In Human, brain asymmetry is an established functional feature that underlies the perception of speech and music. Whereas the left auditory cortex is specialized in processing fast temporal components of speech sounds, the right is more sensitive to spectral modulations. However, circuit features and neural computations behind these lateralized spectrotemporal processes are poorly understood. To answer these mechanistic questions and to determine whether this is unique to Human, we will use mice and probe decoding ability of left and right auditory neurons to perceive temporally and spectrally modulated vocalizations. We will combine large-scale electrophysiological recordings with functional 2-photon imaging to shed light onto the neuronal circuits that drive lateralization of auditory functions. More generally, it will help us to understand how the two hemispheres can share labor to better compute and analyze the sensory world and how this specialization has emerged in vertebrates.

Although the brain displays a symmetric shape, some functions are specialized to a dedicated hemisphere. In Humans, Paul Broca demonstrated the implication of the left hemisphere for language processing 150 years ago. This feature was first thought to be unique to Humans but growing evidence are challenging this view. In the mouse model, there is behavioral evidence that animals use rather their left auditory cortex to process vocalizations while the right counterpart is preferentially used for more general computations such as the detection of sweep direction. Structural studies have found a morphological signature of the asymmetry in the synaptic organization of the two auditory cortices, which was absent in thalamic input to the cortex. These studies suggest that there is some specialization like in Humans where the left auditory cortex computes analysis with high temporal resolution whereas the right auditory cortex is specialized for spectral analysis of sounds. However, this functional asymmetry of the auditory cortex has never been systematically tested with controlled stimuli.

To test whether there is a lateralization of spectrotemporal functions in mice, we will use a framework developed for Human psychoacoustic studies to filter natural mouse vocalizations in the temporal and/or spectral domains. In the team “Neural Coding of the Auditory System” at the Institut de l’Audition, we will record neuronal activity using large-scale microelectrode arrays (Neuropixels) in anesthetized animals at 2 stages along the auditory pathway: the inferior colliculus, where lateralization is supposed to be absent, and in the auditory cortex, where lateralization should emerge. We hypothesize that cortical neurons will be differentially sensitive to temporal or spectral filtering depending on their hemispheric origins. We will assess decoding performances at the level of individual neurons and of the population using statistical model-based methods and machine learning to demonstrate any functional asymmetry of the auditory cortex.

However, several unanswered questions call for a more mechanistic description. Specifically, where does the lateralization emerge in the cortex? Are there specific neuron types that are shaping the functional properties of each cortex? How is this information used by the animal to drive behavior? To tackle these questions, we will combine structural and functional descriptions of the processes at play in the emergence of cortical asymmetry. We will use 2-photon microscopy to record and modulate neuronal activity in different layers of the cortex and in different neuronal types. First, 2-photon imaging will uncover the specific neurons that encode spectral and temporal modulations differentially in the two hemispheres. We will image across cortical layers to identify whether specialization arise already at the thalamic-recipient layer IV or later during the cortical processing. Then, we will make use of 2-photon stimulation to inactivate specifically these neurons and study how it affects asymmetric processing of auditory information both at the neuronal and behavioral levels. We will used mice trained to discriminate between frequency or temporal sound modulations and perturb the identified circuit using targeted photostimulation. By combining viral strategies to express a fluorescent reporter in specific neuronal populations with 2-photon imaging and photostimulation, we will thus be able to describe the neuronal network underpinning cortical lateralization.

The student will learn techniques from 2-photon microscopy, electrophysiology, animal behavior, and data analysis. We are looking for individuals who want to study neural coding in the context of hearing and who meet the following criteria:
1) Master’s degree in neuroscience, biophysics, engineering, or related field
2) Experience in one the following areas would be appreciated: in vivo electrophysiology, animal behavior, optogenetics, microscopy.

Depression is one of the most common mood disorders, affecting 1/3 adults. The proposed project aims to
study the network and circuit level changes underlying the manifestation of depression and the network-level
mechanisms of action of novel atypical and effective psychedelic antidepressants, such as psilocybin.
Towards this goal, we will use large-scale electrophysiological recordings and neuromodulator monitoring in
the limbic circuit of behaving mice, to understand the action of these drugs at the at the level of network
plasticity and functional connectivity of brain regions. The results of this project will be crucial for the
mechanistic understanding of the manifestation of major depression and will catalyze the development of
future projects in this direction. This is a core step in the attempt to identify novel pharmacological approaches
and specific alternative interventions.

Project background and description
Psychiatric diseases, such as depression, manifest in a host of symptoms. The precise underlying causes behind the disease are largely unknown, but it is quite clear that broad network effects are responsible for their cumulative symptomatology. Understanding the changes in the brain network activity during disease and the network effects of treatments is a crucial step for the design of novel effective and specific medications.
Toward this goal, it is imperative to study the mechanisms that underlie the effects of existing effective medications at the level of specific neuronal circuits. A prominent example of such medications is psychedelic antidepressants, such as psilocybin. Psilocybin is an agonist of multiple 5-HT receptors that is characterized by powerful effects on neuronal activity. In contrast to typical antidepressants such as selective serotonin reuptake inhibitors (SSRIs), psilocybin exhibits a rapid and long-lasting amelioration of depression symptoms, starting within hours following a single infusion.
Two of the core systems that are believed to be implicated in both the pathophysiology of depression and the effect of these drugs are the limbic system and the neuromodulatory system. Neuromodulators, such as serotonin, and dopamine, are fundamental building blocks of emotional states and innervate strongly the limbic circuit. Yet, the interplay between the neuromodulatory system and the limbic circuit remains little understood, while the changes in the activity of these systems in depression and following drug delivery remain unknown.
Hypothesis: To fill this major gap in our understanding, we will study the changes in the neuromodulation and the activity of the limbic circuit as well as the acute and protracted changes in neuronal patterns of activity after the administration of psilocybin in naïve and chronically stressed mice.
To achieve that we will combine state-of-the-art neurotechnologies that enable the monitoring of neuromodulator release and neuronal activity, with a behavioral paradigm that we have established for the repeated probing of the hedonic and aversive response and the monitoring of the acute and long-term effects of drug administration. For this, naïve and chronically stressed mice will be trained to collect sucrose rewards, while threatening stimuli (crescendo sounds and bright illumination) will be intermittently presented. This paradigm will enable us to probe the neuronal processing of hedonic and threatening stimuli before and after treatment, the imbalance of which is directly linked to the behavioral phenotype of the disease.
Specifically, we will address the following questions:
1. What are the acute and long-term changes in the neuromodulator release in the amygdala and the medial prefrontal cortex, following the administration of psilocybin?
2. How do psychedelics influence the network activity and functional connectivity of the circuit?
3. Can we causally establish the link between the psychedelic effects on neuromodulator release and the ensuing network changes in the activity of the limbic circuit?
We hypothesize that chronic stress will result in marked changes in the baseline neuromodulatory landscape in the medial prefrontal cortex (mPFC) and the basolateral amygdala (BLA) of behaving mice, as well as changes in the functional connectivity and the neuronal responses of these regions to reward and threat. We expect that psychedelic antidepressants will reverse these neuromodulatory and neuronal changes and will correspondingly reset the limbic network phenotype of stressed mice to the baseline conditions.
Aims and Experimental Plan
Aim 1. Characterize the drug-induced changes in the neuromodulatory landscape of the limbic circuit
To elucidate the effect of psilocybin on the neuromodulator release in the basolateral amygdala (BLA) and the medial prefrontal cortex (mPFC), we will use a novel experimental strategy that we have developed termed “Combinatorial Neuromodulation Fiber Photometry” (CNMFP) to simultaneously record the release of all major neuromodulators (DA, 5-HT, NE, and ACh) in the mPFC and HPC of behaving mice. To record the release of each neuromodulator, we express novel genetically encoded biosensors that report the concentration and dynamics of neurotransmitter release. A cocktail of multi-color indicators, paired with in vivo multi-region fiber photometry and analytical techniques, enables us to record from all four major NMs simultaneously in both regions in behaving mice.
Aim 2. Investigate the drug-induced changes in the functional connectivity of the limbic circuit
To investigate the effects of psychedelic antidepressants on the network dynamics and the information flow between the mPFC and the BLA, we will utilize in vivo electrophysiological recordings using high-density silicon probes (such as Neuropixels). This fine-resolution sampling of hundreds of neurons in each region will enable us to investigate the changes in the functional connectivity between the mPFC and the BLA and the neuronal responses to reward and threat. These longitudinal recordings will allow us to track neuronal activity throughout the development of depressive symptoms following the chronic stress paradigm and the amelioration of the symptoms following the administration of psilocybin.
Aim 3. Establish the causal effects of neuromodulation for the drug effects on network activity
Using the approaches in Aim 1 and Aim 2, we will be able to correlate the changes in the neuromodulator release in the mPFC and the BLA with the changes in the neuronal activity and functional connectivity of the circuit at different time points of the experimental paradigm and the pharmacological treatment. To establish the causal relations between the change in the neuromodulatory landscape and the changes in the neuronal responses and network activity, we will use targeted optogenetic inhibition or excitation of the projections from neuromodulatory centers (e.g., the 5-HT-producing neurons of the dorsal raphe) to each of the target regions. These manipulations, achieved using high-efficiency axon-targeted opsins, combined with electrophysiological recordings, as in Aim 2, will enable us to establish the necessity and sufficiency of specific neuromodulatory signaling in each region for the observed network changes during chronic stress and its amelioration following treatment.
Own work in context (thesis supervisor): In my previous work, I studied the mechanisms allowing the long-range network synchronization of neuronal activity in the limbic circuit and their coordination with bodily rhythms. Over the past years, I have carved a conceptual and technical framework that enables the large-scale and high-throughput investigation of the neuronal correlates of behavioral states and the elaboration of the circuit mechanisms that define and modulate the corresponding brain states. This framework focuses on the synchronization of neuronal ensemble dynamics and its determinants at the level of neuromodulators and receptors.
This project is designed based on strong preliminary data and my extensive expertise in all the techniques involved. This preparatory work guarantees the technical feasibility of the proposed work.
Expected outcome and Impact: The successful implementation of this project will provide invaluable data for the effect of psilocybin on the release of neuromodulators in the mPFC and the BLA and the corresponding changes in the activity and network synchronization of these regions. Importantly, we will identify the causal links between these two subsystems and the contribution of neuromodulators in the circuit organization, which potentially underlies the effectiveness of these drugs for the treatment of depression. This is a core step in the attempt to identify novel pharmacological approaches and specific alternative interventions.

Astrocytes in the brain constitute a predominant influence for the cerebrovascular system controlling perivascular homeostasis, blood-brain barrier integrity, crosstalk with the peripheral immune system, endothelial transport, or vessel contractility in response to neuronal activity. Astroglial regulation towards the cerebrovascular system is set at the level of the gliovascular unit, a specialized interface composed by astrocyte perivascular processes that fully sheath the brain vessels. Such morphological and functional polarity is of critical importance for the brain physiology and is altered in megalencephalic leukoencephalopathy with subcortical cysts (MLC), a rare type of leukodystrophy. The goal of the present project is to decipher gliovascular alterations in this specific pathological context and further explore gliovascular interactions.

MLC1 is an astrocyte-specific membrane protein whose defect leads to megalencephalic leukoencephalopathy with subcortical cysts (MLC), a rare type of leukodystrophy characterized by an early-onset macrocephaly and a progressive white matter vacuolation leading to ataxia, spasticity, and cognitive decline. MLC1 forms a membrane complex at the junctions between astrocyte perivascular processes (PvAPs). In pilot studies, we showed that MLC1 is a marker of the PvAP postnatal maturation. We then analyzed the gliovascular unit (GVU) in a MLC mouse model deleted for MLC1 (Mlc1 KO). We determined that absence of MLC1 modifies the development of astrocytic morphology and polarity and the organization of astrocytic perivascular coverage. These changes correlated with an altered development of vascular smooth muscle cells contractility which regulates the cerebral blood flow. We also observed an altered neurovascular coupling as well an impaired cerebrospinal/interstitial parenchymal fluid circulation, which could contribute to the water accumulation and the megalencephaly observed in the mouse model and in MLC patients. This work suggested that MLC is a developmental pathology of the GVU with PvAP and cerebrovascular defects being primary events in the MLC pathogenesis.

We now propose to further study the physiopathological process of MLC. In particular, we aim to uncover the unknown links between astrocytic perivascular coverage and vascular alterations in MLC. This will be done through a multidisciplinary approach combining transcriptomic analysis of the vascular and astrocytic compartments, gene therapy to restore the expression of MLC1 in astrocytes, imaging analysis of the gliovascular interface and in vivo analysis of cerebrovascular functions such as neurovascular coupling, blood flow and brain drainage. Importantly, this project is based on validated experimental approaches of molecular biology and ex vivo and in vivo imaging as well as on an established collaborative network.

This project will be crucial to characterize the cerebrovascular pathology linked to MLC. It will allow to potentially propose novel therapeutic approaches to cure MLC. It will also highlight how astrocytes and the cerebrovascular systems work in concert to regulate brain physiology.


References of the laboratory for this project:
Alonso-Gardon, M., X. Elorza-Vidal, A. Castellanos, G. La Sala, M. Armand-Ugon, A. Gilbert, C. Di Pietro, A. Pla-Casillanis, F. Ciruela, X. Gasull, V. Nunes, A. Martinez, U. Schulte, M. Cohen-Salmon, D. Marazziti, and R. Estevez. (2021). Identification of the GlialCAM interactome: the G protein-coupled receptors GPRC5B and GPR37L1 modulate megalencephalic leukoencephalopathy proteins. Hum Mol Genet. 30:1649-1665.

Cohen-Salmon, M., L. Slaoui, N. Mazare, A. Gilbert, M. Oudart, R. Alvear-Perez, X. Elorza-Vidal, O. Chever, and A.C. Boulay. (2021). Astrocytes in the regulation of cerebrovascular functions. Glia. 69:817-841.

Gilbert, A., X. Elorza-Vidal, A. Rancillac, A. Chagnot, M. Yetim, V. Hingot, T. Deffieux, A.C. Boulay, R. Alvear-Perez, S. Cisternino, S. Martin, S. Taib, A. Gelot, V. Mignon, M. Favier, I. Brunet, X. Decleves, M. Tanter, R. Estevez, D. Vivien, B. Saubamea, and M. Cohen-Salmon. (2021). Megalencephalic leukoencephalopathy with subcortical cysts is a developmental disorder of the gliovascular unit. Elife. 10.

Gilbert, A., X.E. Vidal, R. Estevez, M. Cohen-Salmon, and A.C. Boulay. (2019). Postnatal development of the astrocyte perivascular MLC1/GlialCAM complex defines a temporal window for the gliovascular unit maturation. Brain Struct Funct. 224:1267-1278.

Slaoui, L., A. Gilbert, A. Rancillac, B. Delaunay-Piednoir, A. Chagnot, Q. Gerard, G. Letort, P. Mailly, N. Robil, A. Gelot, M. Lefebvre, M. Favier, K. Dias, L. Jourdren, L. Federici, S. Auvity, S. Cisternino, D. Vivien, M. Cohen-Salmon, and A.C. Boulay. (2023). In mice and humans, brain microvascular contractility matures postnatally. Brain Struct Funct. 228:475-492.

The proposed PhD project delves into the intricate realm of mitochondrial biology and its crucial role in maintaining overall health. Mitochondria, often referred to as the powerhouses of cells, are known for their role in producing ATP, the cellular energy currency. However, their significance extends beyond energy production, encompassing spatial compartmentalization, homeostasis regulation, and stress response orchestration.

Mitochondrial diseases (MD) are rare, complex disorders stemming from genetic mutations in approximately 400 mitochondrial genes. These disorders manifest as debilitating neuromuscular dysfunctions, posing significant challenges in understanding their underlying mechanisms. Importantly, the tissue-specific nature of MD cannot be solely attributed to ATP deficiency, necessitating a deeper exploration of mitochondrial functions.

This PhD project aims to shed light on the importance of mitochondrial membrane integrity in safeguarding against metabolic dysfunction

Background
The biological hallmarks of health have been holistically defined to include spatial compartmentalization, the maintenance of homeostasis over time, and the integration of adequate responses to stress, all of which are terms that aptly describe functional roles of mitochondria. In humans, pathogenic variants in ~400 different mitochondrial genes cause mitochondrial disease (MD), which are rare and catastrophic multi-systemic disorders that cause neuromuscular dysfunction for reasons that elude us. While the need for mitochondria is undoubtedly ubiquitous, it has become increasingly clear that ATP deficiency alone cannot explain the tissue-specific manifestation in patients nor in the animal models we use to study the etiology of MD disorders. Beyond cellular energy production, mitochondria act as signaling organelles by orchestrating multi-layered homeostatic cellular responses to a wide array of stressors. Mitochondrial membrane integrity ensures the regulated, physical sequestration of mitochondrial derived signaling molecules from the rest of the cell, such as those that trigger cell death and innate immune responses. Therefore, maintaining mitochondrial membrane integrity is critical to ensure the proper function of mitochondria. Our lab has advanced the molecular understanding of the processes that maintain the functional integrity of mitochondria using cell and animal models and have begun to pursue pharmacological approaches to therapeutically enhance mitochondrial resilience in the context of disease.

The Project
In this PhD project, we will investigate how enhancing mitochondrial membrane integrity can safeguard against metabolic dysfunction, cell death and overactive innate immune signaling in cell and animal models of MD. The research program that will apply a combination of cell biology, biochemistry, genetics, and physiology approaches at various interconnected scales.

Aim 1: At the cellular level, the PhD student will use mutant cell lines and/or patient-derived cells to carry out imaging-based phenotypic chemical screens using a combination of proprietary molecules and FDA/EMA-approved small molecules (drug repurposing), to identify compounds that can improve cristae dysmorphology and mitochondrial activity based on an already-identified genetic target (gene X). The host lab has established advanced imaging approaches to study mitochondrial form and function in living cells that can be combined by high-throughput genetic screening (Cretin et al. EMBO Mol Med 2021). The student will use acquire skills from the host lab to employ advanced imaging (STED, SIM, EM), biochemistry, and OMICS approaches to deeply phenotype mitochondrial form and function. To decipher the mode of action of candidate hits, the student will perform deconvolution studies using live-cell imaging siRNA screening approaches to identify the genes required for functional rescue of mitochondrial structure.

Aim 2: At the tissue level, the PhD student will explore the therapeutic value of selected hits using whole-body knockin mice of mitochondrial disease. The host lab has begun creating transgenic mice carrying pathogenic variants in essential cristae-maintaining proteins whose disruption causes mitochondrial and neuromuscular dysfunction in humans. The student will learn how to characterize neuromuscular dysfunction in vivo using phenotyping approaches established at the Institut Pasteur and participate in the description of this new pre-clinical model. Using this model, the PhD student will block the activity of gene X using hits identified in Aim 1 and/or using a whole-body knockout mouse for gene X with the aim of functionally rescuing defective mitochondrial structure and function and physiological function.

The lab
The Mitochondrial Biology Group host lab is located at the Institut Pasteur and is part of the Department of Cell Biology and Infection, consisting of 15 research groups dedicated to exploring fundamental questions in cell biology using various imaging techniques in tissues and cells. In this setting, the PhD student will have ample opportunities to collaborate with fellow scientists, explore new research areas, and engage in valuable networking. Additionally, the lab is associated with the French National Centre for Scientific Research (CNRS), Europe's largest fundamental science agency, providing access to diverse courses and training in various scientific and soft-skill disciplines. The Mitochondrial Biology group is a well-equipped lab space, furnished with all the necessary reagents and equipment for cell and mitochondrial biology and biochemistry. In vivo experiments will take place in the animal facility and all proposed transgenic mouse lines are readily available.
The lab has made significant contributions to our understanding of the role of mitochondria in health and disease using multi-disciplinary and multi-scale approaches. We work closely with clinicians from French reference centres for MD using whole-exome sequencing to identify new disease genes and pathogenic variants, which we then characterize and validate using mitochondrial phenotyping approaches applied directly to patient-derived cells. Using these valuable biological resources, we innovated high-throughput live-cell imaging technologies that enabled us to perform siRNA-based phenotyping screening of patient fibroblasts derived from patients suffering from neurological disorders. We identified 91 mitochondrial proteins that could act as genetic suppressors of mitochondrial fragmentation that could counteract impaired mitochondrial fusion. In addition, our laboratory also uses transgenic mouse models that develop progressive and tissue-specific defects instigated by primary mitochondrial dysfunction, with the aim of understanding disease etiology. In vivo, we apply integrative physiological approaches to delineate the pre-symptomatic/symptomatic transition to guide us in identifying the underlying candidate molecular and cellular defects responsible for triggering disease. We have applied this approach to established and novel mouse models of disease, successfully delineating the dysfunctions that are causal from those that are merely a consequence of organ dysfunction.

The Institute
The Institut Pasteur is a non-profit private foundation and one of the world’s leading research centres. To date, 10 scientists from this Institute have been awarded the Nobel Prize for Medicine and physiology. The Institut Pasteur in Paris is one of the world most productive and respected research institutions, particularly recognized in recent years for its work in medicine and biology. It houses more than 2,800 employees, 59% of whom are women (93/100 Gender Equality Index score). The Institut Pasteur brings together scientists from many disciplines Pasteur also serves as advisor to the World Health Organization (WHO) of the United Nations and has close collaboration with the U.S. Centre for Disease Control and Prevention (CDC). The Institut Pasteur has broad experience in transforming research results into applications, spinning out more than 25 start-ups since 2000. It provides state of the art infrastructure for the experimental study of mouse and cell models, leading-edge core facilities for omics (proteomics, metabolomics, transcriptomics), advanced light microscopy, nanoimaging, image analysis, and bioinformatics, experienced staff but also strong research support departments regarding the administrative and financial aspects as well as the career development.

Locomotion is essential for survival across the animal kingdom as it underlies various goal-directed behaviours, like food search, mating and avoiding danger. Environmental contexts and internal states influence will both animal behaviour by affecting the speed and type of locomotion used, for example. At the neural circuit level, this flexibility in locomotor behaviour may be achieved via neuromodulation. However, understanding the precise neural circuit mechanisms, involved neuromodulators, and their effects on specific circuits remains a challenge.
We seek to understand, experimentally and theoretically, the neural mechanism of neuromodulation and its implementation in a complete sensorimotor circuit of the Drosophila larva, which offers many advantages for neural circuit mapping.

Project

Neuromodulation

At the neural circuit level, behavioural flexibility is thought to be implemented by neuromodulation. This modulation can be done by altering the resting membrane potential of targeted neurons or by modulating the strength of existing synaptic connections. This neuromodulation can alter the dynamics of the behaviour being performed and the type of actions expressed in a given context. The outcome of the network can differ depending on the neuropeptide being released, for example, Refs1–8. Sensorimotor circuits must have evolved to allow flexible information processing and the ability of internal or external stimuli to modulate target neurons within the network to bias its output. We expect these circuits to have characteristic motifs favouring the neuromodulatory activity.
However, detailed neural circuit mechanisms underlying the state and context-dependent behaviour modulation, the neuromodulators involved, and their mechanism on sensorimotor circuits need to be deciphered.
Drosophila offers many advantages to make rapid progress in elucidating the mechanisms of neuromodulation in complete circuits1,2. First, the full neural connectome5 is now available. Genetic tools3,4 available in Drosophila combined with the rapid reproductive cycle makes it possible to manipulate both genetic and neuronal activity in a cell-type-specific manner and quickly detect changes in behaviour. Advances in optical neurophysiology allow the analysis of the physiological properties of neurons and their functional relationships. Finally, deciphering the properties and function of small neural circuits6–8, characterising the larva dynamics with dictionaries of behaviour9,10, learning11 and performing robust operant learning12 has been demonstrated in the larva.

