objectives 1 and 2, we should then be able to describe canonical rules that govern the emergence of divisive and subtractive inhibition provided by specific interneurons. Following our hypothesis that specific response modes are responsible for versatile impact on the network. We will try to dissect out the specific characteristics of these modes to understand the conditions in which division or subtraction are observed. In particular, we expect that during sparse noise presentation the late response of SOM+ interneurons with respect to principal cells will be suited for providing a non-specific inhibitory influence on their target that would result in subtractive inhibition. On the other hand, PV+ interneurons that closely follow the response dynamics of pyramidal cells would be able to provide an inhibitory contribution proportional to the evoked activity therefore generating divisive effect. This logic could then be tested in our theoretical model but also experimentally to try to unify previously published data into a comprehensive framework. In order to do so, we will use similar stimulus ensembles used in previous studies and confirm our mechanisms by activating PV+ and SOM+ interneurons at key moments when their responses modes are identical or divergent. We expect to describe the diversity of reported results in the literature.

Research Objective 4: Study of somatosensory cortical micro-circuits in the awake behaving mouse
In the return phase of the project, we will study the contribution of different interneuron subtypes to the transformation of tactile sensory input into decisional signals. In order to do so, we will design behaviour paradigm in which water-restricted mice will have to learn to discriminate different tactile stimuli in order to get reward with water. We will use large population calcium imaging of identified cell-types to dissect out the contribution of each interneurons and pyramidal cells to the performance of the task. In order to obtain a complete description of the cortical columns dynamics, we will use micro-prism inserted into the cortex to have access to all layers of whisker-related functional column simultaneously. Specific interneurons cell-type would then be identified using reporter mouse lines expressing fluorescent tag. We hypothesize that the performance of the mouse during the task will correlate with information transmission among principal cells that this process will be modulated in a critical way by the local inhibitory circuits. In particular, we expect SOM+ interneurons to strongly modulate the selectivity of pyramidal cells by providing subtractive inhibition through top-down neuro-modulatory inputs whereas PV+ would help linearize the response properties of nearby pyramidal cells by tightly balancing their excitatory synaptic input through feedforward inhibition. By systematically imaging these interneurons during the learning of the task (from naive to expert performance) we will be able to identify how these cell-types help shaping sensory representation of tactile scenes to improve discriminability. These results will then be recapitulated in a computation model of sensory-motor cortices that would reproduce the behaviour and neuronal data.