The project is organised in 3 aims. Two aims (1 and 3 ) belong to this PhD proposal, and aim 2 will feed this PhD, but will be performed by another group of researchers.

Aims 1: Whole-body model of the larva.
We will construct a comprehensive larva model, encompassing its entire neural and muscular system. Data will inform this model on network connectivity, behaviour high, body segment resolution, and muscle pattern activity underlying different locomotor actions. A dedicated software, NYX, will facilitate unified simulations of neural and muscular activities (https://bit.ly/42SR0RU). This aim also investigates the role of neurons at different stages of sensorimotor processing in controlling motor behaviours. Such explorations will evaluate these neurons' influence on larval actions like hunching, turning and crawling. François Laurent and another engineer from INRIA will notably contribute to this aim. In this aim, a PhD student from the Jovanic team, Amit Hasan, will contribute (motor and muscle pattern analysis). Hence, the PhD student will focus only on the implementation of the sensorimotor circuit within our NYX simulation platform.

Aims 2: Mapping the sensorimotor circuit and investigating the neuromodulation targets
This section focuses on mapping larval sensorimotor circuits, especially concerning mechanosensory responses. It intends to feed this project more experimental data than accumulated. The PhD is not dependent on it, but finer exploration of neuromodulation and muscle control can stem from it.
It will discern how various internal and external factors modulate the said network. The objective is to delineate how information travels from the decision circuit at the early processing stage9 to muscular execution. The initial focus will be on primary neuromodulation targets and the roles of inhibitory or excitatory neurons. The research will adapt based on preliminary findings; for instance, should velocity emerge as a pivotal neuromodulation aspect, the study will centre on it. There's an emphasis on understanding the interplay between larval locomotion and various influencing factors, such as food availability or substrate properties.

Aim 3: Theoretical models of neuromodulation
The endeavour is to devise theoretical models illustrating circuit dynamics, from the early sensory processing stages to the muscular output, utilising specific neuron dynamics and synapses10,11. This model will seamlessly integrate with the NYX software and other experimental data. It will serve a dual purpose: inferring neuromodulation effects on the circuit and simulating behavioural outputs. A synergistic approach is highlighted, where the outcomes of Aim 2 inform the models in Aim 3, and vice versa. This aim will benefit from the collaboration with the theoretical PhD project of Astrid Nilson, which focused on exploring the architectural consequences of neuromodulation in artificial and biological neural networks.

Project feasibility and collaborations
This ambitious project benefits from extensive preliminary work on sensorimotor circuitry12–14, an established behaviour analysis setup14, methods from the Jovanic lab, models from Masson lab, and the PhD candidate's expertise with the model and techniques.

References

1. Dickinson, P. S. Neuromodulation of central pattern generators in invertebrates and vertebrates. Curr. Opin. Neurobiol. 16, 604–614 (2006).
2. Eriksson, A. et al. Neuromodulatory circuit effects on Drosophila feeding behaviour and metabolism. Sci. Rep. 7, 8839 (2017).
3. Krasne, F. B. & Edwards, D. H. Modulation of the crayfish escape reflex--physiology and neuroethology. Integr. Comp. Biol. 42, 705–715 (2002).
4. Lin, S., Senapati, B. & Tsao, C.-H. Neural basis of hunger-driven behaviour in Drosophila. Open Biol. 9, 180259 (2019).
5. Marder, E. Neuromodulation of neuronal circuits: back to the future. Neuron 76, 1–11 (2012).
6. Nässel, D. R., Pauls, D. & Huetteroth, W. Neuropeptides in modulation of Drosophila behavior: how to get a grip on their pleiotropic actions. Curr. Opin. Insect Sci. 36, 1–8 (2019).
7. Taghert, P. H. & Nitabach, M. N. Peptide neuromodulation in invertebrate model systems. Neuron 76, 82–97 (2012).
8. Tierney, A. J. Feeding, hunger, satiety and serotonin in invertebrates. Proc. R. Soc. B Biol. Sci. 287, (2020).
9. Jovanic, T. et al. Competitive Disinhibition Mediates Behavioral Choice and Sequences in Drosophila. Cell 167, 858-870.e19 (2016).
10. Lappalainen, J. K. et al. Connectome-constrained deep mechanistic networks predict neural responses across the fly visual system at single-neuron resolution. 2023.03.11.532232 Preprint at https://doi.org/10.1101/2023.03.11.532232 (2023).
11. Miller, K. D. & Fumarola, F. Mathematical Equivalence of Two Common Forms of Firing-Rate Models of Neural Networks. Neural Comput. 24, 25–31 (2012).
12. Jovanic, T. et al. Competitive Disinhibition Mediates Behavioral Choice and Sequences in Drosophila. Cell 167, 858-870.e19 (2016).
13. Jovanic, T. et al. Neural Substrates of Drosophila Larval Anemotaxis. Curr. Biol. CB 29, 554-566.e4 (2019).
14. Masson, J.-B. et al. Identifying neural substrates of competitive interactions and sequence transitions during mechanosensory responses in Drosophila. PLOS Genet. 16, e1008589 (2020).

Monitoring the brain activity is routinely achieved by electro-encephalogram (EEG). However, the exact underlying neuronal network activity is unknown especially during general anesthesia . The Rouach’s group at college-de-France obtained dual recordings applying increasing concentrations of anaesthetic (Propofol and sevoflurane) with Neuropixels probes.
The goal of this PhD thesis is to develop a statistical approach and Machine-Learning interface to quantify possible correlation between EEG and electrophysiology signals. We will also explore whether anaesthetics can impact long-term changes in the brain. This approach will be used to identify novel predictive motifs to design predictive algorithms. The candidate will develop and implement ML-algorithms. The development includes spectral analysis, wavelets, ML on multiple times series.

Background: Brain rhythms are made of various dominant oscillations, relevant during sleep, meditation, and anesthesia, such as alpha [8-12] and delta [0,4] oscillations . We recently developed an approach based on signal processing, wavelet decompositions, and classification methods to remove artefacts from the ElectroEncephaloGram (EEG) and predict in the first 10 minutes the sensitivity to general anaesthesia (GA). The statistical indicators were extracted from signal processing and mathematical morphology, computed from the EEG in the first minutes of exposure to hypnotics, and were correlated with patient sensitivity to anaesthetics and could possibly be used to avoid a too deep anaesthesia.
However, to date, there are no predictive indices to estimate the patient sensitivity in real time, despite several attempts based on the changes in frequency band dominancy (alpha to delta ratio in humans) or the delay in the appearance of delta waves. Indeed, the EEG acquired in real time contains transient markers characteristics of a too deep anesthesia, such as transient patterns known as isoelectric suppressions (IES) which consist of periods of flat EEG lasting from a few seconds to several minutes.

Recently the Holcman’s lab developed novel stochastic modeling approaches to study alpha-band dynamics based on neuronal network dynamical system, however, this approach remains difficult to implement and in particular we are still missing a model of alpha spindle events (transient burst of activity in the alpha band) that could be used to study their predictive values.

Spectral analysis of EEG placed on cortical frontal electrodes relies on spectrograms computed over tens of seconds, with a low-time resolution in a signal containing a large amount of artifacts. We recently developed an optimal procedure based on optimal transport of wavelet coefficients to replace artifacts epochs by surrogated signals, allowing real-time parameter estimations. EEG monitoring tracks during GA can measure the depth of anesthesia (DoA), defined as the degree of central nervous system depression produced by anesthetics. DoA commercial monitors such as the Bispectral Index System (BIS) monitor (Medtronic, Ireland) convert the EEG signal into an index between 0 (cortical silence) and 100 (awake patient).
Today, EEG monitoring and the DoA is used to track the loss of consciousness, but not to predict the personalized time-dependent optimal dosage. Anesthetics can transiently modify dendritic conduction by decoupling signaling along apical dendrites. How confusion or loss of memory can occur during anesthesia remains unclear and we will benefit from in vivo physiological approach using Neuropixels probes, to measure simultaneously hundreds of parallel sensory synaptic input. To identify mechanisms and associated patterns that lead to synaptic possible modification, we will record neuronal rhythms in ex-vivo thalamocortical (TC) slices by using the multi-electrode array (MEA) technique.
Spectral analysis of time series has become a powerful tool for segmenting and extracting predictive features from temporal physiological data. Although the Fast-Fourier Transform and improved methods applied on a sliding time window have remained the gold standard of real-time signal processing, including EEG, recent approaches such as Empirical Mode Decomposition (EMD) wavelet transform, now allow a much accurate time-frequency resolution. As a result, although the alpha-band [8-12] Hz of the EEG appears as a continuous line in the frequency domain, these approaches have revealed in our hands that it is largely fragmented and made of burst sequences of activity called spindles. We propose to study how this fragmentation carries information about brain state using physiology, modeling, signal processing, and classification approaches on clinical data.

Proposed research:
We propose first to model the EEG fragmentation using stochastic differential equations (multi-dimensional Ornstein-UIhenbeck--OU--processes) and not the nonlinear dynamical systems. We will then use this model to identify patterns and predict the dynamics of the EEG signal in real-time. This approach will rely on real-time spectral analysis based on the spectrogram, wavelet, and Empirical Mode Decompositions, tracking the dominant frequency band based on power spectral decomposition and computing the absence of specific bands, based on adaptative thresholding. This approach will allow us studying possible transitions into deep brain states, especially during GA in mice. We will couple the present OU-modeling and the spectral analysis.
To better characterize these brain states and improve both the signal processing and the modeling part, we will investigate how anaesthetics can modify neuronal network activity and induce depression of activity. We will analyse neuronal rhythms in ex-vivo TC slices by using the MEA technique (data gerenated by the Rouach Lab, College-de-France). In TC slices, the anatomical and functional connectivity between the ventrobasal (VB) thalamus and the S1 cortex is maintained intact, thus allowing selective activation of the main ascending cortical input ex vivo.

Finally, based on the patterns that will be extracted in the physiology study, we propose to build and test an algorithm that will evaluate in real-time the brain state to prevent deep sedation and long-term memory changes. The algorithm will be developed for humans and rodents. We will use behavioral tests in rodents after anesthesia to correlate the statistics contained in the EEG with possible post-anesthetic changes and short-term memory impairment. The real-time algorithms will be developed in collaboration with a start-up company dedicated to predicting brain dynamics by combining transient motifs of the EEG signal with Machine learning. The real-time algorithm will be used to evaluate the dose to control the depth of anesthesia during human surgical intervention.

Dysfunction of the Golgi apparatus (GA) due to constitutive GA-proteins deficiencies during development has been associated with a group of disorders named Golgipathies, in which the nervous and skeletal systems are primarily affected. SSMG is a genetic Golgipathy characterized by short stature, micrognathia and microcephaly, resulting from mutations in ARCN1, a gene encoding COP-1D, the delta subunit of the COP-I complex. Why mutations in an ubiquitous protein involved in Golgi organization and trafficking in all cells specifically affect some organs is not understood. The goal of the thesis is to identify specific mechanisms through which COP-1D mutations lead to microcephaly in humans. This will be achieved by modeling cortical development from hiPS cells. Generating patient-derived cortical organoids at various developmental stages the role of COP-1D will be studied in progenitors and neurons, and the impact on GA functions and interaction with other organelles will be determined.

The context : The Golgi apparatus (GA) is involved in a broad spectrum of activities, from lipid biosynthesis and membrane transport to the posttranslational processing and trafficking of most proteins, autophagy, and the stress response. These functions are essential to ensure the homeostasis of adult tissues, but growing evidence including recent disease-gene identifications show the important role of the GA during development. Dysfunction of the GA due to constitutive GA-proteins deficiencies during development has been associated with a group of developmental disorders named Golgipathies (1,2), in which the nervous and skeletal systems are particularly affected. In particular, brain growth failure (microcephaly) with intellectual disability (ID) as well as short stature and bone deformations due to chondrodysplasia are frequently represented in human Golgipathies (3). Microcephaly with ID is characterized by limitations in cognitive functions, language and social behavior, represents a burdensome pain for patients and their family and requires the extensive use of public health resources. GA involvement in microcephaly is an emerging concept to which the team has contributed through recent disease-gene identifications (4–6) but the mechanisms through which GA dysfunction causes brain growth failure and why brain appears more vulnerable to GA dysfunction remains unexplored.

The objectives : The COPI coatomer is a protein complex made up of 7 ubiquitous subunits, involved in vesicular transport, in particular between the GA and the endoplasmic reticulum (ER). Heterozygous mutations in the COPI-delta subunit (ARCN1) have been identified in patients with skeletal defects (dwarfism), facial dysmorphism and microcephaly suggesting that haploinsufficiency of ARCN1 impacts certain organs such as the brain and skeleton in a specific manner (5). In addition to GA-to-ER vesicular transport, COPI has been proposed to function in ER stress pathways and in mitochondrial homeostasis by regulating the localization of mRNAs encoding mitochondrial proteins (7) or by promoting membrane fission events in mitochondria (8). However, whether these functions are specifically altered in patients and whether it has a link with their microcephalic phenotype is unknown. The objective of the project is to explore the various stress pathways and aspects of mitochondrial metabolism to identify specific mechanisms explaining how COP-1D mutations lead to microcephaly in this Golgipathy. Human induced pluripotent stem (hiPS) cell-derived 2D progenitors, 2D neurons and 3D organoids will be used as a model of cortical development. hiPS cell lines from two unrelated patients are already available in the laboratory and an isogenic control mimicking the mutation of one of them was successfully reproduced by CRISPR Cas9 in a control line. The techniques used will include 2D and 3D pluripotent stem cell cultures (hiPSCs), CRISPR Cas technology, immunohistochemical labeling, live imaging and confocal microscopy analyses as well as protein expression and molecular analyzes of ER and GA stress pathways, and mitochondrial activities by spectrophotometry and respirometry.

The candidate : the candidate must have prior experience in cell culture. A background in hiPS cell culture and maintenance will be preferred but is not mandatory. He/She should show strong motivation to work in interaction with a team and develop his/her own initiatives.

The lab : The project will take place at NeuroDiderot, the Inserm unit fully dedicated to brain development located at Robert Debré Hospital in Paris. The candidate will work in the NeuroDev team (https://neurodiderot.u-paris.fr/en/research-teams/) and will benefit from the exciting and warm environment of the NeuroDiderot, skills in hiPS-related differentiation of the HumBO core facility (https://neurodiderot.u-paris.fr/en/humbo/) as well as fruitful interactions with the C-Brains partners.


References
1. Passemard S., Perez F., Colin-Lemesre E., Rasika S., Gressens P., El Ghouzzi V. Golgi trafficking defects in postnatal microcephaly: The evidence for "Golgipathies". Prog. Neurobiol., 2017, 153, 46-63.
2. Rasika S., Passemard S., Verloes A., Gressens P., El Ghouzzi V. Golgipathies in neurodevelopment : A new view of old defects. Dev. Neurosci., 2019, 15, 1-21.
3. El Ghouzzi V., Boncompain G. Golgipathies reveal the critical role of the sorting machinery in brain and skeletal development Nat. Commun., 2022, 13, 7397.
4. Dupuis N., Fafouri A., Bayot A., Kumar K., Lecharpentier T., Ball G., Edwards A.D., Bernard V., Dournaud P., Drunat S., Vermelle-Andrzejewski M., Vilain C., Abramowicz M., Désir J., Bonaventure J., Gareil N., Boncompain G., Csaba Z., Perez F., Passemard S., Gressens P., El Ghouzzi V. Dymeclin deficiency causes postnatal microcephaly, hypomyelination and reticulum-to-Golgi trafficking defects in mice and humans. Hum. Mol. Genet., 2015, 24, 2771-2783.
5. Izumi K., Brett M., Nishi E., Drunat S., Tan E., Fujiki K., Lebon S., Cham B., Masuda K., Arakawa M., Jacquinet A., Yamazumi Y., Chen S., Verloes A., Okada Y., Nakamura T., Akiyama T., Gressens P., Foo R., Passemard S., Tan E., El Ghouzzi V*., Shirahige K*. ARCN1 mutations cause a recognizable craniofacial syndrome due to COPI-mediated transport defects. Am. J. Hum. Genet., 2016, 99, 451-459. * co-last?
6. Uwineza A., Caberg J.H., Hitayezu J., Wenric S., Mutesa L., Vial Y., Drunat S., Passemard S., Verloes A., El Ghouzzi V., Bours V. VPS51 biallelic variants cause microcephaly with brain malformations: a confirmatory report. Eur. J. Med. Genet., 2019, 62, 103704.
7. Zabezhinsky D., Slobodin B., Rapaport D., Gerst J.E. An essential role for COPI in mRNA localization to mitochondria and mitochondrial function. Cell Rep. 2016, 15, 540-549.
8. Nagashima S., Tàbara L.C., Tilokani L., Paupe V., Anand H., Pogson J.H., Zunino R., McBride H.M., Prudent J. Golgi-derived PI(4)P-containing vesicles drive late steps of mitochondrial division. Science 2020, 367, 1366-1371.

One of the holy grails of neuroscience is to understand how complex behaviors arise. However, surprisingly little is known about how behaviors evolve. The visual system of Drosophila has been described extensively in terms of cell type composition, development, and circuitry. In the last years, our lab has also characterized neuronal diversity in different insects using single-cell sequencing. The PhD student will address two questions: What are the differences in the circuitry that underlie specific behaviors in different animals? How do differences in neuronal composition, neuronal features, or circuitry drive different behaviors? S/he will combine cutting edge single-cell sequencing techniques with advanced genetic tools in Drosophila and adapt innovative tools for genetic manipulation and circuit function in non-model insects to understand how cell type composition, neuronal specification and differentiation, as well as circuitry, affect specific behaviors.

Neurons are the most diverse cell type in the animal body in terms of morphology, physiology, molecular identity, activity, function, etc. They assemble into diverse circuits, which drive specific behaviours. These behaviours in turn are under selective pressure to help animals adapt to their environment, avoid predators, forage, find mates, and produce offspring. Pairwise comparisons between closely related species have demonstrated that small differences in neuronal identity, activity, and circuitry can affect behavioural output. However, it is unclear how generalizable these findings are. Therefore, a comprehensive study of the mechanisms that drive changes in behaviour remains elusive. The insect visual system has arisen as an ideal system to address the role of neuronal identity, neuronal features, and circuitry in the evolution of behaviours, using single-cell sequencing and advanced genetic tools as means to achieve this.
Insects inhabit diverse environments with different visual landscapes and stimuli; they have done so while preserving the overall organisation of their visual systems. Thus, insect optic lobes represent an ideal system to address how neuronal systems evolve to trigger different behaviors. Indeed, the Drosophila optic lobe has emerged as an ideal system to address these questions for four reasons: a) It is an experimentally manageable, albeit complex structure, for which we have a very comprehensive catalogue of neuronal cell types. Meticulous work from 30 years ago had identified 100 cell types in the optic lobes based solely on morphological characters. Recent work from our and other labs took advantage of elaborate genetic tools, as well as single-cell sequencing, to expand the number of neuronal cell types to 169, based on both morphology and molecular identity. b) We know the basic principles by which most of these neuronal cell types are generated during development and how they are specified. The interplay of spatial and temporal patterning with a Notch binary fate decision is responsible for the generation of the 169 different neuronal types. c) Detailed connectomic maps of the optic lobe as well as a thorough enumeration of the synapses between different cell types can be used to probe circuitry. d) Finally, the neuronal circuits responsible for different activities and behaviours are relatively well characterized. Nonetheless, studies in one species alone cannot address evolutionary questions; the Drosophila visual system is, hence, an invaluable starting point for comparative studies across insects.

The PhD candidate will address two questions:
Aim 1: What are the differences in the circuitry that underlie specific behaviors in different animals?
Neurons assemble in circuits that act as units to control behaviours. Changes in neuronal circuitry, such as eliminating or introducing new synapses can lead to the evolution of new behaviours, even in the absence of any clear difference in the individual neurons that form the circuit.
1a) Adaptation of genetic tools for trans-synaptic tracing and cell-type labeling in Musca and Tribolium. To study neuronal circuits in non-model insects, the student will adapt genetic tools that have been developed in Drosophila. One will be trans-Tango which labels the postsynaptic targets of a neuron of interest; the other is split-Gal4 which will be used to generate neuronal type-specific drivers from broader driver lines. For this goal, s/he will focus on two species: a) Musca domestica, which is a fly whose behaviours and circuits should be comparable, albeit different, to Drosophila and b) the flour beetle Tribolium castaneum, which has been developed as a genetic model organism and for which many genetic tools are already available.
1b) Identifying the presynaptic circuit of lobula neurons. To study circuitry and behaviour, the student will focus on lobula neurons, which in Drosophila are responsible for interpreting visual input and triggering a wide range of behaviours18. S/he will take advantage of available connectomic data to identify the cells presynaptic to lobula neurons in Drosophila. These will mainly be transmedullary neurons – there exist ~50 different and morphologically recognizable transmedullary neuronal types. S/he will then generate split-Gal4 lines in Tribolium and Musca that are specific for different transmedullary subtypes using either Drosophila enhancer fragments or species-specific fragments based on markers identified using available single-cell sequencing data. To compare the circuits that control specific behaviours, s/he will express trans-Tango under the control of the split-Gal4 lines that were identified earlier in Musca and Tribolium. This will lead to the expression of tdTomato in the postsynaptic lobula neurons, which will be sorted, sequenced, and annotated as done routinely in the lab. The student will then be able to compare the lobula neurons’ presynaptic circuits that are responsible for triggering specific behaviours, which will give us a clear idea of the conservation of neuronal circuits across different time scales that span the insect tree and whether new behaviours may rely on different circuits.

Aim 2: How do differences in neuronal composition, neuronal features, or circuitry drive different behaviors? We do not know whether homologous neurons control the same behaviours in different species. The objective of this aim is to address this question in the visual system of different flies.
2a) Do the same neurons drive different behaviours? To address this question, the student will use flies that have similar body plans and have the potential to display similar behaviours, but have split from Drosophila melanogaster ~50 and ~120 million years ago, Drosophila virilis and Musca domestica. They will use the technique that was described in Aim 1b to identify split-Gal4 lines that are expressed in lobula neuronal types. These lines will be used to drive the expression of channel rhodopsin to permit optogenetic manipulation in specific lobula neuronal types in D. virilis and Musca, as has already been done in D. melanogaster. The studemt will record the behaviour and classify it using an exhaustive suite of locomotion (such as reaching, jumping, walking) and social behaviours (such as chase, court, follow). S/he will compare triggered behaviours between the three species and evaluate the extent of same neurons driving different behaviours. Since the three fly species occupy different micro-habitats with different visual stimuli, they will probably have distinct behaviour programs downstream of the feature detection neurons, such as strengthening or weakening of some behaviours or even evolution of different behaviours.
2b) Do new behaviours require new neurons? Working under the assumption that lobula neurons are the main visual output neurons that directly impact behaviour, the student will use available single-cell sequencing data to identify new lobula neurons, based on known markers. S/he will then generate split-Gal4 lines (based on their transcriptome) to a) drive GFP expression and verify that they are lobula neurons and b) drive channel rhodopsin and record the behaviour they control. Delving into this behaviour will open new avenues of research for my lab beyond the timeframe of this proposal.
2c) The role of cell type composition, cell identity, and circuitry in the evolution of new behaviours. As a final goal, to understand how behaviours evolve, the student will integrate all the cross-species data regarding a) cell type composition, b) cell type transcriptomic similarity, c) neuronal circuitry, and d) neuronal type behavioural output, to evaluate the contribution of each of these processes in the evolution of new behaviours at time spans ranging from tens to hundreds million years of species diversification.

Multiple sclerosis (MS) is an autoimmune and demyelinating disease of the central nervous system. One factor contributing to the wide variability in disease severity among MS patients is their ability to initiate a self-repair mechanism known as remyelination. When remyelination is unsuccessful, it results in various types of lesions, including a specific category called mixed active/inactive lesions, which are particularly prominent in individuals with severe forms of the disease.
Recently, we established a humanized animal model that replicates the mixed active/inactive lesions by transplanting lymphocytes from MS patients into a focal demyelinating lesion in the nude mice spinal cords. Our next objective is to elucidate, through spatial transcriptomics, the chronological shifts in molecular and cellular components at the origin of mixed lesion progression using advanced computational modeling.

Role of lymphocytes in remyelination:
An efficient remyelination of Multiple Sclerosis (MS) lesions is critical for a favorable disease evolution as suggested in pathological investigations. This was confirmed in vivo in a recent longitudinal study applying specific imaging myelin markers and showing for the first time that the remyelination capacity of patients was inversely correlated with clinical severity(1). This study also clearly highlighted the existence of a large inter-individual capacity of patients for myelin repair. Remyelination failure is therefore a feature observed in MS and the presence of demyelinated lesions have been well documented. In MS post mortem tissue different type of lesions were described, among which mixed active/inactive lesions, characterized by an inactive center and an active rim with iron loaded microglia (2). These lesions are hallmark of patients with severe disease and at risks of having higher disability. Interestingly, perivascular cuffs of lymphocytes (LY) are often observed in mixed active/inactive lesions, indicating the potential role of these LY in perpetuating an inflammatory activity detrimental to remyelination(2).
Furthermore, T LY were identified as key players in the remyelination process, since remyelination developed to a much lesser extent in mice lacking CD4+ or CD8+ T LY than in normal mice(3). Depending on their activation state, LY can exert different effect on myelin repair. On one hand, it was demonstrated that regulatory T LY are crucial for the process of remyelination(4), while on the other hand, the transfer of pro-inflammatory Th17 LY in cuprizone-fed mice resulted in a decrease in remyelination (5). All these data point to an essential role of T LY in the remyelination process.
A much needed model of mixed active/inactive lesion:
The classical model used to study remyelination, such as the injection of various gliotoxic agents including lysolecithin, ethidium bromide and bacterial endotoxins in white matter tracts of the central nervous system, improved our understanding of the different steps leading to remyelination. Nonetheless, because of the high regenerative potential observed in these models, there is a lack of chronic lesion model mimicking important features of MS.
Therefore, we undertook an original approach to examine the neuro-oligo-immune interactions leading to remyelination failure or success in conditions closer to human pathological condition. We established new in vivo experimental paradigms to investigate how human MS LY influence remyelination (6).
We show that 3 weeks after their grafting, LY from MS patient impede remyelination in vivo when grafted in demyelinated lesions of Nude mouse spinal cord. By deciphering the mechanism of remyelination failure in vitro, we found that MS LY induce a higher pro-inflammatory activation of mouse microglial cells compared to LY from healthy individuals (HD), this process leading to an early blocking of oligodendrocytes progenitor cells (OPC) differentiation. Moreover, we found that patients LY influenced differentially the remyelination process, some showing a beneficial (High Patients) and some a deleterious effect (Low patients) on the repair process, a pattern mimicking what is observed in MS patients(6).
Three months after the grafting, we observe that only LY from MS patients form perivascular cuffs and contact both the vascular wall and mouse immune cells, and not HD LY. MS LY graft impairs oligodendrocytes (OL) differentiation and remyelination, induces the development of a glial scar in the center of the lesion and maintains the development of a demyelination process. Interestingly, in the rim of the lesion in the normal appearing white matter, we also observe that microglia cells are loaded either with myelin (known as MIMS-foamy) or with iron (known as MIMS-iron) as described in MS patient’s post mortem tissue and in MS patients MRI images. Behavioral analysis showed that while non-grafted mice and mice grafted with HD LY improved over time, probably due to normal remyelination, mice grafted with MS patient LY continued to make errors on the notched beam and had a weakened forelimb grip, probably due to the persistence of inflammation and demyelination. We also observed an increase of sensory evoked potential latency in mice grafted with MS patients LY.
In this context, our objective is to understand what are the factors that impede remyelination and perpetuate an inflammatory context in favour of continuous demyelination.

METHODS:
For the first time, MS patients at distinct stages of the disease, either relapsing or progressive (more aggressive form), will be profiled for their remyelination potential through a follow-up with recently developed imaging technologies using magnetic resonance imaging (MRI) and positron emission tomography (PET) to precisely visualize and quantify myelin and inflammation. Blood collection from this unique cohort of MS patients will be used to isolated their LY and graft them in our humanized models.
Next, we will perform a spatial transcriptomic analysis of the demyelinated lesion and its surroundings using the GeoMx DSP Nanostring system in HD or MS conditions 3 weeks, 1.5 months and 3 months after the lesion graft to assess different cell’s (mouse microglia, oligodendrocytes as well as human LY) signatures within the lesion vs border vs normal appearing tissue of mice grafted with HD or MS patients.
We will investigate how molecular changes evolve over time in various cell populations by examining differentially expressed genes. Taking our research a step further, we aim to comprehend the components of multicellular interactions and how distinct subpopulations within the lesion collaborate to either promote remyelination or contribute to lesion development. Consequently, we will analyze the data with a focus on ligand-receptor pairs to deduce intercellular communication patterns based on the coordinated expression of specific gene subsets, establishing a cellular interactome. We will assess the interaction between microglia, oligodendrocytes, and lymphocytes in different groups of grafted mice to gain insights into these processes.
Focusing on each cell datasets, we will also use ontology-based reconstructions to construct molecular networks in response to all the different conditions. Indeed, it is crucial to identify the drivers nodes, i.e. the genes that have the potential to influence the rest of the network, as they can then be key targets for treatments. For this purpose, we have developed a new method called “stepwise target controllability” that introduces a ranking among the target nodes (e.g., effector molecule such as chemokines/cytokines). Then, we iteratively evaluated the controllability of the system by adding one target node (e.g., transcription factor or signaling molecule) at a time in a descending order in order to identify whether the path between the driver node and the target node was altered in pathological conditions.

1. B. Bodini et al., Dynamic imaging of individual remyelination profiles in multiple sclerosis. Annals of neurology, (2016).
2. S. Luchetti et al., Progressive multiple sclerosis patients show substantial lesion activity that correlates with clinical disease severity and sex: a retrospective autopsy cohort analysis. Acta Neuropathol 135, 511-528 (2018).
3. A. J. Bieber, S. Kerr, M. Rodriguez, Efficient central nervous system remyelination requires T cells. Annals of neurology 53, 680-684 (2003).
4. Y. Dombrowski et al., Regulatory T cells promote myelin regeneration in the central nervous system. Nat Neurosci 20, 674-680 (2017).
5. E. G. Baxi et al., Transfer of myelin-reactive th17 cells impairs endogenous remyelination in the central nervous system of cuprizone-fed mice. J Neurosci 35, 8626-8639 (2015).
6. M. El Behi et al., Adaptive human immunity drives remyelination in a mouse model of demyelination. Brain 140, 967-980 (2017).

Rett syndrome (RTT) is a devastating Autism Spectrum Disorder affecting 1/15000 female live births. One of the most frequent and severe MeCP2 mutation associated with RTT (T158M) affects its ability to bind DNA and regulate gene expression. RTT is incurable but could be reversed in adult mice following MeCP2 re-expression. MeCP2 gene therapy appears as a promising strategy but many attempts failed to fine-tune MeCP2 levels and avoid symptoms caused by its overexpression. MeCP2 is localized in the nucleus but the mechanisms involved in its nuclear transport have never been targeted in RTT. Importins play a crucial role in the subcellular transport of proteins towards the nucleus. We will develop a new gene therapy in the T158M model of RTT combining optimized AAV viral vectors and ultrasound methods. We will also manipulate the nuclear shuttling of MeCP2 to define how we can avoid its overexpression or potentiate gene therapy effects at the molecular, neuronal, and behavioral levels.

Introduction.
Rett syndrome (RTT) is a devastating autism spectrum disorder (ASD) caused by mutations in the gene MECP2 (methyl-CpG binding protein 2). RTT accounts for 10% of cases of profound intellectual disability of genetic origin in women. Amid normal early growth, RTT girls undergo a significant regression of early milestones between 6 to 18 months and develop a myriad of neurological symptoms and autism traits. Thousands of MECP2 mutations have been reported. Mutations of the T158 residue, within the Methyl Binding Domain are frequent (~10% of all RTT) and severe. RTT is incurable but researchers demonstrated its reversibility in mice. In this context, MeCP2 gene therapy appeared a promising strategy to restore its function but this approach failed in rescuing optimal protein levels while preventing deleterious overexpressions. MeCP2 protein is a key nuclear player in gene expression regulation. Yet, the mechanisms of its transport were not targeted to alleviate MeCP2-related disorders. Importins are nuclear transport factors with pivotal roles in the subcellular transport of proteins. Using a battery of approaches from behavior to molecular mechanisms, we discovered a role for importin alpha-5 in the regulation of anxiety via control of nuclear shuttling of MeCP2 and of other importins in chronic pain and memory.

Hypothesis and aims.
MeCP2 gene replacement strategies have been challenged by the poor brain penetration of the vectors, brain cells’ transduction, and fine-tuning of expression to avoid phenotypes reminiscent of MeCP2 duplication syndrome. Clinical trials are initiated to test gene therapy in RTT girls. However, there is no one-size-fit-all solution to treat girls who feature a diverse repertoire of mutations. It is thus pivotal to develop new and optimized MeCP2 gene therapy in models featuring common and severe MeCP2 variants. We are making here the hypothesis that:
1. Optimized MeCP2 gene therapy (i.e., combined with focal-ultrasound technology) will boost brain penetration of the Mecp2 vector, increase the transduction of brain cells, and ameliorate RTT phenotypes in T158M mice.
2. Manipulation of the importin pathways will allow to adjust MeCP2 nuclear levels within physiological range and ameliorate behavioral, cellular, & transcriptional impacts of the gene therapy.

Methods.
We will use a multi-level strategy combining the expertise of the Panayotis and Oheim teams at the SPPIN (CNRS, Université Paris Cité) and inputs from the Roux group at the Marseille Medical Genetics (Inserm, Aix-Marseille Université). The proposal is organized around 3 levels, from the evaluation of behavioral & physiological changes upon modulation of MeCP2’s expression and nuclear shuttling to assessments of the consequences of these manipulations on neuronal morphology and gene expression. The proposed study will use molecular biology, biochemistry, and histology methods (design/validation of AAVs), stereotaxic approaches, comprehensive mouse behavioral profiling (mouse video-tracking), neuronal pathway labelling and optical imaging of the neuronal morphology, and evaluation of gene expression through RNA-seq and bioinformatic methods.

Conclusion.
If successful, the proposed work will enable the development of a new MeCP2 gene therapy in a severe clinical context. The research question is driven by our very intriguing published works showing that manipulation of specific importin alphas can help to control the shuttling and amount of transcription factors in the nucleus of specific neuronal cell types or brain areas. Given the difficulty to design a safe and efficient gene therapy in all RTT girls, we believe that an optimized approach based on the maximization of viral vector delivery (FUS) and the fine-tuning of MeCP2 nuclear levels through a dynamic post-translational process such as nuclear transport should be used to tailor and increase the potency of the treatment. Research insights from this work have the potential to be translated to other pathologies where the subcellular levels and/or localization of transcription factors is impaired.

The vestibular system participates in essential functions such as balance, postural control and gaze stabilization. Beyond these, vestibular signals also contribute to cognitive processes e.g spatial orientation and navigation. Because of its involvement in many basic functions important in our daily life, inner ear pathologies affecting the vestibular system are associated with a significant deterioration of the well-being of patients and represent an important public health concern. The compensation taking place after transient or permanent vestibular loss is known to involve dynamical multisensory reweighting of proprioceptive and visual inputs and of internal efferent copies. The proposed PhD thesis will consist in investigating in animal models of vestibular pathologies, the dynamic of the sensory substitution process and determine the role of sensory/motor stimulation in the capacity of the vestibular system to cope with pathway-specific deficits.

Detailed program:
A/ Context:
Vestibular pathologies constitute a major public health concern: balance dysfunction occurs frequently in aging people (85% prevalence above 80; Agrawal et al., 2013) but also in younger people (prevalence of 35% in the 40 year-old). While animal models often use permanent vestibular lesions (Simon et al., 2020), many diseases in fact consist in a gradual, transient and/or partial loss of vestibular function. For instance, Menière’s disease, which represents ~9% of all vestibular pathologies in the adult and occurs in the <60 years old, is characterized by recurrent episodes with brief (<24h) fluctuating symptoms and otherwise normal vestibular function before long-term deterioration arises (Lopez-Escamez et al., 2015). Transitory vestibular symptoms are also commonly reported in vestibular neuritis where, although symptoms tend to rapidly disappear due to compensation, vestibular function can recuperate up to a year after the initial loss (Welgampola et al., 2019). Gradual and partial vestibular loss is also encountered as a side-effect of some ototoxic anti-cancer treatments (prevalence in treated patients ~40%; cisplatin; Paken et al., 2016; Prayuenyong et al., 2018). There is therefore a crucial need to develop animal models mimicking permanent/complete or transitory/partial vestibular loss, respectively.
Role of internal efferent copies in vestibular compensation: research on post-lesion plasticity following permanent vestibular loss has shed light on the neural plastic mechanisms that follow a chronic unilateral or bilateral vestibular lesion, a process referred as “vestibular compensation” (Brandt et al., 1997; Cullen et al., 2010; Beraneck and Idoux, 2012). The compensation taking place after the lesion is known to involve dynamical multisensory reweighting of proprioceptive and visual inputs and of internal efferent copies (Cullen et al., 2010; Sadeghi et al., 2012; Sadeghi and Beraneck, 2020). We have recently demonstrated that during locomotion, an efference copy signal originating in the spinal cord participates to gaze stabilization. Specifically, we demonstrated in the amphibian (Xenopus) (Bacque-cazenave et al. 2022) and in neonatal mouse spine-brainstem isolated preparation that an efference copy command originating from the spinal cord CPG drives compensatory eye movements. In the adult mice, we demonstrated that the spinal efferent copy generates eye movements coupled to forelimb movements with a robust horizontal component (França de Barros et al. 2022). The existence of this new ascending pathway raises intriguing questions concerning its potential implication in the functional recovery that occurs following a vestibular loss.
Objectives: the primary goal of this PhD will be to characterize in different animal models of vestibular pathologies, the dynamic of the sensory substitution that occurs following functional vestibular loss. Specifically, the objective will be to determine the degree of vestibular loss necessary for visual, proprioceptive, and locomotor efferent signals substitution to occur, and their relative importance to the different vestibular-specific pathways. The second objective will be to determine how visual or locomotor stimulation (rehabilitation-like intervention) help the recovery of gaze stabilization and postural reflexes, respectively. Video-oculography quantification will measure the reweighting of vestibular, visual, and spinal-extraocular signals to ensure functional vestibular control. Comparison of the different models of vestibular loss will shed light on the ponderation of the different signals in relation to the intervention, and on the role of the spinal motor command in functional homeostasis. Once successful, these experiments will be completed with electrophysiological recordings of central vestibular neurons, and anatomical tracing.

B/ Methodology:
Animal models of vestibular lesions: our research group holds a unique expertise in the generation of animal models of permanent vestibular pathologies (Simon et al. 2020; 2021). To better model fluctuating inner ear function, we recently implemented protocols based on exposure to an ototoxic substance, 3,3?-iminodiproprionitrile (IDPN) (Greguske et al., 2019)(Schenberg et al. In Press in eLife). Subchronic exposure to IDPN in drinking water at low doses allowed for progressive ototoxicity, leading to a partial and largely reversible loss of function. The subchronic IDPN protocol was shown to cause reversible postural and locomotor deficits (Martins-Lopes et al., 2019a), and to trigger sensory substitution in gaze stabilizing reflexes (Schenberg et al. 2023).
Behavioural quantifications: quantification of postural control, vestibulo-ocular reflex & optokinetic reflexes using motion-capture and video-oculography techniques in adult mice constitute the core expertise of the team. Vestibular stimulation is achieved using a 5-axis turntable allowing the specific stimulation of the semicircular canals (yaw and roll rotations) or of the otoliths (static ocular-counter roll or off-vertical axis rotation) at different frequencies (0.2-2Hz) and velocities (Up to 70°/s). Optokinetic stimulation is achieved using a servo-controlled projector surrounding the animal and coupled to the turntable making it possible to combine visual and vestibular stimulation. During the PhD this proven technology already used in passive conditions will be used in new behavioural situations: active locomotion. To that end, a servo-controlled wheel or a rotatory ball will be coupled to the existing apparatus in order to quantify the OKR and VOR responses in passive vs active conditions.
Electrophysiology and anatomical tracing: to further describe the sensory reweighting occurring following vestibular loss, electrophysiological recordings of vestibular neurons implicated in gaze stabilization or postural controls will be implemented. Chronic recordings in passive vs active (locomotion) conditions of functionally identified vestibular neurons, completed with anatomical tracing and immune-histochemistry, will reveal the pathway-specific features of sensory substitution.
Rehabilitation protocols: The student will adapt to the animal model established rehabilitation protocols to determine how visual and/or locomotor stimulation modifies the weight of the different neural signals involved in vestibular compensation.

C/ Candidate skills: the candidate should hold a master in Neuroscience or equivalent, and an authorization for animal handling. He/She will justify of a previous experience in scientific animal research, having ideally already combined behavior and cellular investigations. Specific skills: Surgery; behavioural quantification; electrophysiology; schedule, prepare and manage scientific experiments; analyse data; data computing; critical thinking.
Team references in relation with the proposed project:
1. Schenberg L, et al. Subchronic alteration of vestibular hair cells in mice: implications for multisensory gaze stabilization. eLife (In press)
2. França de Barros F, et al. Conservation of locomotion-induced oculomotor activity through evolution in mammals. Current Biology. 2022
3. Bacqué-Cazenave J, et al. Locomotion-induced ocular motor behavior in larval Xenopus is developmentally tuned by visuo-vestibular reflexes. Nature Communications. 2022
4. Simon F, et al. Implication of Vestibular Hair Cell Loss of Planar Polarity for the Canal and Otolith-Dependent Vestibulo-Ocular Reflexes in Celsr1-/- Mice. Frontiers in Neurosciences. 2021.
5. França de Barros F, et al. Long term visuo-vestibular mismatch in freely behaving mice differentially affects gaze stabilizing reflexes. Scientific reports. 2020

Hearing deficits, when they occur early in life, prevent normal shaping of neuronal circuits by acoustical stimulation especially during the critical period, an early developmental window of high plasticity. This might partly explain why the late fitting of hearing aids or cochlear implants is less successful in terms of speech comprehension than early fitting. Using a new generation genetic model of ‘ideal’ hearing restoration, the candidate will investigate the extent to which sound processing, but also auditory performance and social behaviors are restored as a function of the age of hearing restoration. The PhD candidate will assess peripheral and central auditory processing of complex sounds after hearing restoration using in vivo electrophysiology. In parallel, the neurophysiological changes will be correlated with the behavior of animals over weeks and following hearing restoration in a colony of freely-moving mice using an innovative “Big Brother” behavior system we designed.

Sensorineural or perceptive hearing loss affects one infant in 700 at birth and approximately one child or young adult in 500 before the age of 20. It is mostly caused by damage to the cochlea, the peripheral auditory organ. Sensory deprivation early in life prevents shaping of the brain microcircuits to the environment through plasticity mechanisms, progressive deterioration of the auditory neuronal relays, cross modal reorganization of the brain and dramatic changes in the vascular system.

These deficits are thought to impair the outcome of interventions in deaf patients. Auditory rehabilitation in hearing impaired patients, currently with neuroprosthetics through the fitting of hearing aids or cochlear implants (a device that bypasses the cochlea to directly stimulate the auditory nerve) are not always satisfactory. Its success largely depends on the ability to revive maturation of the central auditory pathway by neuronal plasticity for optimal processing of the restored sensory stimulation. In addition, these indirect central auditory deficits may also compromise the success of upcoming gene therapies to cure genetic forms of deafness. The goal of the PhD is to determine the reversibility of peripheral, central neuronal and cognitive deficits in genetic forms of deafness upon hearing restoration.

To tackle this question, the PhD student will use a new generation genetic mouse model of hearing restoration. In this model, cre-dependent mutant mice are born deaf and recover their hearing after tamoxifen injection to activate a creERT2. This new genetic model allows hearing restoration at any age, making possible the probing of auditory perception recovery throughout lifetime during the critical period, an early developmental window of high plasticity, or during adulthood.
The candidate will focus on both neurophysiology and behaviour to address whether hearing deficits are reversed or halted upon hearing restoration. To this aim, after hearing restoration and depending on the age of the tamoxifen injection for induction, he will investigate i) the functioning and molecular structure of the auditory sensory organ in the cochlea; ii) the neurophysiological response of the auditory cortex to sounds of various complexity; iii) the auditory performance level and social behavior of animals following hearing restoration. For these two later parts, the candidate will benefit from a very innovative system “Big Brother”, providing a socially enriched environment to a colony of mice and allowing a daily tracking of changes in neural activity, physiological parameters, and behavioral performance in the absence of the experimenter. Briefly, this setup combines artificial intelligence, video monitoring, and animal localization by RFID implants to decipher the individual behaviors and social interactions within the colony. Microphones continuously record vocalizations while animals are conditioned with operant walls. Wireless neural implants help recording the same neurons across days. With this system, the candidate will test how durable these benefits are and how social interactions and acoustic communication improve with hearing restoration.

The knowledge gleaned using this model will increase our understanding of the plasticity mechanisms involved in sensory deprivation and restoration. The expectations of patients are high and the results obtained will evaluate the extent to which central auditory deficits may impair the outcome of the upcoming gene therapies. Moreover, the results should provide a scientific basis for new paradigms for auditory rehabilitation, which is currently empiric in nature, with the adaptation of auditory training to the cortical or perceptual deficits identified and the possible development of plasticity-related therapies.
The PhD will take place in the “Plasticity of Central Auditory Circuits” team at the Hearing Institute, a research center of Institut Pasteur. The team uses a combination of approaches including in vivo electrophysiology and in vitro electrophysiology (patch-clamp), behavior paradigms to test auditory perception, and different sets of cutting edge immunohistochemical techniques.

Parkinson's Disease (PD) entails the specific loss of dopaminergic neurons in the substantia nigra pars compacta, marked by intraneuronal protein aggregates and activated microglia. Current treatments focus on symptom relief, given the absence of a cure. The scarcity of accurate models hinders research progress. We propose employing induced pluripotent stem cell-derived brain organoids to replicate PD pathology and vulnerability. Long-term culture midbrain organoids, inclusive of microglia, will be subjected to neurotoxic challenges. Multiplex immunohistochemistry and single-cell RNA sequencing will facilitate a comprehensive analysis, revealing genetic, metabolic, and immunological contributors to PD. This reproducible model aims to bridge preclinical and clinical realms, advancing drug development.

Introduction
Parkinson's disease (PD) is characterized by the selective loss of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNpc). One of the key pathological features observed in the majority of PD cases is the presence of intraneuronal protein aggregates called Lewy bodies, primarily composed of ?-synuclein (A-SYN). Furthermore, the presence of activated microglia is a common finding in PD. Despite several efforts, there is currently no cure for PD, and available therapeutic approaches focus on managing symptoms. One of the major challenges lies in the limited availability of models that accurately replicate the complexity of A-SYN pathology at different disease stages. The translatability of findings from mouse models to humans is often limited due to divergent biology between species. For instance, there are notable differences between human and mouse A-SYN in terms of their structure and aggregation propensity. Furthermore, mice and humans exhibit variations in genes and signaling pathways involved in microglial function and activation. For example, the expression and activation of specific receptors involved in A-SYN-related inflammation and clearance of aggregates, such as Toll-like receptors and pattern recognition receptors, vary between species. To address these limitations, we propose to develop a human midbrain organoid model integrating microglia (MbO-Mg) from human and non-human primate induced pluripotent stem cells that specifically captures the selective vulnerability of SNpc DA neurons and microglial contribution to ?-synucleinopathy.

Research Plan
The goal of the thesis project is to develop accurate MbO-Mg models that recapitulate PD pathology, selective SNpc DA neuronal vulnerability, and microglial contribution to neurodegenration. We will generate long-term culture midbrain organoids integrating isogenic microglia. Toxic, chemical, inflammatory and A-SYN challenges will be used to induce pathology. Using a variety of methods, including multiplex immunohistochemistry, single-cell RNA sequencing and spatial transcriptomics, we will perform a detailed comparative analysis of molecular and cytological features to identify human-specific genetic, metabolic and immunological features responsible for disease.

Aim 1: Development of long-term cultures of midbrain organoids integrating microglia (MbO-Mg)
In this first objective, the primary goal is to establish robust and durable cultures of midbrain organoids (MbOs) by incorporating isogenic induced pluripotent stem cell (iPSC)-derived microglia (referred to as MbO-Mg). To achieve this, we will utilize a state-of-the-art motorized spinning bioreactor system that facilitates the advanced maturation of dopamine-producing substantia nigra pars compacta (SNpc) neurons, ensuring their long-term viability and functional integration with iPSC-derived microglia. To capture human-specific vulnerabilities to neurodegenerative processes, we will extend our investigations to genetic forms of PD and non-human primate organoids. We aim to comprehensively evaluate midbrain development in these two different contexts. To map the progression of MbO-Mg development, we will employ multiplex immunohistochemistry, specifically using the PhenoCycler™-Fusion platform. This will allow us to analyze multiple cellular markers simultaneously, providing a comprehensive view of the differentiation of SNpc A9 DA neurons. We will quantify their abundance and their distribution within the organoids, providing critical insight into their spatial arrangement. In addition, we will analyze the integration of iPSC-derived microglia into hMbO. This analysis will include a thorough examination of specific microglial identity markers, with a strong focus on the intricate interactions between microglia and neuronal cells. Our investigation will also include an assessment of the presence and functionality of synaptic connections, shedding light on the dynamic network of cellular communication within the organoid system. To measure the immunomodulatory response within the MbO-Mg environment, we will use intracellular multiplex flow cytometry. Using this approach, we will measure the production of inflammatory cytokines and chemokines, providing critical insight into the immune signaling dynamics underlying cellular interactions within the organoids. Overall, this first goal lays the foundation for the development of advanced midbrain organoid cultures incorporating microglia, providing a powerful platform for studying neurodegenerative processes and bridging in vitro models to the complex human neural system.

Aim 2. Development and characterization of a midbrain organoid model of A-SYN pathology
To achieve this goal, we will perform a number of critical steps: 1) MbO-Mg cells will be exposed to labeled A-SYN PFF treatment. We will then characterize the resulting pathological changes. Immunohistochemistry will be performed 1, 2 and 3 weeks after PFF treatment. The PhenoCycler™-Fusion system will be used for multiplex immunohistochemistry, allowing precise evaluation of the cell type-specific distribution of A-SYN and pSer129 A-SYN, as well as assessment of A-SYN uptake and intracellular distribution. At the latest time points, we will determine the emergence of features reminiscent of late-stage PD pathology. Immunohistochemical techniques will be used to selectively stain Proteinase K-resistant A-SYN and ubiquitin deposits. In addition, we will perform quantitative analysis of SNpc A9 and A10 DA neurons during the course of this study to elucidate their respective vulnerabilities. This comprehensive approach will not only improve our understanding of PD pathology, but also lay the groundwork for the development of a robust midbrain organoid model for A-SYN-related research.

Aim 3. Investigation of cell-type specific stress responses contributing to Parkinson’s disease pathology

The aim of this proposal is to identify the genetic and metabolic factors that contribute to the differential susceptibility to PD. To this end, we will generate long-term culture midbrain organoids integrating isogenic microglia. Toxic, chemical, inflammatory and A-SYN challenges will be used to induce pathology. Using a variety of methods, including multiplex immunohistochemistry and single-cell RNA sequencing, we will perform a detailed comparative analysis of molecular and cytological features to identify cell type-specific genetic, metabolic and immunological features responsible for disease. At the subcellular level, compartmental distribution of RNAs within cells provides an efficient way to identify transcripts at the site of function and in response to local stimuli. This will be key to identifying the unique transcriptome signature of different cell types to stress stimuli. It will also shed light on the intricate mechanisms underlying A-SYN uptake and spread, and the role of glial cells in early disease. These advanced MbO models, representing both idiopathic and genetic PD, will provide valuable insights into the underlying disease mechanisms. They will also provide novel targets for therapeutic intervention and potential early disease biomarkers that could aid in diagnosis and monitoring.

Feasibility of the project
The student will receive training in stem cell biology and disease modeling. This will include the generation and characterization of brain organoids. The project will involve multiple techniques such as cell biology, molecular biology, imaging, molecular biology, biochemistry, and introduction to bionformatics. As the proposed research builds on the expertise and resources of the host laboratory, all experiments are feasible. This patient-oriented research will significantly benefit from the collaborative environment and platforms at Imagine/SFR Necker (bioinformatics, scRNA-Seq, cell imaging).

Anatomical and functional brain-imaging studies have revealed that a complex brain architecture is already present at birth but little is known about the depth of information processing and the richness of mental representations in human infants. In this thesis, the aim will be to study the dynamics of conscious access in 3–4-month-old-infants. Using high-density EEG (128-280 channels), combined with new methodological approaches such as decoding, single-trial analyses of variability patterns, and new paradigms that monitor infants’ attention with frequency tagging, pupillometry and eye-tracking, the goal will be to separate infants’ short-lived, externally triggered versus sustained, internally driven responses. The thesis will be based on the adaptation to infants of experimental paradigms used to test conscious access in adults (ignition, top-down amplification, serial bottleneck). This project aims to enhance our understanding of early human cognition.

Thesis in human cognitive development:
After centuries of considering infants’ mental life as either empty or confused, research in cognitive development has shown surprisingly sophisticated cognitive competencies from birth on, hidden beyond infants’ poor motor behavior. Anatomical and functional brain-imaging studies have revealed that features associated with adult brains such as structural and functional asymmetries, long-distance connectivity and activations in prefrontal cortex are observed early on. Despite such advances, little is known about the depth of information processing and the richness of mental representations in human infants.
The aim of the thesis is therefore to fill this gap and study whether human infants possess high-level cognitive processes capable of selecting information, sustaining it, and using it to guide their attention from the first months of life on and a capacity to explicitly use these internal representations in subsequent cognitive operations.
Because infants cannot report on what they see, hear or feel, we will rely on neural signatures which are associated with conscious access in adults using high-density EEG (128 to 280 channels) and test 3–4-month-old infants, because this age is a first important milestone in infants’ brain and cognitive development. The different experimental paradigms used in adults to test conscious access will thus be adapted to infants. Thanks to a visual localizer, we will be able to train decoders that will be subsequently tested on the main experiments to follow the dynamics of information processing, notably looking for metastable states. Neural responses will be correlated with behavioral and physiological responses (saccades, pupil dilation, heart rate modulation). Through a systematic investigation of the defining properties of conscious perception and the cognitive processes it enables, we should obtain a better appreciation of the dynamics within an immature brain and know when (within the first few months of life, or later) humans first benefit from the distinct learning characteristics that allow them to bypass slow associative learning through explicit and symbolic internal operations.

Requirements: The student must be comfortable with programming in Python and working with young infants and their parents.

Supervisor: The thesis will be directed by Ghislaine Dehaene-Lambertz (researcher CNRS, director of the developmental Neuroimaging Laboratory (INSERM / CEA), foreign member of the National Academy of Science (USA), https://scholar.google.fr/citations?user=hB-2od0AAAAJ&hl=fr and https://orcid.org/0000-0003-2221-9081).
The lab is located in Neurospin, Paris-Saclay University, a brain imaging center dedicated to the understanding of the human brain and to the development of brain imaging technologies (e.g. the largest MRI for humans 11.7T). Five laboratories currently share their expertise on site and provide a stimulating research environment. In the lab itself, high-density EEG, MEG and MRI are routinely used in infants and children to study language development, social cognition, attention, mathematical cognition, etc.

Lab Website https://www.unicog.org/ghislaine-dehaene-lambertz/ and http://moncerveaualecole.com/

The cornea is the outermost layer of the eye and it forms the central part of the ocular surface. It is composed of three tissue layers derived from two embryonic germ tissues: a stratified corneal epithelium of surface ectoderm origin, a stromal layer populated by keratocytes and composed of highly aligned collagen fibrils, and a monolayer of endothelial cells covering the posterior corneal surface. The latter two layers derive from the neural crest cells.
Corneal development has been extensively studied through histology and transmission electron microscopy, enabling the tracking of various tissue structures from early embryonic stages to postnatal stages. We hypothesize that specific molecular events and signaling pathways regulate the sequential determination of neural crest cell fate, leading to the formation of distinct corneal layers and the acquisition of functional properties during corneal development. The project aims to establish associations between neural crest cell fate,

The cornea is the clear front surface of the eye. It lies directly in front of the iris and pupil, and it allows light to enter the eye. It provides two-thirds of the optical power of the eye and refracts and focuses incident light on the retina. The cornea serves three main physiological functions: transparency, refractive power, and biomechanical behavior. Its transparency allows over 80% of incident visible light to transmit to the retina. Additionally, the cornea contributes to 2/3 of the eye's refractive power, focusing the image on the retina through its anterior curvature and refractive properties. Lastly, the cornea's biomechanical behavior ensures the maintenance of its curvature despite external stresses, such as changes in intraocular pressure and eye rubbing. These physiological functions are essential for visual acuity. While direct measurement of transparency, refractive power, stiffness, and viscoelasticity during development is challenging due to the small sample size, a precise assessment of corneal structure can determine when these functions become established.

The cornea is composed of three tissue layers derived from two embryonic germ tissues: a stratified corneal epithelium of surface ectoderm origin, a stromal layer populated by keratocytes and composed of highly aligned collagen fibrils, and a monolayer of endothelial cells covering the posterior corneal surface. The latter two layers derive from the neural crests.
The stroma constitutes 90% of the corneal thickness and is composed of collagenous lamellae consisting of tightly packed collagen fibrils embedded in a hydrated matrix of glycoproteins and proteoglycans. The parallel arrangement and the uniform spacing of the collagen fibrils are thought to result in destructive interference of incoming light rays, thereby reducing scatter and promoting corneal transparency. The keratocytes are a population of mesenchymal neural crest–derived cells, sandwiched between the lamellae, responsible for the secretion of the stromal extracellular matrix (ECM), including collagen fibrils and proteoglycans.

The development of the cornea has been extensively studied through histology and transmission electron microscopy, enabling the tracking of various tissue structures from early embryonic stages to postnatal stages. Multiple cell lineages and signaling pathways involved in corneal development have been identified. Investigating the roles of these pathways has been facilitated through the use of knockout or other animal models, associating specific pathways with resulting corneal phenotypes in adults. Consequently, associations between signaling pathways and physiological functions have been established. However, there remains a need to directly investigate the interplay between neural crest cell fate, signaling pathways, and the development of physiological functions. We hypothesize that specific molecular events and signaling pathways regulate the sequential determination of neural crest cell fate, leading to the formation of distinct corneal layers and the acquisition of functional properties during mouse corneal development. We further propose that elucidating this chronological association will offer critical insights into the underlying mechanisms governing corneal morphogenesis and tissue homeostasis.
Objectives of the project
Our project tends to the:
1. Tracking Neural Crest Cell Migration and Differentiation: The first objective is to track the migration and differentiation of neural crest cells during the formation of stromal and endothelial cell layers in both mouse and human corneal development. This involves closely examining how these neural crest cells move and transform into specific cell types that are crucial for corneal structure. By characterizing the chronological sequence of neural crest cell fate determination, we aim to uncover the precise timing and mechanisms that guide the development of these essential corneal components. This knowledge can shed light on how developmental disorders may disrupt these processes and lead to corneal diseases.

2. Investigating Molecular and Cellular Events for Tissue Organization: The second objective involves a detailed investigation into the molecular and cellular events that govern tissue organization within the cornea. This includes understanding processes such as primary stroma secretion, collagen deposition, and the organization of these collagen fibers. Additionally, we will explore tissue innervation, studying how nerve cells become integrated into the cornea. Understanding these intricate processes will provide insights into the factors that influence the structural integrity of the cornea, potentially revealing new avenues for interventions to enhance tissue organization or repair damaged corneas.


3. Examining Corneal Physiological Features: The third objective aims to examine the development of essential corneal physiological features, including transparency, refractive properties, and biomechanical behavior. We will correlate these features with cell fate determination and tissue structural changes. This objective is crucial for understanding how the developmental processes studied in the first two objectives impact the function of the cornea. It will provide valuable insights into how changes in corneal structure relate to its ability to transmit light, refract it properly, and maintain its structural integrity. This knowledge could ultimately inform interventions to improve vision and address issues related to corneal function.

A comprehensive approach involving various techniques such as histological analyses, gene expression profiling, and immunohistochemistry will be employed to achieve these objectives. These methods will allow us to study the chronological dynamics of cell fate determination during corneal development, covering key embryonic and postnatal stages. Additionally, functional assessments, such as measuring corneal transparency and analyzing refractive index changes, will provide critical data to correlate physiological properties with structural changes, further enhancing our understanding of corneal development and function.


Impact and benefit of the project
This research project's findings will provide valuable insights into the spatiotemporal events governing mouse and human corneal development, offering a better understanding of the molecular cues and signaling pathways involved in tissue patterning and homeostasis. Moreover, identifying the chronological association between neural crest cell fate determination and corneal structure establishment may have implications for studying corneal diseases and disorders with developmental origins. Ultimately, this knowledge could pave the way for the development of targeted therapeutic approaches aimed at restoring corneal function and treating vision-related pathologies.
One of the key questions addressed by this research is the determination of neural crest cell fate and their chronological association with the establishment of the corneal structure. The neural crest is a group of cells that plays a fundamental role in the development of many structures in the body, including the cornea. Understanding how these cells influence corneal formation can have significant implications for understanding corneal diseases and disorders that have developmental origins.
By studying these links between the neural crest and the cornea, we could uncover underlying mechanisms that are disrupted in certain medical conditions. For example, this could help explain why certain individuals develop specific corneal diseases and how these diseases form at the molecular level. This information is crucial for the development of targeted treatments aimed at restoring corneal function and addressing vision-related pathologies.
In summary, this research's findings can transform our understanding of corneal development by identifying the molecular mechanisms and temporal

All animals need to constantly tune their energy consumption to the metabolic challenges they face, posed by internal needs, and external environmental factors. Recently, the communication network between the brain and the gut has emerged as a key regulator of this energy balance. Excitingly, I have discovered that such communication, along the “brain-gut axis”, is less static than we anticipated. Namely, gut-neuron function is changed by internal needs, in this case reproduction, and this change is responsible for driving increase in food-intake. We will now investigate the long-standing question on how lifestyle habits can impact neuron function. Taking advantage of state-of-the-art imaging, genetic and molecular techniques, we will ask how changes in this communication perturb physiology. This work will establish, at least in part, why animals respond differently to the same environmental challenge. An attribute of adaptability that can lead to prosperity of failure of a species.

Project background and description
A fundamental question in biology is how the environment of a living organism shapes its energy metabolism. Food intake (influx) and energy consumption (outflux) determine our body’s energy stores. A distorted influx/outflux ratio can lead to pathophysiological conditions, such as obesity, and their detrimental impact on animal health. Regulating this energy flow is central to animal survival and prosperity, yet we only partly understand it. This is due to the difficulty of untangling the web of feedback mechanisms involved.
In recent years, the “brain-gut axis”, has surfaced as a global regulator of energy stores. Still, we know very little about this type of regulatory input. Work over the last decade has begun to shed light on the molecular identity and function of gut neurons, which are thought to have a central role in mediating this inter-organ communication. Yet, it remains elusive if and how environmental factors modulate the function of gut neurons. Answering this question requires a system in which the neuronal subtypes and their discrete functions can be selectively analysed.
The gut innervation in mammals is astonishighly complex and often termed “the second brain”. A major challenge is how to parse this complexity and perform functional studies on individual neuronal subtypes in vivo. Drosophila melanogaster has a less complex neural network and offers a powerful system for addressing these questions. Fruit-flies have been used for over a century by geneticists to spearhead the field of gene regulation. This has resulted in a vast array of genetic tools, that allow control of gene expression, with single cell-resolution. Moreover, flies have many relevant similarities to mammals. They employ the same basic metabolic pathways and their internal organs include a gut, equivalent to its mammalian counterparts. Flies exhibit complex dietary choices and benefit from increased physical activity. For this project we will capitalize on these commonalities and study how enteric neuron function is changed by environmental factors, such as changes in food choice and physical activity. We will use the fly “brain-gut axis” as a model to gain a mechanistic insight on how lifestyle choices shape animal physiology via altering the brain-gut communication.
Own work in context (supervisor)
My work led to the two following advances in the field of the function of the brain-gut axis. (1) Through an anatomical screen, I have generated an atlas of the adult fly brain-gut axis. (2) I discovered an unexpected feature of gut neuron biology. Namely, that gut neurons are functionally plastic, meaning that they change their neural activity patterns in a context depended manner. This change then alters the gut function to allow for an increase in food intake. This constitutes a physiological feature highly relevant for the adjustment of food intake by the animal to meet its energy demands. I explored this in females in the context of reproduction, where mechanisms underlying regulation of feeding are evolutionary conserved across multiple species. For this work, I focused on a set of neurons expressing the neuropeptide myossupressin (Ms, namely Ms-neurons). More recently, I have uncovered that Ms-neurons also change their activity patterns after exposure to nutritional sugars.
The proposed project builds on these findings to explore the role of gut-neurons in metabolic adaptation. We will use two environmental paradigms, namely (a) consumption of a diet high in sugar (HSD), known to lead to fly obesity and (b) an endurance training regime as a paradigm of elevated physical activity.
Hypothesis
Using the fly brain-gut axis we will explore if environmental factors and lifestyle habits, such as consuming an unhealthy diet or leading a sedentary life, are sensed by gut-innervating neurons. We will use a specific group of gut innervating neurons, the Ms-neurons, as a candidate neural substrate for this sensing and explore the neural circuits and molecular mediators involved.

Aims and Experimental Plan
We will use genetic neuronal tracing, optogenetics, high-speed behavioural recordings, 2-photon functional imaging, connectomic and single-cell transcriptomic analyses, to uncover the organisational principles of these environmental sensors distributed along the brain-gut axis.
Aim 1. Identify the relevant neural sensors
We will expose adult flies to either HSD or elevated levels of physical activity. To confirm the potency of these manipulations on animal physiology we will use two behavioural methods as read-outs, the “flypoo” and “flyPAD” assays. Once regimes that produce distinct changes in feeding behaviour and gut physiology have been identified, we will examine these flies for changes in gut neuron function to understand how the output of neuronal activity leads to the behavioural/physiological adaptations. For these we will use genetically encoded calcium sensors. As a starting point, we will profile changes in Ms-neuron activity after exposure to these environmental stressors. Next, we will manipulate the activity of these neurons and test for sufficiency in mediating the physiological changes. By genetically altering their neural activity, we will be able to change the feeding behaviour of these flies to match the ones described as a result of our stressor paradigms (i.e., in the absence of the stressor). This work will uncover the logic of gut-neural sensing i.e., are Ms-neurons “master sensors” of changes in energy demands? How do different metabolic signals change Ms-neuron activity?
Aim 2. Explore the neural circuits involved
To identify relevant neural networks, we will make use of the genetically encoded circuit tracing tools and test for connectivity between Ms-neurons and other candidate groups of neurons, such as taste receptor neurons. The subesophageal zone (SEZ) in the fly brain is important for taste processing and feeding. Taste receptor genes that are expressed in neurons targeting the SEZ were found to be altered by diet and endurance training. I have also shown that Ms-neurons project dendrites to the SEZ making this a site for homeostatic regulation. Elucidating the organizational principles of these gut-neuron circuits will pave the way towards understanding the nature of behavioural changes associated with diet and exercise. This is an entirely novel aspect as to date and the connectivity of gut innervating neurons with the taste processing centres has never been proven.
Aim 3. Uncover the molecular effectors involved in this inter-organ communication
To uncover molecular effectors, we will profile Ms-neurons transcriptionally using single cell approaches. Comparing flies exposed to the environmental stressors to flies grown under normal laboratory conditions will illustrate differentially expressed genes. These will encode factors produced by Ms-neurons in response to our stressors, e.g., specific receptors or biosynthetic neurotransmitter pathways. This approach has the potential to uncover previously unknown pathways that contribute to energy metabolism regulation. After narrowing down the list of candidates, their functional roles will be probed. Using genetic methods, we can downregulate expression of the relevant gene(s) specifically in Ms-neurons. This is feasible in flies as we can readily alter gene function at the single cell level (i.e., by downregulation/overexpression), so that many 10s of genes can be analysed to determine those that are of major functional relevance.

Expected outcome and Impact
This work will lead to new discoveries on basic gut neuron biology. It will uncover, anatomical sites within the brain whose interconnection is important for mediating changes in energy metabolism. Such insights can help us to understand why animals respond differently to the same environmental stressor. An attribute of adaptability that can lead to prosperity of failure of a species.

Paramecium is a unicellular organism that swims in fresh water using cilia. When it is stimulated (mechanically, chemically, optically, thermally, etc), it often swims backward then turns and swims forward again. This “avoiding reaction” is triggered by a calcium-based action potential. As Paramecium is both an organism and a cell, it offers a unique opportunity to investigate the relation between physiology and behavior. This project aims at understanding the physiological basis of one piece of its rich behavioral repertoire: how it escapes from a capillary. Indeed, Paramecium can detect when it is trapped in a dead end and swim backward to escape, and it can also escape from a vertical tube and improve its performance over successive trials. The investigation will use a combination of modern techniques, including behavioral analysis in microfabricated environments, genetic techniques, electrophysiology, modeling.

Paramecium is a unicellular organism that swims in fresh water by beating thousands of cilia. When it is stimulated (mechanically, chemically, optically, thermally…), it often swims backward then turns and swims forward again. This “avoiding reaction” is triggered by a calcium-based action potential. For this reason, some authors have called Paramecium a “swimming neuron” (Brette, 2021). Although it is a single cell, Paramecium displays a surprisingly rich array of behaviors: chemotaxis, gravitaxis, photophobia, avoiding obstacles, thermoregulation, behavior switching, social behavior, adaptation to changing conditions (temperature, chemical composition), escaping from dead ends, as well as learning, including classical conditioning. Physiologically, it has most classes of ionic channels found in neurons, it has both glutamate and GABA receptors, various calcium signaling pathways, etc. Technically, it benefits from many techniques such as whole genome sequencing, RNA interference by feeding, genetic screening, intracellular electrophysiology, calcium imaging. As Paramecium is both an organism and a cell, it offers a unique opportunity to investigate the relation between physiology and behavior, in particular between cellular plasticity and behavioral plasticity.
Specifically, this project consists in investigating how Paramecium escapes when it is trapped. This behavior has been observed in two slightly different conditions. In the first one, Paramecium swims in a narrow capillary that does not allow it to turn, so that it gets trapped at the end. There it will give the avoiding reaction repeatedly, alternatively moving backward and forward against the wall (Kunita et al., 2014). But after a minute, the avoiding reaction suddenly becomes much longer (several millimeters), potentially allowing the organism to escape. Our working hypothesis is that Paramecium detects confinement by secreting some substance that accumulates until it activates some receptor triggering the escape, possibly through purinergic signaling.
The second observed behavior is called tube escape learning (reviewed in Brette, 2021). When a large capillary (much bigger than the cell) is immersed above a swimming paramecium, the fluid and the cell are taken up by capillary action, and the organism swims towards the meniscus by gravitaxis. After about 30 s, Paramecium swims downward and escapes. When the organism is picked up again with the capillary, it escapes after about 15 s, and this improvement is retained for hours.
To investigate these two escape behaviors, the student will use different techniques. For behavioral observation, custom PDMS swimming pools with channels will be built thanks to our collaboration with Laboratoire Jean Perrin (physics lab in Paris). Movies will be analyzed with tracking software, which extracts detailed trajectories. Several kinds of experimental manipulation will be used to investigate the basis of the behavior: manipulation of the ionic composition extracellular medium, pharmacological blocking (of ionic channels, mechanoreceptors, etc), genetic silencing (with RNA feeding), testing with mutants (for example, one mutant does not make action potentials). The genetic part will be done thanks to our collaboration with Eric Meyer, specialist of Paramecium genetics at ENS, Paris. This investigation could be complemented with calcium imaging (thanks to a genetically engineered indicator), and electrophysiology on immobilized cells.
The results will be integrated with a basic biophysical model of the action potential and electromotor coupling developed in the lab (Elices et al., 2022), so as to obtain a detailed model of the phenomenon, relating physiology with behavior.



Brette R. 2021. Integrative Neuroscience of Paramecium, a “Swimming Neuron.” eNeuro 8:ENEURO.0018-21.2021. doi:10.1523/ENEURO.0018-21.2021
Elices I, Kulkarni A, Escoubet N, Pontani L-L, Prevost AM, Brette R. 2022. An electrophysiological and behavioral model of Paramecium, the “swimming neuron.” doi:10.1101/2022.02.15.480485
Kunita I, Kuroda S, Ohki K, Nakagaki T. 2014. Attempts to retreat from a dead-ended long capillary by backward swimming in Paramecium. Front Microbiol 5. doi:10.3389/fmicb.2014.00270

The metabolic state of an organism exerts a large influence on how its brain functions. Neuronal circuits are thought to be under energetic constraints, which may limit their ability to encode brain functions to maximal capacity unless such a barrier is lifted. Moreover, neuronal energy deficits are a key pathological feature in several neurological diseases, thus interventions selectively boosting metabolism in altered circuits may rescue pathology. This project will focus on studying how facilitating neuronal metabolism may enhance brain function. To do so, we will i) generate a Ca2+-permeable channelrhodopsin localized to the inner mitochondrial membrane to boost metabolic rates, on demand, in genetically-defined neuronal populations and ii) we will study the power of boosting neurometabolism in improving memory formation in rodents using genetic and optogenetic strategies. This work will generate groundbreaking knowledge for future studies on metabolism and brain function.

Cognitive function is famously related to one’s metabolic state: even at rest, the energetic demands of the nervous system are enormous [1]. Nevertheless, how metabolism governs neurotransmission and organismic behavior remains largely unknown. Several lines of evidence indicate that trade-offs between information processing and energy consumption have shaped the evolution of neurons and neural circuits, favoring submaximal information processing to avoid excess energy consumption without any clear behavioral benefits [2,3]. However, this implies that in healthy states there exists potential for increasing neural circuit function and thus complex brain tasks could be improved by alleviating the metabolic constraints that curb circuit performance. Supporting this idea, experiments conducted in rodents and humans have demonstrated that the consumption of glucose prior to the establishment of a new memory enhances the process of memory formation [4,5], suggesting that energy constraints could indeed govern brain function. However, the neural underpinnings of this process remain mysterious and a substantial debate persists regarding the validity of these conclusions [6].

Neuronal bioenergy is essential for sustaining the cell biology of the neuron and its capacity to transmit information. Around 90% of energy in neurons is generated by mitochondria, an organelle highly abundant in neurons that can be found in all locations of the complex neuronal arborization, even millimeters away from the soma. Not surprisingly, decades of research have strongly connected mitochondrial dysfunction and neurological diseases [7], suggesting that bioenergetic defects cause neuronal dysfunction and thus are believed to be central to many neurological disorders. However, while theoretically improving cellular energy may help combat neurological diseases, a major limitation lies in the difficulty of precisely targeting energy-deficient neural circuits, which limits the translational potential of this approach.

We hypothesize that the capacity of boosting neuronal metabolism in selected neural circuits holds the key to i) understanding the fundamental governing role of metabolism in shaping brain function and ii) potentially rescuing neuronal pathology caused by bioenergetic defects. Leveraging the established capacity of the host lab in developing novel molecular tools to study neuronal physiology and metabolism [8–10], this project will both work on 1) developing innovative optogenetic tools to manipulate neuronal metabolism on demand using light and 2) dissect the role of metabolic states in governing complex brain functions, such as forming long-term memories. To do so, we will work on two interconnected but separated aims:

i) Development of novel optogenetic approaches to boost neuronal metabolism.

The host lab recently showed that Ca2+ is a potent activator of mitochondrial metabolism [8]. Thus, the PhD student will generate a Ca2+-permeable channelrhodopsin localized to the inner mitochondrial membrane to manipulate metabolic rates, on demand, in genetically-defined neuronal populations. Leveraging the recent development of Ca2+-permeable channelrhodopsins located at the plasma membrane (CapChR2) [11], this aim will test a series of targeting strategies to achieve both (a) selective localization of CapChR2 at the inner mitochondrial membrane and (b) preserved optogenetic function. The host lab has engineered a first strategy for expressing these at the inner mitochondrial membrane using 79 amino acids of MIC60, a protein located at the inner mitochondrial membrane. Preliminary data in cultured neurons shows that 470 nm light increases mitochondrial Ca2+, measured using mitochondrial Ca2+ sensors developed by the host lab (Mito-jRCaMP1b) [8], validating this approach. The student will benchmark the new tools generated against current approaches increasing mitochondrial metabolism with light that have only been validated in the worm, C. elegans [12], by adapting those for mammalian systems. Unlike classical channelrhodopsins, which would dissipate the mitochondrial proton motive force [13], our tools will escalate metabolism with light, providing a novel strategy to causally study how metabolism shapes function. This aim will provide for the first time a series of optical tools that allow boosting neuronal metabolism on demand, enabling the future exploration of the role of metabolism in circuit physiology and brain computation. This technology has the potential to revolutionize the study of brain metabolism in health and disease.

ii) Dissect the metabolic control of a complex brain function: memory formation.

Leveraging the governing role of mitochondrial Ca2+ in controlling metabolism in neurons [8], the host lab has discovered that reducing mitochondrial Ca2+ extrusion rates after firing results in increased neuronal metabolism. Mitochondrial Ca2+ extrusion is controlled constitutively by an exchanger known as NCLX. However, the host lab has discovered that after neuronal activity, the Ca2+/H+ exchanger Letm1 plays an essential role in returning mitochondrial Ca2+ levels back to baseline. Ablation of Letm1 in neuronal mitochondria results in increased mitochondrial Ca2+ transients, causing ATP overproduction in cultured rodent neurons.

The metabolic state of neurons at the Drosophila’s mushroom body (MB) is known to gate long-term memory formation. The host lab found that conditional ablation of Letm1 in the mushroom body of adult flies increased their capacity of forming novel long-term memories (unpublished), suggesting that metabolic boosting facilitates brain function. In aim 2, the PhD student will explore the role of increased neuronal metabolism in supporting memory formation in rodents. We will approach this complex problem with two separate paradigms: i) we will dissect the role of mitochondrial Ca2+ extrusion in controlling associative aversive memories in rodents and ii) we will explore the role of light-mediated increases in neuronal metabolism on facilitating aversive memories using fiber photometry in mice. This aim will leverage the use of robust and controlled manipulation of neuronal metabolism to answer a long-standing question in the field: can brain function be enhanced through metabolic physiology?

In summary, the PhD student will work on an impactful project that will develop and deploy the latest technologies to reveal the intricate relationship between mitochondrial metabolism and neuronal physiology. Outcomes of this research have the potential to revolutionize our understanding of neuronal physiology and brain function, generating an experimental and theoretical framework that will be essential for future studies exploring the molecular link between metabolism and the brain. Additionally, the design of novel optogenetic tools offers a novel strategy to modulate the metabolic rates of neurons and other cell types in the brain on demand and thus may open the door to therapeutic strategies for both i) curbing neurometabolic dysfunction and ii) improving healthy brain function.

References:

1 Pontzer, H et al. (2021). Science 373, 808–812 //
2 Bosman, C & Aboitiz, F. (2015). Frontiers in Neuroscience 9, //
3 Mann, K et al. (2021). Nature 593, 244–248 //
4 Gold, P. (1995). The American Journal of Clinical Nutrition 61, 987S-995S //
5 Messier, C. (1997). Neurobiology of Learning and Memory 67, 172–175 //
6 Peters, R et al. (2020). Neuropsychol Rev 30, 234–250 //
7 Collier, JJ et al. (2023). Trends in Neurosciences 46, 137–152 //
8 Ashrafi, G et al. (2020). Neuron 105, 678-687.e5 //
9 de Juan-Sanz, J et al. (2017). Neuron 93, //
10 Cuhadar, U et al. (2022). bioRxiv 2022.07.03.498586 doi:10.1101/2022.07.03.498586 //
11 Fernandez Lahore, RG et al. (2022). Nat Commun 13, 7844 //
12 Berry, BJ & Wojtovich, AP. (2020). FEBS J 287, 4544–4556 //
13 Tkatch, T et al. (2017). PNAS 114, E5167–E5176 //

The project POD aims to investigate the neural correlates of short-term ocular dominance (OD) plasticity in the primary visual cortex (V1) of adult ferrets. Short-term OD plasticity has been characterised in adult humans, but the underlying neural mechanisms are still unclear due to limitations in non-invasive neuroimaging techniques in humans. We will combine functional UltraSound neuroimaging (a cutting-edge technique) with electrophysiological recordings to investigate functional properties of the ferret V1 and track plastic changes following short-term monocular deprivation (MD). Our ultimate goal is to bridge the gap between neural mechanisms and perceptual outcomes in the context of OD plasticity, with potential clinical applications for treating amblyopia.
The work is organised into three work packages (WP):
WP1: Mapping functional properties of ferret V1
WP2: Tracking plastic changes in visual cortex after MD
WP3: Linking cortical changes to perceptual sensitivity after MD

Neuroplasticity is the ability of the brain to change over different timescales in response to ongoing sensory experiences. The visual system has been historically used as a gold standard system to investigate neuroplasticity, with the paradigm of monocular deprivation (MD) and its impact on ocular dominance (OD). In animal models, OD plasticity has been observed during development and after long-term (days to months) monocular deprivation, which induces a permanent OD shift in favor of the open eye [1]. More recently, studies in humans have found that a particular form of OD plasticity is preserved after development:  short-term (2-2.5 h) monocular deprivation counterintuitively shifts OD in favor of the deprived eye, both at the perceptual [2], and at the neural level [3,4], reflecting visual homeostatic plasticity. Homeostatic plasticity is thought to play a key role in balancing responses to both eyes in primary visual cortex (V1), and as a consequence enhancing neural processing of the deprived eye. This homeostatic boost of the deprived eye is transitory (2-3h after MD), but can become permanent in patients with amblyopia [5], opening the way to new treatments for this widespread neurodevelopmental disease. Understanding the neural underpinnings of short-term OD plasticity is therefore instrumental for potentially important clinical applications.  

While short-term OD plasticity has been investigated in adult humans, the underlying neural mechanisms are still unclear, because of the limited spatio-temporal resolution of the non-invasive neuroimaging techniques available. Here we propose to investigate the neuronal mechanisms of short-term monocular deprivation in the primary visual cortex of the adult ferret with a combination of high-resolution neuroimaging and electrophysiological recordings. 
In a set of preliminary experiments, we set to map functional responses to monocular stimuli over the surface of ferret visual cortex. For this, we used functional UltraSound (fUS) neuroimaging [6], a new neuroimaging modality that provides high spatiotemporal resolution hemodynamic brain responses (100-µm side voxels, 2.5 Hz sampling rate) over extended spatial dimensions (3.6 mm x 15 mm in surface, 2 cm in depth). Notably, it is the first use of fUS for this purpose in the ferret, following up previous characterizations with optical imaging. However, this imaging modality offers distinct advantages, including its unique capabilities in capturing high-resolution hemodynamic responses.
Our laboratory is expert in the use of fUS imaging in the awake ferrets [7,8]. Using the technique, we demonstrated the presence of bands of ocular dominance in the V1 of adult ferrets [9]. We also identified cortical patches responding to both eyes, at locations matching the center of the retinotopic field. Preliminary experiments with MD suggest a boost of responses to the deprived eye in a subset of recordings, consistent with the human literature [3,4].

— Work package 1: Mapping functional properties of ferret V1 at multiple spatial scales
We aim to comprehensively map the functional properties of the primary visual cortex (V1) in adult ferrets across multiple spatial scales. To achieve this goal, we will couple functional UltraSound (fUS) neuroimaging, a cutting-edge technology known for its exceptional spatiotemporal resolution in capturing hemodynamic brain responses, with chronic electrophysiological recordings in the targeted region. 
+ Task 1.1. Defining retinotopic responses with fUS imaging: We will perform fUS imaging to precisely delineate retinotopic responses within the ferret V1. This step is crucial for establishing a foundational understanding of how the visual cortex processes spatial information.
+ Task 1.2. Delineating a region exhibiting binocular responses with fUS imaging: Our next objective is to identify cortical regions that exhibit binocular responses using fUS imaging. Once identified, we will perform electrophysiological recordings with chronic arrays (floating microelectrode arrays, in use in the laboratory) to validate the findings obtained through fUS imaging. This approach will allow us to gain a deeper insight into the neuronal activity within the identified binocular regions of the ferret V1.

— Work package 2: Tracking plastic changes in visual cortex after monocular deprivation
We will then focus on understanding the plastic changes that occur in the visual cortex of adult ferrets following monocular deprivation. We will combine fUS imaging in one hemisphere and electrophysiological recordings in the other hemisphere to investigate the neural mechanisms underlying ocular dominance (OD) plasticity at multiple scales.
+ Task 2.1. Mapping the neural correlates of OD plasticity: We will map the neural correlates associated with OD plasticity at the microscale (electrophysiology) and mesoscale (fUS imaging). In addition, we will track the timecourse of OD plasticity, both during monocular deprivation and after patch removal. This will provide insights into the evolution of neural changes in the visual cortex over time.
+ Task 2.2. Exploring the influence of arousal states during OD: Arousal states is a key player in modulating OD plasticity [10,11]. We will investigate how changes in arousal levels affect neural responses during monocular deprivation. Arousal state will be monitored through different measurements (LFP recordings in olfactory bulb for sleep scoring, heart and breathing recordings, eye tracking).

— Work package 3: Linking cortical changes to perceptual sensitivity after monocular deprivation
We will aim at establishing a connection between the observed cortical changes in the ferret V1 and the resulting perceptual effects of monocular deprivation.
+ Task 3.1. Probing perceptual sensitivity after OD: We will probe perceptual sensitivity in both eyes of the ferrets after OD. We will develop a simple detection task in which the stimulus contrast will be adjusted to assess the perceptual threshold in each eye (deprived and non-deprived).
+ Task 3.2. Testing whether neural changes explain perceptual reports: We will conduct experiments to determine if the neural changes observed in V1 can explain and account for the perceptual reports and sensitivities experienced by the ferrets following monocular deprivation. This step aims to bridge the gap between neural mechanisms and perceptual outcomes.

References:
1. Wiesel T. & Hubel, D. H. Single-cell responses in striate cortex of kittens deprived of vision in one eye. J. Neurophysiol. 26, 1003–17 (1963)
2. Lunghi, C., Burr, D. C. & Morrone, C. Brief periods of monocular deprivation disrupt ocular balance in human adult visual cortex. Curr Biol 21, R538-9 (2011)
3. Lunghi, C., Berchicci, M., Morrone, M. C. & Russo, F. Di. Short-term monocular deprivation alters early components of visual evoked potentials. J. Physiol. 593, (2015)
4. Binda, P. et al. Response to short-term deprivation of the human adult visual cortex measured with 7T BOLD. Elife 7, (2018)
5. Lunghi, C. et al. A new counterintuitive training for adult amblyopia. Ann. Clin. Transl. Neurol. 6, 274–284 (2019)
6. Macé, E. et al. Functional ultrasound imaging of the brain. Nat. Methods 2011 88 8, 662–664 (2011)
7. Landemard, A. et al. Distinct higher-order representations of natural sounds in human and ferret auditory cortex. Elife 10, (2021)
8. Bimbard, C. et al. Multi-scale mapping along the auditory hierarchy using high-resolution functional UltraSound in the awake ferret. Elife 7, (2018)
9. Issa, N. P., Trachtenberg, J. T., Chapman, B., Zahs, K. R. & Stryker, M. P. The critical period for ocular dominance plasticity in the Ferret’s visual cortex. J Neurosci 19, 6965–6978 (1999)
10. Lunghi, C. & Sale, A. A cycling lane for brain rewiring. Curr Biol 25, R1122-3 (2015)
11. Wang, M., McGraw, P. & Ledgeway, T. Attentional eye selection modulates sensory eye dominance. Vision Res. 188, 10–25 (2021)

Early visual areas contain segregated parallel networks of columns carrying highly specific functional signals. However, the mesoscale organization of higher-level cortices such as Infero-Temporal (IT) remains unknown, along with differences between humans and monkeys. Additionally, little is known about how neural representation evolves from posterior to anterior IT cortex. This project aims to study and compare the mesoscale organization of macaque and human IT cortices using sub-mm functional MRI at 11.7 Tesla, combined with stimulus representations from deep neural networks. Specifically designed shim arrays and modular close-fit flexible RF receive arrays will be used to improve temporal-lobe Echo-Planar Imaging signals jeopardized at such high field. The PhD candidate will be advised by experts in Ultra-High-Field (UHF) methodologies, cognitive neuroscience and machine learning, to prepare, run and analyse some of the first ultra-high-resolution functional acquisitions at 11.7T.

Context:
Prevailing theories of primate neocortex organization are dominated by large-scale area parcellations (1) while early sensory-motor areas consist of anatomo-functionally segregated columns/mesoscale units with highly specific connectivity patterns (2). These mesoscale sub-areal units have much greater explanatory power for cortical signal processing than the areas they reside in. However, information about mesoscopic functional modules is limited for the remainder of primate neocortex, especially in higher-level regions such as IT cortex (3). Recently, using sub-mm resolution 3T fMRI, we discovered that two category-selective areas in macaque IT cortex, face-selective ML and body-selective MSB, contain functionally clustered subunits with distinct connectivity properties and functional profiles for 10 different object categories. Electrophysiology results showed similar tuning profiles in clusters of cells within MSB. This suggests that high-level face- and body-selective areas may contain column-like organization like early visual cortex, where functionally similar face/body cells are spatially grouped into mesoscale units with segregated connectivity.
With boosted SNR, UHF MRI opens new opportunities to study the mesoscale organization of primate high-level visual cortex. However, at UHF, the EPI signal shows exacerbated distortions and dropout around the ear canals due to static field inhomogeneity, jeopardizing fMRI of the IT cortex at 11.7 T. At 7T, such artifacts have been significantly reduced with an in-house prototype multi-coil whole-brain shim array called SCOTCH, which is equivalent to a partial 7th-order SH shim set (7, 15) (unparalleled performance worldwide). It should work at 11.7T with similar performance with recent upgrades. A SCOTCH system dedicated to the monkey brain has also been built and will be used here. Additionally, we have recently developed High-Impedance Coil (HIC) technology to increase temporal lobe SNR (by a factor 2) and robustness against subject’s head variability (10). Preliminary results from our human-targeted flexible hat are very encouraging (9).

Objectives:
Several critical questions will be addressed: 1) Is column-like organization a general architecture in macaque and human IT cortices? 2) Does a better-sampled object space lead to finer-grained columnar organization than previously observed? 3) Are there object spaces on the ventral surface of the macaque IT, apart from extensively studied macaque STS? 4) How similar are the macro- and mesoscale organizations across primate species? 5) The object space is represented multiple times along the IT cortex (4); how do the mesoscale neural representation and inter-columnar/areal connectivity evolve from posterior to anterior IT cortex, and how do they compare to the topographic deep artificial neural network (TDANN) representation that models cortical topographical organization (5)? To address these questions, we will use sub-mm fMRI at 11.7 T to scan macaque and human IT cortices with stimuli from the THINGS database (6) covering a wide range of categories evenly sampled from a high-dimensional object space, incorporating CORnet-Z and TDANN (5) to model stimulus representations. We will use RSA, clustering methods, hyperalignment or optimal transport to compare and align representations between primate brains and DNNs.
On the hardware front, we aim to develop a temporal-lobe-specific SCOTCH shim array, since simulation results show field homogeneity can be improved with respect to the whole-brain system (8). We will also develop a modular close-fit flexible HIC receive array for monkeys at 11.7T.

PhD tasks:
1. Participate in designing and building a new shim array dedicated to the human temporal lobes, with the help of an undergraduate student. As mentioned, the proof of concept and methodology have already been demonstrated in the past at NeuroSpin.
2. Participate in the development of a HIC-based receive array dedicated to the macaque temporal lobes at 11.7 T, with the help of another PhD student already present. The proof-of-concept is on the verge of being validated for human brain MRI at 11.7 T.
3. Study the mesoscale organization of human and monkey IT cortex using BOLD fMRI at 11.7T: monkeys will be scanned under anesthesia (optionally awake macaques if authorization is granted – the equipment is prepared and the coils are suitable for awake imaging). A functional localizer (block design) with a resolution of 1.2 mm will be used to localize known category-specific clusters in both species. The main experiment (slow event-related design) uses 200 stimuli evenly sampled over high-dimensional object space from the THINGS database and human/monkey faces/bodies from existing stimulus sets. Human scans will include 3 participants, 2 sessions each. Monkeys will have more sessions. Targeted resolution will be higher than 0.8 mm for humans and 0.6 mm for monkeys at 11.7 T, ensuring robust data for single-stimulus analyses. Resting-state data will be acquired using the same resolution as the main experiment. Apart from the mapping approach, we will use the LAYNII layer fMRI package (11) to generate “columns” to conduct a fine-grained searchlight analysis within the cortical sheet. All experimental designs and processing pipelines have been established in previous or ongoing projects (12–14).

For the first two parts, the candidate will benefit from the help of Alexis Amadon, expert in UHF and head of the MRI hardware development team. For the third part, (s)he will be advised by Minye Zhan (for humans) and Qi Zhu (for macaques), experts in neuroscience and cognition at UNICOG.

In conclusion, this study will provide some of the first high-resolution functional maps of human and monkey brains at 11.7T, the highest field ever reached for MRI of primate brains. It will not only yield macroscale cortical patches, but also reveal mesoscale column-like structures and their functional connectivity in vivo. The stimuli evenly spanning the object space may provide improved alignments of homologue patches/columns between monkeys and humans.

References:
1. M. F. Glasser et al., Nature. 536, 171–178 (2016).
2. S. He, R. Dum, P. Strick, J. Neurosci. 13, 952–980 (1993).
3. J. H. Kaas, Proceedings of the National Academy of Sciences of the United States of America. 109 Suppl 1, 10655–10660 (2012).
4. P. Bao, L. She, M. McGill, D. Y. Tsao, Nature. 17, 4302–6 (2020).
5. H. Lee et al., “Topographic deep artificial neural networks reproduce the hallmarks of the primate inferior temporal cortex face processing network” (preprint, Neuroscience, 2020), , doi:10.1101/2020.07.09.185116.
6. L. M. Stoinski, J. Perkuhn, M. N. Hebart, Behav Res (2023), doi:10.3758/s13428-023-02110-8.
7. B. Pinho Meneses et al., NeuroImage. 261, 119498 (2022).
8. B. P. Meneses, A. Amadon, Magn Reson Mater Phy. 35, 923–941 (2022).
9. J.-F. Gapais, thesis, NeuroSpin (2023).
10. B. Zhang, D. K. Sodickson, M. A. Cloos, Nat Biomed Eng. 2, 570–577 (2018).
11. L. R. Huber et al., NeuroImage. 237, 118091 (2021).
12. M. Zhan, C. Pallier, A. Agrawal, S. Dehaene, L. Cohen, Sci. Adv. 9, eadf6140 (2023).
13. A. Koizumi et al., eNeuro, in press, doi:10.1523/ENEURO.0429-19.2019.
14. X. Li, Q. Zhu, I.D. Popivanov, R. Vogels, W. Vanduffel, In preparation.
15. Bruno Pinho Meneses, Alexis Amadon, Static-magnetic-field shimming coil system for magnetic resonance imaging (2020) (https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2020212463).
16. Q. Zhu, W. Vanduffel, Proceedings of the National Academy of Sciences of the USA. 116, 2306–2311 (2019).
17. X. Li, Q. Zhu, W. VaVanduffel, Progress in Neurobiology. 211, 102230 (2022).
18. A. Amadon, M. Cloos, Method and apparatus for compensating for B1 inhomogeneity in magnetic resonance imaging by nonselective tailored RF pulses (https://patentscope.wipo.int/search/fr/detail.jsf?docId=WO2011128847).
19. M. A. Cloos et al., Magn. Reson. Med. 67, 72–80 (2012).

Mutations in the SPG11 gene are responsible for the most frequent autosomal recessive form of hereditary spastic paraplegia, a neurodegenerative disease of genetic origin characterized by rigidity in lower limbs, often associated with cognitive impairment. The neurodegenerative phenotype is caused by lysosomal dysfunction due to the absence of the SPG11 product, spatacsin. However, the disease is also suggested to have developmental abnormalities. The objective of the project is to investigate whether the lysosomal dysfunction caused by the loss of spatacsin may also contribute to developmental defects. To address this question, we will use a model of cortical organoids derived from induced pluripotent stem cells of SPG11 patients or isogenic controls.

Mutations in the SPG11 gene are responsible for the most frequent autosomal recessive form of hereditary spastic paraplegia, a neurodegenerative disease of genetic origin characterized by rigidity in lower limbs, often associated with cognitive impairment. This disease is due to the loss of function of the SPG11 product, spatacsin (Stevanin et al, 2007). This protein is implicated in lysosome membrane trafficking after induction of autophagy, but also in basal conditions. We also obtained evidence that lysosomal calcium homeostasis is altered in absence of spatacsin. Furthermore, we demonstrated that loss of spatacsin function leads to accumulation of some lipids such as cholesterol or gangliosides in lysosomes, which could contribute to neurodegeneration (Boutry et al, 2018; Boutry et al, 2019).
There are some evidences that loss of spatacsin could also have an impact during brain development. Loss of spatacsin impaired proliferation of neural progenitor cells, and cortical organoids derived from induced pluripotent stem cells of SPG11 patients appeared smaller than control organoids (Pérez-Branguli et al, 2019). Furthermore, recent evidence from the host laboratory suggested the presence of slight alterations in the organization of the cortex in mouse at birth. However, it is not known how the loss of spatacsin impairs the development of the cortex.

The project aims at investigating how the lysosomal dysfunction caused by loss of spatacsin contributes to alteration in cortical development. We will use as a model cortical organoids derived from induced pluripotent stem cells of SPG11 patients or isogenic control. Preliminary data from the host laboratory have shown that loss of spatacsin alters the formation of cortical organoids, with an impairment of the proliferation of neural progenitor cells as well as impaired cortical lamination. Furthermore, we obtained evidence that restoring calcium homeostasis in lysosomes restored the development alteration in SPG11 organoids.
Using single cell RNA sequencing as well as cell biology approaches and imaging, we will investigate the signaling pathways linking lysosomal dysfunction to the aberrant cortical development in SPG11 organoids, and we will explore how restoring lysosomal calcium homeostasis rescue the phenotype.
In the long term, the possibility to restore normal cortical development will allow us to investigate whether abnormal development has an impact on the neurodegenerative phenotype that occurs during adulthood.


References
Boutry M, et al. Loss of spatacsin impairs cholesterol trafficking and calcium homeostasis. Commun Biol. 2019 Oct 17;2:380. doi: 10.1038/s42003-019-0615-z
Boutry M, et al. Inhibition of Lysosome Membrane Recycling Causes Accumulation of Gangliosides that Contribute to Neurodegeneration. Cell Rep. 2018 Jun 26;23(13):3813-3826. doi: 10.1016/j.celrep.2018.05.098.
Pérez-Brangulí F, et al. Human SPG11 cerebral organoids reveal cortical neurogenesis impairment. Hum Mol Genet. 2019 ;28(6):961-971. doi: 10.1093/hmg/ddy397
Stevanin G, et al. Mutations in SPG11, encoding spatacsin, are a major cause of spastic paraplegia with thin corpus callosum. Nat Genet. 2007 ;39(3):366-72.

In recent years, neuroscience and AI have met in two ways. First, artificial neural networks have achieved such good performance that some now see them as plausible models of human behavior. Second, even functional similarities have been claimed, on the basis that network activations can be used to predict brain fMRI data. However, mere prediction doesn’t tell us what makes model representations similar to brain activations, and in which aspects they diverge.

The goal of this project is to understand biological and artificial neural representations in parallel. For this, we develop a new framework of analysis applicable to both. We will test the framework using language as a case study, tracking the representations of linguistic features and operations in brains and neural networks, and comparing them. This will provide a detailed description of what makes a brain and a model, or two models, or two brains, process language alike or not alike.

OPAQUE AI. Neural language models have demonstrated impressive abilities in numerous tasks, including fundamentally human tasks. This culminates today in the capacity to generate relevant, human-like text. However, their underlying neural mechanisms remain unknown. Some models consist of billions of parameters, making it challenging to understand how exactly they perform the tasks they excel at (e.g., generate language), and how the computational solutions they find compare to actual human processing, if at all.

NEUROSCIENCE MEETS AI. Despite the lack of mechanistic understanding of the models, alignment is found between the neural activations observed in actual brains (e.g., fMRI data), and the artificial activations extracted from such highly performing artificial neural networks. For instance, Eickenberg et al. (2017) show that during a task of object recognition, the brain fMRI data are well predicted from corresponding artificial neural network activations. Similar results have been obtained in other domains, such as speech and music perception (Kell et al, 2018), semantic and syntactic processing (Caucheteux et al., 2021; Pasquiou et al., 2022) and even used for the improvement of AI models, by better aligning them with brain data (Toneva et al., 2019).

GOAL. The goal of this project is to go beyond mere brain-model alignment and to provide a detailed understanding of the aspects in which model and human neural representations resemble each other or differ. We will focus on language processing as a case study and develop a new machine-learning method that identifies which properties of the linguistic input (e.g., tense, word frequency, sentence structure, agreement realizations) best explain observed differences among high-dimensional neural representations. We will advance along three main axes:

> Subgoal 1 (single model) - Identify which properties of the linguistic input best explain observed differences among high-dimensional representations of a neural language model.
> Subgoal 2 (single brain) - Identify which properties of the linguistic input best explain observed differences among high-dimensional representations of different brain regions.
> Subgoal 3 (brain-model alignment) - Identify which properties of the linguistic input best explain observed differences between high-dimensional representations of a neural model and those of the human brain.

METHODOLOGY

BACKGROUND. In both human and model studies, two general approaches to study neural representations can be discerned: “decoding” and “encoding” methods (see, King et al., 2020, for a review). In the decoding setup (aka, ‘diagnostic probes’), the goal is to predict linguistic features (e.g., grammatical number, grammatical complexity) from neural activations, typically using machine-learning classifiers (Hupkes et al., 2018; Arps et al., 2022). In the encoding setup, the arrow is reversed, whereby the goal is to predict neural activity from a set of features. For this, regression methods are typically used, where linguistic features and neural activity are the independent and dependent variables, respectively.

CURRENT LIMITATIONS. Decoding and encoding methods have different merits and limitations. One main limitation of decoding methods is that the decodability of a given feature doesn't ensure its causal role. For instance, a certain feature can be decodable from neural activations not because it has a mechanistic role but only because it correlates with another feature that has such a role.
In encoding models, this limitation can be addressed to some extent by introducing into the model both the feature of interest and confound features, testing their relative importance in predicting neural activity. On the other hand, a common limitation of encoding models is that they are uni- rather than multi-variate, unlike decoding methods. In encoding methods, the regression typically predict very local patterns independently of each other: the neural activity recorded from a *single* unit of the model, from a single electrode in the brain or from an fMRI voxel (Caucheteux et al., 2021; Pasquiou et al., 2022; Caucheteux and King, 2022). Therefore, encoding methods are limited in their ability to study distributed representations across many units.

METRIC LEARNING TO STUDY COMPUTATIONAL REPRESENTATIONS. We introduce a simple approach, which preserves the good from both worlds, by extending encoding methods to the multivariate case. In a nutshell, we first compute pairwise distances between the neural representations of a given set of stimuli (e.g., sentences). These neural distances indicate for which pair of stimuli the model finds critical similarities or differences.
We then model these neural distances in a metric-learning fashion, by optimizing a feature-based metric function. The parameters of this feature-based metric function are weights over each feature and, after optimization, these parameters reveal which features are predominant in the model representations in the first place. This can be done globally or for regions of interest, e.g., neural networks layers. Possible confounding linguistic features can be disentangled through ablation studies (leaving one linguistic feature at a time). Overall, this approach allows to study distributed representations of linguistic features, in a neural network model. It reveals which linguistic features are represented, and where.

CASE STUDY OF GRAMMATICAL MECHANISMS. We will test the metric-learning approach in the case of language processing, focusing on how neural networks encode grammatical information, and in particular sentence structure. This question was studied in neural networks, and to an even greater extent in humans, investigating which brain regions encode grammatical structures (Ben-Shachar et al. 2004; Pallier et al., 2011; Fedorenko and Thompson-Schill, 2014; Friederici, 2017), what are the possible neural mechanisms (Desbordes et al.), and whether they encode grammatical information separately from semantic information (Dapretto; Siegleman; Pasquiou). However, despite decades of research, how sentence structures are encoded in neural networks, biological or artificial, remains an open question.

Subgoal 1 - Analyzing an artificial neural network. We will apply the method above to study the linguistic representations of artificial neural networks.

Subgoal 2 - Analyzing a brain. The exact same method described above can be used to analyze brain data. Above, linguistic features are weighted to best fit the distance between the activations generated by any two sentences in a given layer. For brains, we will find the optimal weighting of linguistic features to best fit the distance between the fMRI activations generated by any two sentences in a given brain area.

Subgoal 3 - Comparing brains and artificial neural networks. The method of analysis is thus entirely parallel for brains and neural networks. It provides a weighting of linguistic features per brain region/network layer. We can thus compare these regions and layers on the basis of the weighting they correspond to, and find analogous types of representations from a linguistically informed point of view.

A major goal of sensory neuroscience is to understand how neurons can extract complex features from natural scenes. Even as early as in the retina, this is still an open challenge. We have recently shown that, during natural image stimulation, many ganglion cells can change their selectivity for light increase (ON response) or decrease (OFF response) depending on the natural image. These ON-OFF selectivity changes are compatible with the robust encoding of a more abstract feature: contrast.

However, we do not understand how the contrast signal is built up from the different components of the neuron’s receptive field. Here, we will study the center and the surround contribution of neurons receptive fields when responding to natural images. We will use a new perturbative approach and deep convolutional neural network models to reveal the putative nonlinear combination of center and surround signals when a neuron process natural images, using multi-electrode recordings in mice retina.

A major goal of sensory neuroscience is to understand how visual areas process natural scenes. A long-standing hypothesis is that each visual area extracts features from the visual scene that will be essential for this task. However, feature extraction has mostly been characterized with simple artificial stimuli, and it is difficult to understand what features are extracted by neurons during complex, natural scene stimulation. An important question in the field is thus to understand what features are extracted from more complex, natural stimuli, and how.

However, even in the retina, this question remains unanswered. The retina is a complex transparent tissue consisting of several layers of neurons that covers the back of the eye. Light passes through the overlying layers and stimulates the photoreceptor cells. Between photoreceptors, the retinal input, and ganglion cells, the retinal output, there are several layers of cells shaping the retinal computations. And it is still unclear how different cells contribute to retinal processing.

To process the spatial information in a visual scene, each ganglion cell has a receptive field center, the region of visual space whose stimulation evokes the strongest responses. But ganglion cell responses can be also modulated by stimuli outside of their receptive field center, a classical phenomenon termed surround modulation. This is a ubiquitous feature of sensory computation. In the retina, surround modulation is present in most ganglion cells and might help enhancing edges in a visual scene and optimize information transmission to the brain.

In this thesis project, we want to understand how ganglion cells integrate spatial information within the retinal circuit to be able to extract complex information from natural scenes. We are interested in separating the contributions from the center and surround of the cell’s receptive field. For this, we will focus in a particular ganglion cell type we recently described as “contrast encoding” (Goldin et al., 2022). This cell increases its response either to light increases or decreases depending on the context (i.e. the natural image used as stimulus). To reveal the center vs surround contributions to the nonlinear selectivity of these cells, we will probe them with perturbations added on top of natural scenes.

We will record ganglion cells with a multi-electrode array, and stimulate the retina with flashed natural images where small noise-like perturbations will be added. We will use natural images as a reference stimulus and add a small amplitude checkerboard to it. We will present many times the reference stimulus, each time with a different perturbation added on top of it. The checkerboard amplitude added to the reference image will be calibrated to be large enough to evoke a significant change in the response, but not so much to remain a small perturbation of the reference image. These perturbations will evoke small changes on the responses of retinal ganglion cells.

For each cell and reference image, our perturbative approach will allow us to calculate the gradient of the stimulus-response function of a neuron. In other words, we will be able to obtain the best local linear approximation of the stimulus-response function. We have found that these gradients depend on the images, even for the same cell, preferring light increments for one natural image and light decrements for another. This is in contrast with the classical view obtained by flashing spatially uniform stimuli on the retina, where ganglion cells are either ON (sensitive to light increase) or OFF (sensitive to light decrease).

Here, we will isolate the role of the surround during the processing of flashed natural images. For this we will use two conditions: in the normal condition, the perturbed natural image will cover the entire visual field. In the “masked” condition, the stimulation will be restricted to the receptive field center of a recorded ganglion cell, and the rest of the visual field will remain gray. The difference in the response of ganglion cells between these two conditions will be a clear signature of the role of surround modulation.

In order to fulfill this project we need to detect online the center of the receptive field of ganglion cells, to be able to target the masking of the natural image to its surround. Due to time limitations on the experiment length, we will use a novel closed-loop approach that uses fewer trials than traditional ‘open-loop’ approaches, (Goldin, Virgili & Chalk, 2023). This method allows to discriminate between models, and will thus enable us to select the specific cell types that we aim in this study.

Finally, we will use deep convolutional neural network (DCNN) models to model ganglion cell responses to natural images in order to gain insight into the role of the retinal circuit components. DCNN are artificial intelligence models, and they are the most successful biologically inspired models of the visual system. We encounter them in a daily basis, each time we use a picture tagging software (google, facebook, etc.). They have shown impressive performances emulating visual function (e.g. object recognition) or predicting the response of single neurons in visual areas.

The candidate for this project should have a degree in neuroscience, physics, engineering, biology or a related relevant background. The ideal candidate needs to have a strong will to be trained in experimental and quantitative systems neuroscience. Any previously acquired skills will be taken into account, but this is not exclusive.

The time organization of project will go as follows.
From months M1-M6: reading the literature, experimental training, producing the stimulus, setting the online protocol for center-surround detection
M7-M12: recordings to fine tune the stimulus, training in the analysis of extracellular recordings and data, first data analyses, participation in further training in a summer school
M13-M18: recordings of the masked and unmasked perturbations, thorough data analysis and data preparation for modeling
M19-M24: further recordings, presentation of preliminary results in international conference/s, analysis of modeling results, preparation of a last round of experiments
M25-M30: last round of recordings and analysis using the full closed loop protocol, further modeling, manuscript preparation training and writing
M31-36: manuscript submission, thesis writing and defense

Our experiments, analysis and modeling will allow teasing apart different hypotheses of sensory processing. On the one hand, the surround may act modulating the gain, i.e. to normalize the input coming from the center. Alternatively, when surround is stimulated it may change the feature extracted by ganglion cells. Deciding between these hypothesis is a key step to better understand the role of surround modulation to process natural stimuli. Our novel approach has the ability of challenging classical views of retinal processing derived from artificial stimuli, and to uncover that retinal processing is strongly dependent on visual context. Our results, if successful, may be applied to other sensory areas of the brain, as for example in the somatosensory cortex, for which we have an ongoing collaboration.

References:
Goldin MA, Lefebvre B, Virgili S, Pham Van Cang MK, Ecker A, Mora T, Ferrari U, Marre O. Context-dependent selectivity to natural images in the retina. Nature Communications. 2022; 13,1

Goldin MA, Virgili S, Chalk M. Scalable gaussian process inference of neural responses to natural images. PNAS. 2023; 120,3

Uncovering the neural mechanisms behind subjective experience is a significant challenge of neuroscience. Our lab has demonstrated the existence of dynamically recurrent brain patterns (BP) characterizing the state of consciousness. Using fMRI, we have shown that in conscious subjects’ brain dynamics is predominated by BP characterized by long-range functional communication and anti-correlation between areas. In contrast, in unconscious subjects, the predominant BP are associated with sparse and low inter-areal communication.

This project aims to understand the nature and limitations of conscious processing according to ongoing BP. The candidate will work towards (1) better defining the brain dynamics of inter-area coordination to describe the BP accurately and their association to subjective experience and (2) testing in real-time healthy subjects and disorders of consciousness patients’ capacity to consciously integrate information according to the ongoing brain dynamics.

1 Introduction

Consciousness is central to subjective human experience and a key pillar of human society. It traverses our healthy daily experiences and defines most neurological and psychiatric diseases. Uncovering the neural mechanisms that allow subjective experience is a major challenge of neuroscience. The study of consciousness has significantly improved in the last 20 years(Michel et al. 2019). This topic becomes critically important in the clinics regarding patients with disorders of consciousness (DOC).
Our lab has recently demonstrated that distinct recurrent brain patterns obtained from functional MRI (fMRI) are directly associated with different states of consciousness. In a series of works using human and non-human primate functional neuroimaging data(Barttfeld et al. 2015; Uhrig et al. 2018; Demertzi et al. 2019), we have shown that conscious states are associated with the exploration of a rich repertoire of distinct brain patterns. Within this rich dynamic exploration, conscious individuals exhibit a combination of two types of brain patterns: i) 'high' with long-range functional cortico-cortical communication (correlations and anti-correlations) between brain areas, and ii) 'low' with sparse and low inter-areal communication that mirror anatomical connectivity. In contrast, unconscious individuals (anesthesia and DOC) only express the latter.
The overarching goal of this project is to understand the nature and limitations of conscious processing according to the ongoing brain dynamics. The hypothesis is that the dynamics of brain patterns will shape the dynamics of states of consciousness and subjective experience. Testing this prediction will directly impact understanding the neuronal mechanisms of subjective experience and the diagnosis, prognosis, and potential therapeutics of patients with DOC.

2 Context and positioning: can we track the state of consciousness from brain dynamics?

2.1 Brain patterns dynamics and state of consciousness
Using dynamical fMRI connectivity, We tested the association between recurrent functional connectivity patterns and states of consciousness. We acquired resting state data for this objective in awake and anesthetized non-human primates (NHP) (Barttfeld et al. 2015; Uhrig et al. 2018). Our results showed that in conscious conditions, brain dynamics was characterized by a sequential exploration of a richer repertoire of functional configurations – including "high" and "low" -. In contrast, during anesthesia, the 'low' configurations are predominant. These results gave rise to potential clinical applications for detecting awareness in anesthesia and brain-lesion patients.
More recently, we extended these results to determine if the identified patterns of coordinated brain activity could distinguish the state of consciousness in the clinics. In a multicentric study, we performed a dynamic analysis of fMRI data in healthy humans and DOC patients (Demertzi 2019). Analogous to the results in NHP, conscious individuals exhibited a rich exploration of brain configuration and the predominance of the previously described 'high' brain patterns. In contrast, vegetative state patients mainly explored the 'low' pattern. Our results establish that human consciousness relies on the brain's ability to sustain rich dynamic configurations of neural activity that lose prevalence in unconscious states.

2.2 Brain patterns dynamics and subjective experience

More recently, we demonstrated that the temporal sequence of brain patterns synchronizes across healthy individuals when attentive to a movie (Türker et al. 2022) – an effect lost during resting state or for scrambled movies -. This result goes beyond the previous reports of cross-subjects correlation of region interest activity (Naci et al. 2014; Hasson et al. 2004) and demonstrates synchrony at the level of whole brain configurations, suggesting that the described brain patterns reflect the individuals' subjective experience.

3. Workplan

Primary objective: To understand the interaction of ongoing brain patterns on the dynamics of states of consciousness and subjective experience.

Using real-time EEG/fMRI, the candidate will do unprecedented experiments testing the cognitive processing of external stimuli, contingent on the ongoing brain patterns.

This project aims to understand the links between healthy participants' brain activity patterns and conscious information processing. First, we will record subjects in resting state (RS) to individually validate the expected recurring brain activity patterns using non-supervised clustering. Next, those patterns will be detected using real-time fMRI.

All participants will undergo four sessions of combined fMRI and EEG recording. In each session, the experiment will start with a first block of 5-minute RS, followed by a second block of 15-minute RS where we will interrupt the subjects and sample their subjective state of consciousness (see objective 2), a third block of 60 minutes with a specific cognitive task (see objective 1), and a final fourth block of 5-minutes RS. The four experimental sessions will only differ regarding the task performed in the third block.

Objective 1: How is the capacity to integrate information influenced by ongoing brain activity patterns?

In the third block of the experiment, participants will be asked to perform cognitive tasks designed to test the limits of different dimensions of conscious information integration while exhibiting different brain patterns. In all cases, the stimulation will be triggered by the detection in real-time of targeted brain patterns (e.g., 'high' or 'low' patterns) from the ongoing fMRI activity (with the help of external collaborator Vincent Tascherau from the University of Montreal expert in real-time fMRI processing). The differential neural response to the stimulus according to the ongoing brain pattern will be analyzed using behavioral responses, EEG-evoked activity, and dynamics of brain patterns after the stimulus. We predict a modulation of the capacity of conscious information integration in all tasks with different cognitive characteristics of 'high' and 'low' brain patterns.
We plan four different cognitive tasks for the third block: (1) a perceptual detection and metacognition task, adapted from (Sergent 2020) and designed to test the auditory perceptual threshold and metacognitive judgment; (2-3) two lexical decision tasks; with target stimuli preceded by a priming cue which can be either informative or not, allowing us to test expectancy and abstract discrimination, and (4) a temporal integration task, adapted from(Otto, and Herzog 2006) and designed to test the limits of temporal integration of information parametrically. We will study how ongoing brain patterns modulate performance in all those tasks.

Objective 2: Are there specific mental contents associated with brain activity patterns?

When not engaged in a task, our consciousness entertains stimulus-independent thoughts. This phenomenon called "mind-wandering" has been investigated in recent studies, showing that neural activity in the 'default mode network' correlates with the intensity of wandering (Christoff et al. 2009). While detecting the target brain patterns, we will interrupt the subjects from their rest and interrogate them on their current mental content. To sample the participants' experience, we will use a forced choice between proposed mental states (subjective test) and a rapid word association task (objective test). Previous evidence indicates that word associations reflect the current mental state of the subject (Van den Driessche et al. 2019). We will confront the results of the objective and subjective tests with the current brain activity, allowing us to identify the psychological correlates of brain patterns.

Memories of salient experiences can last a lifetime but we only retain the most pivotal elements of an event, while other components of the original memory tend to fade.
The brain's ability to selectively preserve certain memory details is disrupted in disorders like Alzheimer's or PTSD, where memories are either lost or excessively vivid. Exploring how the brain makes these critical decisions is essential to understanding long-lasting memory consolidation in typical and pathological memory consolidation.
In this research project, the PhD candidate will adopt an innovative approach to investigate the role of distinct subpopulations of hippocampal and cortical neurons during the encoding process, to determine where salient aspects of memories are processed in the brain.

During the encoding of a new memory, structural changes in specific subsets of neurons active during the experience, called engrams, result in these neurons becoming sufficient and necessary for the subsequent recall of the formed memory (1). 
Extensive literature has shown that, after encoding, these memory traces are both consolidated at a synaptic level and re-organized in brain circuits through a process called system consolidation. According to this model, memory traces can be recalled at different time points with the support of different networks of brain regions (2). In particular, it has been shown that memory traces are initially stored in the hippocampus and their retrieval at recent time points, when the memory of the episode is enriched with details, is dependent on this region (3). Subsequently, they endure and integrate into the cortices becoming part of the remote memory storage. At this point, recall appears less detailed and more generalized. Consistent with this hypothesis, visualization of engram cells in the prefrontal cortex (PFC) showed that they respond preferentially to the aversive information (i.e. footshock delivery in a contextual fear conditioning) already during memory acquisition but are reactivated only at remote time points while remaining silent in between. 
Several pieces of evidence suggest that in rodents engram cells in the cortex are “early tagged” during encoding despite the fact that this region is not yet necessary for memory expression (4,5). Changes in plasticity at a pre-synaptic level have been observed in the anterior cingulate cortex (ACC) already 24 hours after training, and activation of these neurons during encoding promotes post-synaptic remote-like plasticity processes even at recent time points.
These results suggest that episodic memory encoding drives the early selection of cortical engram cells that will store a specific memory trace in the future. However, the mechanisms through which only some aspects of the memory episode are selected to be transferred in the cortex and become part of the remote memory storage remains unknown. Do the cortex and the hippocampus encode different aspects of the same memory during memory acquisition? Or do they initially encode all details present at the time of memory acquisition, with the hippocampus later retaining the more episodic information and the cortex retaining the more semantic details? Engram labeling technology coupled with optogenetics allowed the selective manipulation of brain cell populations holding specific memory representations to determine their involvement in memory processes. However, given the temporal limitations of the labeling technology used so far, it has been impossible to target different ensembles of cells active at specific moments within the same task, since the time window of tagging is in the range of hours or even days. 

Here the Ph.D. candidate will use FLiCRE (6), a novel tool based on optogenetics, that enables labeling of neuronal ensembles associated with specific events, precise stimuli presented, or behaviors observed within a task. This cutting-edge system induces the expression of excitatory or inhibitory opsins in active cells when both calcium and light are present, with a temporal precision of seconds. By limiting labeling to small discrete time intervals, such as the onset of stimulus delivery or behavioral response, this technique allows to tag and manipulate for the first time the multiple components that differently contribute to the final recent and remote representation of the memory.
In particular, during contextual fear conditioning, mice learn to associate a context and all the details present at the time with an aversive experience. During this task, a heterogeneous response is observed in different hippocampal neurons: some cells are activated during footshock presentation, others are preferentially active when the animal is freezing, others when the animal is exploring the context, etc.
In REMGRAM, using FLiCRE, the Ph.D. student will determine whether these components are all part of the engram in the first place and whether they play different roles in the storage of recent versus remote memory in CA1 and in the ACC. Furthermore, by artificially mimicking or blocking the activity mechanisms that underlie the process of remote memory formation, the candidate will try to induce a recent memory to become ‘remote-like’ (generalized, aCC-dependent) and, vice-versa, restricting a memory to its recent-like form (episodic, HPC-dependent).
The main objective of this research is to reveal for the first time the mechanisms through which a component of a memory (a subpopulation of engram cells) is selected to become part of the long-term storage of memory in the brain.
The 3 main objectives are:

1. Histological quantification of sub-engram cell reactivation in ACC and HPC at recent and remote time points.
Ca++ imaging recordings will be used to observe the online activation of ACC and CA1 engrams at recent and remote time points.

2. Determining the functional role of ACC and CA1 sub-engrams in recent and remote memory recalls. 
To this purpose, FliCRE inhibitory opsins will be injected in WT mice and optic fibers implanted in animals’ CA1 and ACC. The inhibition of the different memory components (shock cells, freezing cells, no-shock cells) will inform us about the necessity of those in the storage of recent and remote memories.

3. HPC-ACC sub-engrams manipulation to revert generalized/remote to more specific/recent memories and vice versa.
To this purpose, After CFC, sub-engrams in ACC and CA1 will be optogenetically stimulated in homecage to induce the coordinated replay that has been shown to mediate the transition of memory from the HPC to the cortex.
In a different cohort of mice, to prevent a recent memory from becoming remote and preserve its episodic and contextual information, we will inject an inhibitory DREADD in HPC or in the ACC. Then, mice will be injected daily with Clozapine-N-Oxide (CNO) to inhibit the communication between the two regions, impairing the functionality of either the input or the output region. Mice will be then tested after 28 days to assess whether the memory holds recent-like features.

REMGRAM requires the integration of different techniques and technologies. It will combine behavior with optogenetics, circuit connectivity, and imaging of neuronal activity.

Expected outcomes: 1 peer-reviewed publication for aim 1 and 1 peer-reviewed publication for aims 2 and 3.

Skills that will be acquired during the Ph.D.: Behaviour analysis in mice; Optogenetics and miniaturized microscope surgery and behavioral application; Data analysis using Matlab and R code; Statistical analysis; Management and writing up of the project under mentor supervision; Participation in conferences to present the scientific data.

1. Josselyn, S.A. et al., (2015). NatureReviews Neuroscience 16, 521-534
2. Frankland, P.W., & Bontempi, B. (2005). Nat Rev Neurosci 6, 119-130
3. Vetere, G. et al., (2017). Neuron 94(2):363-374.e4
4. Vetere, G. et al., (2019). Mol Neurobiol 56, 8513-8523
5. Vetere, G.et al. (2011). PNAS 108(20), 8456–8460
6. Kim, C.K. et al.(2020). Cell 183, 2003-2019.e2016

The complement, a key component of the immune system, refines brain circuits in the developing thalamus by shaping microglia-neuron interactions. Its role in the maturation of the cortex is controversial. The 6-layered neocortex of mammals contains several different types of pyramidal cells (PCs), each type populating a specific layer.
We have observed that the complement system remodels synapses in layer 3 PCs, but not in layer 5 PCs of developing cortex. These intriguing results suggest that complement-dependent refinement is neuron type-specific. Importantly, microglial cells are diverse, but the function of different subtypes of microglia remains largely unknown. We hypothesize that specific microglia subtypes target layer 3 PCs for complement-dependent refinement.
The PhD project will assess whether neuronal identity determines complement-dependent synapse refinement, and whether distinct subtypes of microglial cells mediate neuron type-specific synaptic refinement.

Host Team and Institute
The host team is located in the Latin Quarter in Paris at the Institut du Fer à Moulin (IFM). The IFM is devoted to the study of the development and plasticity of the nervous system. The team is headed by Corentin Le Magueresse and Anne Roumier, and focuses on immune-brain interactions.

Candidate
For this project we are looking for a highly motivated student with excellent academic credentials. Techniques include ex vivo patch-clamp, in utero electroporation, stereotactic injections of lipid nanoparticles, immunohistochemistry, confocal imaging, 3D image analysis. Prior knowledge of the techniques is not required, but the student should be eager to learn and possess good teamwork skills.

Context and objectives
Neuronal connections are substantially reorganized during adolescence in the mammalian brain. This process involves large-scale synaptic over-production during the early postnatal period, followed by large-scale synaptic elimination (“synaptic pruning” or “synaptic refinement”). The 6-layered neocortex of mammals contains several different types of pyramidal cells (PCs), each type populating a specific layer. Notably, synaptic pruning in neocortex differs across layers: in the neocortex of adolescent mammals, cortical synapses of layer 3 PCs (L3 PCs) undergo considerably more synaptic pruning than those of layers 5 and 6 (L5 and L6 PCs).
The complement system, a key mediator of innate immunity, plays an essential role in the refinement of excess synapses by microglia in the developing thalamus through the tagging of synapses. Tagged synapses are recognized by complement receptor 3 (CR3) which is expressed by microglial cells(1).
Recent results from other teams have shown that mice lacking CR3 (CR3 KO) do not display changes in synapse number in PCs from cortical layers L4 and L5, which called into question the role of the complement in synapse refinement in neocortex. We confirmed those results in L5 PCs of the frontal cortex. However, in striking contrast, our preliminary results show that CR3 ablation results in a strong increase in synapse density in L3 PCs. Together, these results suggest that CR3-dependent synapse pruning specifically affects L3 PCs.
In the brain, CR3 is exclusively expressed by microglial cells. Importantly, recent results show that microglial cells are diverse. Single-cell transcriptomics consistently identified several distinct populations of microglial cells in neocortex, but their functions remain largely unknown(2). We hypothesize that specific microglia subtypes target L3 PCs for CR3-dependent pruning during postnatal development.
Thus, the PhD project seeks to answer two important questions:
1) Does neuron identity determine its targeting by complement-dependent synapse refinement?
2) Do specific subtypes of microglial cells mediate neuron type-specific synaptic refinement?

Part 1. Role of neuronal identity in synapse refinement by the complement.

1.1. Further characterization of CR3-dependent synapse pruning.
To study the effect of synaptic refinement at different dendritic locations in the same neuronal type, dendritic spine density will be quantified in the apical tuft and in basal dendrites of L3 and L5 PCs in WT and CR3 KO mice. The PhD candidate will also assess whether layer-specific synaptic alterations extend to inhibitory synapses (miniature inhibitory PSCs, density of synaptic GABAergic markers) as a consequence of CR3 ablation in L3 and L5 PCs.

1.2. Testing the hypothesis that neuronal identity dictates synapse/neurite refinement using Fezf2-induced respecification of layer 3 neurons
To test the hypothesis that neuronal identity determines complement-mediated synaptic pruning, we will use in vivo lineage reprogramming and generate “L5-like” PCs in L3 (i.e., change the molecular and morpho-functional identity of L3 PCs). Large L5 PCs express the transcription factor Fezf2. We and others have shown that Fezf2, a terminal selector gene, is capable of instructing L5 PC identity(3). Here, in utero electroporation of Fezf2 will be used to convert L3 PCs into “L5-like” PCs. Control animals will receive an EGFP-encoding plasmid only. The PhD candidate will verify that the molecular and functional properties of Fezf2-expressing L3 PCs resemble that of L5 PCs in both WT and CR3 KO mice. She/He will then compare synaptic activity (mEPSCs and mIPSCs) and spine density in induced CFPNs, in WT and CR3 KO juvenile mice. If neuronal identity determines pruning, the effect of CR3 ablation on synapses should be abolished in respecified neurons (Fezf2+ L3 PCs), but not in L3 neurons expressing the control EGFP plasmid.

Part 2. Probing the interactions between specific pairs of neuron-microglia subtypes

Single-cell RNA-seq identified different microglia subpopulations(2), as well as ligand/receptor pairs expressed by different neuron (including L3 PCs and L5 PCs) and microglia subtypes(2). This study provides candidate ligand/receptor pairs (in particular Sema3a/Plxna4 in Layer 2-3) that are differentially expressed in distinct microglia-neuron subtypes.
We hypothesize that identify specific microglia cluster(s) that mediate CR3-dependent pruning in L3 PCs but not in L5 PCs. To achieve this aim, distinct ligand/receptor pairs will be selectively altered in neurons and microglial cells to demonstrate that specific microglia subtypes mediate complement-dependent refinement in L3 PCs.

2.1. Investigating the ligand/receptor candidates in neurons
Previous work has established that Sema3a is highly expressed in L3 PCs but almost absent in large L5 PCs(2). To knock down the expression of Sema3a in L3 neurons, we will use in utero electroporation of shRNA-encoding plasmids at embryonic stage E14.5 in WT and CR3 KO mice. Since Sema3A guides axon/dendrite growth, neuronal migration and neuronal polarization in the embryo, in utero knock-down of Sema3A would likely result in migration and polarization deficits. Therefore, GFP-expressing, Cre-dependent shRNA constructs against Sema3A will be co-electroporated with a plasmid expressing Cre under the control of the human Synapsin promoter (hSyn). hSyn guarantees the expression of Cre once synapses have begun to form in the neonatal cortex. Thus, shRNA-mediated knock-down of Sema3A will start only in electroporated neurons at the neonatal stage, before the period of synaptic refinement.
The readout of this experiment will be mEPSC and mIPSC frequency and spine number in electroporated L3 PCs from WT vs CR3 KO mice. If the tested microglia/neuron, ligand/receptor pair is responsible for synapse/neurite pruning in L3 PCs, knock-down of Sema3a in L3 PCs from CR3 KO mice should not lead to further pruning deficits (compared with the knock-down in WT mice). As a control, we will use a similar shRNA knock-down strategy in L3 PCs to interfere with a ligand/receptor pair (Vegfb/Nrp1) that is not specific of L3 PCs. This should not normalize pruning in L3 PNs from WT vs CR3 KO mice.

2.2. Investigating ligand/receptor candidates in microglial cells
Similarly, we will verify that the inhibition of the partner ligand or receptor in microglia alters CR3-dependent pruning in L3 PNs. We will take advantage of novel Lipid Nano-particles tethered with antibodies developed in collaboration with the team of Anders Etzerodt at Aarhus University (Denmark), to deliver siRNA specifically to microglial cells. Lipid Nano-particles against the gene of interest (Plxna4) together with a GFP reporter protein will be injected in the frontal cortex of neonates (P0-P1), prior to synaptic pruning.
The readout of CR3-mediated pruning will be as part 2.1 (i.e., does pruning still differ in L3 PCs from WT and CR3 KO mice after Plxna4 knock-down in microglia?).

References
1. Schafer DP, et al. (2012) Neuron 74(4):691-705.
2. Stogsdill JA, et al. (2022) Nature 608(7924):750-756.
3. Rouaux C & Arlotta P (2013) Nature cell biology 15(2):214-221.

Developmental defects of the cerebellum are associated with motor but also non-motor disorders such as intellectual disability and autism. About half of patients with these defects don’t receive an etiological diagnosis. Additionally, the disease and developmental mechanisms involved are rarely understood when a causative mutation is identified. In our laboratory, we combine human genetics approaches with disease modeling to better understand the genetic causes and underlying pathological mechanisms. We recently identified several patients with a peculiar hindbrain malformation and/or cerebellar symptoms and de novo loss-of-function mutations in new disease gene. The goal of this project is to use human induced pluripotent stem cells-derived models and zebrafish to characterize a new genetic disorder and uncover new cellular and molecular mechanisms involved in human hindbrain development.

UNDERSTANDING THE PHYSIOPATHOLOGY OF CEREBELLAR DEVELOPMENTAL DEFECTS USING HUMAN CELLULAR MODELS AND ZEBRAFISH

PROJECT DESCRIPTION
Developmental brain disorders (DBD) encompass a highly heterogeneous group of diseases characterized by impairments in cognition, communication, behavior or motor functioning as a result of atypical brain development. The clinical assessment, including brain MRI of numerous DBD cases suggests that the cerebellar development can be specially affected. The human cerebellum is a brain region known to play a key role in motor development but also, a less understood one, in cognition (1). Currently, for almost half of the patients with DBD of suspected genetic origin, there is no genetic diagnosis. Additionally, the disease and developmental mechanisms involved are rarely understood when a causative mutation is identified. In our laboratory, we combine human genetics approaches with disease modelling using human induced Pluripotent Stem Cells (hiPSC)-derived models and zebrafish to better understand the genetic causes and underlying pathological mechanisms of DBD (2, 3). Recently identified gene defects suggest that haploinsufficiency of key genes involved in cerebellar development is a cause of DBD with broad clinical consequences including intellectual disability and autism. In our cohort of DBD patients, we have identified multiple cases with heterozygous loss-of-function mutations in genes known or suspected to play a key role in human neurogenesis and latter cerebellar developmental steps.
One of these patients is affected with psychomotor developmental defects and a rare and peculiar malformation of the pons and cerebellum called pontine tegmental cap dysplasia (PTCD). It is characterized by an abnormal shape of the brainstem and cerebellum on MRI. However, so far, no genetic or environmental causes have been identified for this defect. Diffusion tensor imaging in this patient detected white matter fiber tracts disorganization in the hindbrain, as previously observed in other cases with PTCD (4). Genetic investigation identified a de novo mutation in a gene not previously associated with a Mendelian disorder, highly expressed during cerebellar development and playing a role in axon guidance and transcription regulation. Investigation of other cohorts with DBD identified at least one additional case with de novo loss-of-function mutation and similar symptoms associating motor-coordination defects with cognitive impairments. Among others, these observations support the identification of a new DBD syndrome with disrupted cerebellar development and caused by a new genetic defect.

RESEARCH PROGRAM
1- Development of cellular models using patient-derived induced pluripotent stem cells (iPSCs) and genome base editing
The cellular reprogramming of cell line from the index patient with PTCD is ongoing at the stem cell core facility of the Imagine Institute. We have established adenosine and cytosine base editing technics in the lab as well as other genome editing methods. These technics will be used to correct patient’s mutation and also to create cell lines with the other loss-of-function mutation. The goal is to obtain mutant cell lines with their respective isogenic controls.
2- Differentiation of iPSCs into cellular models for cerebellar development to study the impact of the mutations at the protein, transcriptomic and neuronal levels
The PhD student will use 2D and 3D cellular models for cerebellar development. This work is based on recently published protocols for the study of cerebellar granule cells (5) and Purkinje cells (6). In parallel, 3D cerebellar organoids will be cultured based on a protocol that we have fully established and optimized in the lab. First, the consequence of the mutations will be studied at the transcript and protein levels to detect an alternative transcript-specific impact and a decrease of protein expression levels. Then, cellular differentiation will be studied with emphasis on neurites and axon developments. Finally, the candidate will study the consequences of the mutation at the transcriptomic level during differentiation of 2D and 3D cell models.
3- In vivo consequences on cerebellar development using Zebrafish
Preliminary work identified the fish orthologous gene and characterized its expression. Genome editing will be used to create loss-of-function mutant/crispants at the heterozygous and homozygous states. Hindbrain structural defects will be investigated (3) as well as transcriptomic deregulation during development.
This project will be dedicated to the investigation of models for a new DBD syndrome. This study can identify the first genetic factor causing PTCD and it can uncover new molecular mechanisms involved in human hindbrain development.

REFERENCES
1- Haldipur P, Millen KJ. What cerebellar malformations tell us about cerebellar development. Neurosci Lett. 2019 Jan 1;688:14-25. doi: 10.1016/j.neulet.2018.05.032. PMID: 29802918.
2- Ucuncu E, Rajamani K, Wilson MSC, Medina-Cano D, Altin N, David P, Barcia G, Lefort N, Banal C, Vasilache-Dangles MT, Pitelet G, Lorino E, Rabasse N, Bieth E, Zaki MS, Topcu M, Sonmez FM, Musaev D, Stanley V, Bole-Feysot C, Nitschke P, Munnich A, Bahi-Buisson N, Fossoud C, Giuliano F, Colleaux L, Burglen L, Gleeson JG, Boddaert N, Saiardi A, Cantagrel V. MINPP1 prevents intracellular accumulation of the chelator inositol hexakisphosphate and is mutated in Pontocerebellar Hypoplasia. Nat Commun. 2020 Nov 30;11(1):6087. doi: 10.1038/s41467-020-19919-y. PMID: 33257696
3- Coolen M, Altin N, Rajamani K, Pereira E, Siquier-Pernet K, Puig Lombardi E, Moreno N, Barcia G, Yvert M, Laquerrière A, Pouliet A, Nitschké P, Boddaert N, Rausell A, Razavi F, Afenjar A, Billette de Villemeur T, Al-Maawali A, Al-Thihli K, Baptista J, Beleza-Meireles A, Garel C, Legendre M, Gelot A, Burglen L, Moutton S, Cantagrel V. Recessive PRDM13 mutations cause fatal perinatal brainstem dysfunction with cerebellar hypoplasia and disrupt Purkinje cell differentiation. Am J Hum Genet. 2022 May 5;109(5):909-927. doi: 10.1016/j.ajhg.2022.03.010. PMID: 35390279
4- Engle EC. Human genetic disorders of axon guidance. Cold Spring Harb Perspect Biol. 2010 Mar;2(3):a001784. doi: 10.1101/cshperspect.a001784. PMID: 20300212
5- Behesti H, Kocabas A, Buchholz DE, Carroll TS, Hatten ME. Altered temporal sequence of transcriptional regulators in the generation of human cerebellar granule cells. Elife. 2021 Nov 29;10:e67074. doi: 10.7554/eLife.67074.PMID: 34842137
6- Buchholz DE, Carroll TS, Kocabas A, Zhu X, Behesti H, Faust PL, Stalbow L, Fang Y, Hatten ME. Novel genetic features of human and mouse Purkinje cell differentiation defined by comparative transcriptomics. Proc Natl Acad Sci U S A. 2020 Jun 30;117(26):15085-15095. doi: 10.1073/pnas.2000102117. PMID: 32546527

Chronic ocular pain significantly decreases the quality of life of patients who develop associated anxiety and depression. To develop tailored treatments, it is essential to explore the underlying pathophysiology of ocular pain.
Stress and anxiety can both contribute to the development of chronic/neuropathic ocular pain and exacerbate the symptoms. The primary hypothesis suggests that central sensitization initiates maladaptive mechanisms, resulting in the development of chronic pain. Both neuroinflammation and the descending pain modulatory systems play a critical but yet not defined role in central sensitization.
The objective of this project is to decode the influence of stress and anxiety specifically on ocular pain and its chronicization processes. It will be conducted through behavioral, anatomical, biochemical, and innovative imaging approaches. This innovative PhD project will provide valuable knowledge for future research and potential therapeutic strategies.

Chronic ocular pain, characterized by symptoms such as irritation, dryness, burning, and aching, significantly decrease the quality of life of patients who develop associated anxiety and depression. To develop tailored treatments, it is essential to explore the underlying pathophysiology of ocular pain.
Corneal nociceptive pain arises from the activation of primary sensory neurons located in the trigeminal ganglion, which transmit signals to the trigeminal brainstem sensory complex, a pivotal hub in the processing of trigeminal pain.

Stress and anxiety can both contribute to the development of chronic ocular pain and exacerbate its symptoms by triggering central sensitization, a fundamental process that underlies the chronicization of pain and requires thorough understanding.
In this context, the descending pain modulatory systems, whether inhibitory or facilitatory, play a critical role in both acute and chronic pain. However, if they have been under specific attention the characterization and action of top-down processes on central sensitization and inflammation in the trigeminal nucleus related with ocular pain is less well even not defined.

The objective of the proposed project is to decode the influence of stress and anxiety on ocular pain and its chronicization processes. Thus, the candidate will decipher the maladaptive and inflammatory mechanisms induced by stress. This will involve conducting behavioral (stress-induced paradigm and place preference), advanced imaging technique (functional ultrasound imaging) to assess plastic brain modulation via correlation matrices and functional neuroimaging, as well as performing anatomical approaches (tissue analysis, sectioning, immunohistochemistry, confocal analysis) along with biochemical (Eliza) studies in rodent pre-clinical models of chronic or neuropathic ocular pain.
This innovative PhD project will provide pioneering insights into the impact of stress and anxiety on the initiation and maintenance of chronic ocular pain, offering valuable knowledge for future research and potential therapeutic strategies.

The synapse is a nanoscale machine, which transfers, integrates and stores information in brain circuits. Its proper function relies on multimolecular networks of interactions whose composition and dynamics shape synaptic transmission. A large body of evidence indicates that synapses specialized in humans. Human synapses are more densely distributed along dendrites and they mature over longer time scales than in rodents or non-human primates. The rules governing their plasticity also differ from the other mammalian species studied so far. These traits reflect new regulations of neural cell biology and contribute to the formation and function of complex circuits supporting human cognitive abilities. Yet, the molecular underpinning remains largely unknown. This project will tackle the role and mechanisms of atypical neurotransmitter receptors at the intersection between human brain evolution and neurodevelopment disorders in synaptic development and plasticity.

The most striking feature of the human brain is its size, which has expanded since the separation with non-human primates. However, the human brain is not just a large brain and it is now clear that unique properties allow human neurons to integrate much more information than in other species. The synapse, where two neurons get in contact and communicate, is critical in this context. Synapses are molecular nanomachines whose composition and organization determines which neurons are connected and how signals are transmitted in the brain. As a consequence, subtle changes induced by mutations in synaptic genes are a major cause of neurodevelopmental and psychiatric disorders such as intellectual disability, epilepsy and autisms. Subtle changes also appeared during human evolution, leading synapses to be much more numerous than in other primates and to mature over extended periods of time, which enhances the influence of the environment of brain development. These features contribute to the uniqueness of the human brain, its distinctive cognitive abilities and its susceptibility to diseases. It is therefore critical to understand their molecular basis if we want to treat brain diseases.

The lab has previously shown that a human-specific gene, SRGAP2C, underlies to the emergence of features that are characteristics of human neurons, such as delayed synaptic maturation and increased synapse density (Charrier et al. Cell 2012, Fossati et al., Neuron 2016; Assendorp et al., BioRxiv 2022). Using proteomic and transcriptomic approaches, the lab has identified a convergence of human-specific modifiers on atypical neurotransmitter receptors implicated in neurodevelopmental disorders whose expression peaks during synaptogenesis. This project will tackle the role of one of these genes in the formation of cortical circuits, focusing on the regulation of trans-synaptic interactions and synaptic transmission.

To achieve these goals, the project will employ a multi-scale approach that goes from the molecule to the circuit. The methodology is based on in vivo manipulations in intact mouse cortical circuits, high-resolution microscopy, electrophysiology, quantitative proteomics and engineering of cortical neurons derived from human pluripotent stem cells. The combination of mouse and human models will allow us to establish a robust framework to decipher fundamental mechanisms of neuronal and synaptic development, and better understand what makes human neurons special.


Publications:
Assendorp N, Depp M, Fossati M, Dingli F, Loew D, Charrier Cc, CTNND2 moderates neuronal excitation and links human evolution to prolonged synaptic development in the neocortex, bioRxiv 2022.09.13.507776; doi: https://doi.org/10.1101/2022.09.13.507776

Gemin O, Serna P, Zamith J, Assendorp N, Fossati M, Rostaing P, Triller A and Charrier C, Morphologically constrained modeling of spinous inhibition in the somato-sensory cortex (2021) PLoS Biol, 19(8):e3001375.

Fossati M and Charrier C, Trans-synaptic interactions of ionotropic glutamate receptors (2020) Curr Opin Neurobiol, 66:85-92.

Fossati M, Assendorp N, Gemin O, Colasse S, Dingli F, Arras G, Loew D and Charrier C, Tans-synaptic signaling via the glutamate receptor delta-1 mediates inhibitory synapse formation in cortical pyramidal neurons (2019) Neuron, 104(6):1081-1094.e7.

Fossati M, Pizzarelli R, Schmidt ER, Kupferman JV, Stroebel D, Polleux F, Charrier C, SRGAP2 and its human-specific paralog co-regulate the development of excitatory and inhibitory synapses (2016) Neuron, 91(2):356-69.

Charrier C, Joshi K, Coutinho-Budd J, Kim JE, Lambert N, de Marchena J, Jin WL, Vanderhaeghen P, Ghosh A, Sassa T and Polleux F, Inhibition of SRGAP2 by its human-specific paralogs induces neoteny during spine maturation (2012) Cell, 149(4):923-35.


Expected profile of the candidate: We are seeking a highly motivated and creative student with strong team spirit. The candidate should have a solid background in neurobiology and cellular/molecular biology.

Environmental changes drive drastic evolutionary changes in behavior and brain function. While the genetic basis of evolution have been studied, the evolution of brain computations, are unknown. Astyanax is a model for studying genetic mechanisms underlying trait evolution. Astyanax consists of a river and cave populations that independently evolved in largely isolated caves. Cavefish exhibit prominent changes in sensory systems including loss of vision and expansion of smell, taste and lateral line. Despite the robust changes in behavior and morphology, the shifts in processing sensory information within the brain has been unexplored.
Here, we will use light-sheet Ca2+ imaging and optogenetics to investigate the evolutionary shifts in sensory processing in the largest multisensory processing brain region, the optic tectum. This study will uncover principles on how evolution shaped brain computations of sensory systems to cope with the sensory constraints of caves.

INTRODUCTION
The Mexican tetra Astyanax mexicanus represents a unique biological system to study the genetic and neural basis of sensory evolution. A. mexicanus consists of eyed surface population that inhabit rivers in Northeast Mexico and at least 30 isolated populations of cavefish that have independently evolved in markedly different habitats. Cavefish have evolved a series of morphological, physiological, and behavioral changes including eye loss, albinism, reduced sleep and social behaviors, and expansion of non-visual sensory systems (e.g. olfaction, lateral line, etc.). Many of these phenotypes and behaviors have evolved repeatedly in geographically isolated cavefish. Thus, A. mexicanus is powerful model to investigate the genetic and ecological factors that shape the evolution of traits.
Despite over 80 years of research investigating evolved differences in behaviour and morphology of sensory systems between the surface and the cavefish populations, the effects of evolution on neural circuit dynamics and brain computation, have never been studied.
Here, we will address these open questions by studying the differences in functional connectivity and sensory processing between surface and cavefish populations.
This proposal focuses on sensory processing in the optic tectum, a brain region that is homologous to the mammalian superior colliculus. The tectum receives sensory inputs from the retina, as well as other sensory modalities. Its main functional role is to detect sensory stimuli, process them, and generate appropriate motor behaviors such as prey capture. Previously, we have used functional imaging in zebrafish to assess the functional role of the tectum in vision processing. In cavefish the optic tectum is still present (just a ~20% reduction in size), raising the possibility that this brain region has been repurposed for functions other than vision. Therefore, the optic tectum is an ideal system to study evolutionary changes in sensory circuits. A. mexicanus is closely related to zebrafish, they are easy to raise in the lab, suitable for embryo injection, and are transparent. These traits, in combination with a recently sequenced genome, permit implementation of all zebrafish genetic and imaging advantages in the A. mexicanus model. In addition, surface and cavefish populations are interfertile allowing for the generation of hybrid fish for genetic mapping and trait segregation studies. This proposal compares ancestral surface fish to the Pachòn and Molino populations of cavefish that have independetly evolved many shared traits including eye loss. The proposed experiments will address how visual systems evolve, and whether evolution repeatedly occurs through shared genetic and neural mechanisms.

Aim. Defining the evolutionary rewiring and repurposing of visual circuits – a calcium imaging approach
Here, we will study the evolved changes in sensory processing at the level of neuronal circuits. In zebrafish, neurons in the optic tectum, the main multimodal sensory processing area in teleost fish, are selective for different stimulus features in the visual field such as size , position, or direction. Some neurons also respond to auditory and water-flow stimuli. However, these two modalities act primarily to modulate visual processing, rather than representing stimulus-specific information on overlapping sensory maps. In surface fish, the retina represents 90% of the tectal inputs, but it is unclear how sensory representation and processing in the optic tectum has changed in cavefish that lack eyes. Using functional imaging, we will study how this classically visual center has repurposed and rewired for processing non-visual sensory modalities.
Aim 1. Sensory processing repurposing in the optic tectum
Here, we will test whether the optic tectum has been repurposed to process non-visual sensory modalities surface fish with Pachón and Molino cavefish larvae. Our collaborator, Alex Keene (Texax A&M, USA) has already generated transgenic cavefish with pan-neuronal expression of GcaMP6s in surface fish, Pachón and Molino cave populations. In combination with light-sheet microscopy we will monitor the entire tectal circuit, with single-neuron resolution, in an intact behaving surface and cavefish larvae (8 days post fertilization), while presenting different types of sensory stimulation. We will compare the differences in responsiveness to sensory stimuli across the tectum to address the following questions:
i) Has the tectum of the cavefish shifted to principally process to non-visual modalities (e.g. water-flow, auditory, olfaction, taste) ? Specifically, we will study the differences in neuronal sensory tuning curves (e.g. sound frequency, amplitude modulation, taste or gustatory) between surface fish and cavefish. And whether they exhibit sensory maps at the circuit level (e.g. tonotopy, auditory spatial representation.

ii) We will define whether changes in sensory processing in the tectum are accompanied by differences in the ratio between excitatory, inhibitory and modulatory neurons ? For this purpose, the Keene lab will perform immunostaining for Vglut2a, Gad1b and ChAT to quantify at the whole-tectum level, the ratio between the different types of neurons, in surface and cavefish morphs.

iii) Alternatively, it is possible that the optic tectum evolved to become a modulatory or motor brain center rather than a sensory processing area. If this is the case, we will record the tectal responses to arousal stimuli (e.g. a mild electric shock), and cross correlations between tectal spontaneous activity and behaviour to test for motor repurposing.

iv) Comparing behavioural and tectal sensory responses results between Pachón and Molino cavefish will shed light on whether their evolution converged on shared changes in tectal repurposing. In case of convergence, is this reached through similar or different underlying mechanisms ?

Finally, will test causality of the potential repurposing of the optic tectum by assessing cavefish sensory induced behaviours in intact or tectum-ablated larvae. Ablations will be performed using a two-photon microscope.


Aim 2. Evolutionary shifts in the tectum’s functional connectivity
Spontaneous activity can be used to infer the circuit’s functional connectivity, for example calculating the correlations between the spontaneous activity of different neurons. We have developed approaches for imaging and data analysis of the dynamics of large neuronal circuits in zebrafish that will be applied to A. mexicanus. The lab has shown that the spontaneous tectal activity in zebrafish displays a spatiotemporal structure with distinct functional neuronal assemblies (highly functionally interconnected groups of neurons), coding the spatial position of potential prey, and showing attractor-like dynamics that could improve the signal-to-noise ratio for visual detection. Here, we will further compare the spontaneous tectal activity of the surface and the two cavefish populations to learn about the evolutionary changes in the tectum’s functional connectivity associated with the sensory constraints of caves.
We will specifically ask:
i) Do cavefish display neuronal and populational properties similar to those of surface fish (e.g. calcium events duration and rate) ?
ii) Does the spatiotemporal structure represent functional neuronal assemblies? If cavefish tectum processes a non-visual modality or will be repurposed for a non-sensory computation (e.g. modulation).
iii) Preliminary data shows that assemblies exits in the cavefish optic tectum. If this is indeed the case, we will determine whether assemblies in cavefish have adapted for processing auditory, lateral line or olfactory modalities ? Which properties of these modalities will they represent (e.g. frequency, spatial localization, intensity). Do they still show attractor-like dynamics, a mechanism present in zebrafish for improving visual detection ?

We recently reported that astrocyte calcium (Ca2+) signals in the nucleus accumbens (NAc) are necessary for the acquisition of cocaine seeking behavior and require the matricellular protein hevin, which is a secreted components of the extracellular matrix implicated in excitatory synaptogenesis. Moreover, hevin knockdown in astrocytes blocks cocaine-induced structural and synaptic plasticity in NAc medium spiny neurons (MSN). Mechanistically, our data suggest that cocaine-induced Ca2+ rise in astrocytes may regulate the secretion of hevin, which in turn modifies the glutamatergic synapses, hence regulating reward.
The thesis objectives are:
1 determine the dynamics and the extracellular cues mediating hevin secretion
2 characterize the role of hevin in shaping neuronal activity of MSN
3 identify the binding partners of hevin that mediate the cellular and rewarding properties of cocaine

The findings of the lab, together with the existing literature, led to the working hypothesis that emotionally relevant stimuli such as drugs of abuse recruit brain astrocytes and hevin to induce neuroplasticity. Mechanistically, our data suggest that cocaine-induced calcium rise in astrocytes may regulate the secretion of hevin, which in turn modifies the glutamatergic cortico-striatal synapses on medium spiny projection neurons (MSN), hence regulating reward processes and addiction.
The objective of this PhD project is now to decipher the mechanism of action of hevin in the neuroplasticity of the rewarding properties of cocaine. The selected PhD fellow will pursue the following specific aims:

Aim 1. Determine the dynamics and the extracellular cues mediating hevin secretion
He will quantify the effects of astrocyte calcium activation on the secretion of hevin. Astrocyte calcium signals play a role in regulating secretion of various molecules by these cells, which may contribute to the synapse remodeling associated with cocaine rewarding effects (1,2). He/She will dissect the effects of increased intracellular calcium signals on hevin secretion dynamics using TIRF live-imaging (imaging facility, IBPS, Sorbonne Université), in primary astrocytic cultures of mouse striatum (3-5). The advantage of TIRF imaging compared to confocal imaging is to allow the visualization of individual secretory vesicles close to the plasma membrane6. We use a novel protocol containing heparin-binding EGF-like growth factor that generates more mature astrocytes than classical cultures methods (5,7). This medium promotes the development of astrocytes with a stellate morphology, fine processes, and spontaneous calcium transients. Astrocyte cultures will be co-transfected on day 14 in vitro (DIV 14) with the AAV construct expressing hevin-SEpHluorin. To mimic the calcium rise induced by cocaine in astrocytes, he/she will use a DREADD chemogenetic approach. Transfected cells will be identified with DREADD-mCherry staining. TIRF confocal live imaging will allow us to visualize the action of astrocyte activation on the dynamics of individual vesicles upon addition of CNO into the imaging solution.
Astrocytes express a multitude of receptors and can be activated by neurotransmitters and cytokines (8), known to be increased by cocaine. These factors increase astrocyte calcium signals in slices and astrocyte cultures (9-12), which indicate that they could mediate hevin secretion. We will test the effect of the class I mGluR agonist (RS)-3,5-dihydroxyphenylglycine (DHPG), dopamine, TNFa, and IL-1b on hevin-SEpHluorin secretion in astrocytes by TIRF imaging.
In parallel, identification of the vesicles expressing hevin in astrocytes will be determined by co-expression of hevin with markers of different organelles and vesicles tagged with fluorescent proteins using confocal imaging, such as Syt7-mCherry, DsRed-Rab7 and VAMP2-mCherry.

Aim 2. Characterize the role of hevin in shaping neuronal network activity of MSN
Behavioral adaptations to drugs of abuse are associated with changes in NAc MSN activity (13). Optogenetic manipulation of the two subtypes of MSN (D1 and D2) has confirmed their opposite effects on reward-related behavior (14). Importantly, recent study shows that modulation of astrocytic activity in striatum differentially affects D1- and D2-MSNs (1) suggests that pre-existing differences in astrocytic activity in NAc could influence sensitivity to cocaine by modulating D1- and D2-MSNs. We postulate that D1 and D2 MSN activity may be differentially altered upon astrocytic hevin knockdown. The PhD student will test the consequences of hevin down-regulation in astrocytes on MSN activity in vivo using fiber photometry. He/She will inject AAV-GFAP-EmGFP-miR-hevin and AAV2.2-EF1a-DIO-GCaMP6f in the NAc of D1-Cre or A2A-Cre transgenic mice, two mouse lines expressing Cre-recombinase under the control of D1 MSN and D2 MSN neuron-specific promoters, respectively. Mice will be habituated to handling before in vivo calcium signal recording by fiber photometry. On the test day, recording will be done at basal state and after acute cocaine injection. Since cocaine was shown to activate D1 MSN and to inhibit D2 MSN, we postulate that hevin knockdown reduces D1 MSN calcium activity after cocaine injection, and that D2 MSN calcium activity is not altered following hevin knockdown. This would demonstrate the specific role of hevin on the cocaine-induced reinforcement of glutamate synapses on D1 MSN. Importantly, advanced imaging in the laboratory allowing simultaneous recording of two cell-type (astrocytes: GFAP-GCaMP and neurons: hSyn-RGECO1a) in in two distincts brain regions (NAc and prefrontal cortex), will allow us to analyze the key astro-neuronal network (PFC to NAc MSN) involved in drug addiction.

Aim 3. Identify the binding partners of hevin that mediate the cellular and rewarding properties of cocaine
Certain components of the synapse or extracellular matrix were shown to interact with hevin to promote synaptogenesis. Hevin has been shown to mediate synaptogenesis via interaction with the synaptic adhesion molecule nlgn1B and with the vesicular neuronal protein calcyon (15,16). Nlgn1B is a synaptic organizer molecule involved in the formation of excitatory synapses. Calcyon is a neuronal vesicular transmembrane protein involved in vesicle trafficking and endocytosis of D1 receptor. We selected a third potential target, the adhesion molecule b3-integrin, which can bind to components of the extracellular matrix and other ligands. This receptor was shown to be required for cue-induced reinstatement to cocaine and its level increased in the NAc after cocaine reinstatement (17). The PhD student will address whether these interactions could mediate the effects of hevin on cocaine-induced plasticity, using co-immunoprecipitation (IP) of hevin on NAc extracts of mice treated with vehicle or cocaine, and compare the levels of nlgn1B, calcyon and b3-integrin in these two conditions. Samples will be analyzed by SDS-PAGE and the binding of hevin with each potential partner quantified by Western blotting.
To further verify the relevance of the interactions between hevin and potential partners in vivo, he/she will perform in situ proximity ligation assay (PLA), which detects the physical proximity between two proteins in situ. He/She will quantify this interaction by counting the number of puncta revealing the proximity of the proteins before and after injection of cocaine. The time course used after cocaine injection to evaluate the interaction of hevin with each potential candidate will be based on the dynamics of hevin secretion observed in Aim 1.

Conclusions
This PhD project will provide a thorough understanding of the role of hevin in activity-dependent plasticity in the context of exposure to cocaine, from its regulated secretion by astrocytes to its synaptic effects on the neuronal and astrocytes network activity of basal ganglia. This project relies on novel techniques and tools that have been developed through the years and well validated in the laboratory (18), making this ambitious project feasible.

1 Yu X et al. Neuron 2018
2 Nagai J et al. Cell 2019
3 Li D et al. The Journal of Physiology 2015
4 Li D et al. The Journal of Neuroscience 2008
5 Pham C et al. Acta Neuropathologica Communications 2021
6 Becherer U et al. PLoS One 2007
7 Dean C et al. Journal of Visualized Experiments 2018
8 Kofuji P et al. 2021
9 Beskina O et al. American Journal of Physiology-Cell Physiology 2007
10 Bezzi P et al. Nature Neuroscience 2001
11 Corkrum M et al. Neuron 2020
12 D'Ascenzo M et al. Proc Natl Acad Sci U S A 2007
13 Wheeler RA et al. Neuropharmacology 2009
14 Lobo MK et al. Science 2010
15 Kim JH et al. Cell Death Differ 2021
16 Singh Sandeep K et al. Cell 2016
17 Wiggins A et al. The Journal of Neuroscience 2011
18 Mongrédien R et al. bioRxiv 2023

Autism spectrum disorders (ASDs) include a heterogeneous group of pervasive developmental disorders characterised by alterations in communication, socialisation, and by the presence of repetitive behaviours. ASD has a strong genetic component and hundreds of risk genes have been associated with the disease. Multiple genes encoding synaptic proteins have been found mutated in ASD and synaptopathy has emerged as one of the major pathophysiological mechanisms involved in the disease. Mutations or deletions in Reelin gene (Reln) have been associated not only with cortical malformations (MCD) but also with ASD, schizophrenia and epilepsy. In mouse, reduction of Reln expression in the heterozygous Reeler mutant (HRM) lead to decreased synaptic plasticity together with altered behavior. Our project relies on structure/function analyses of human Reln variants in mouse and also on the studies of its signaling in neurons and glia in order to better understand the pathophysiology of the ASDs.

Reelin (Reln) is a secreted protein well known for its function in neuronal migration during cerebral cortex development. Its loss of function in mice is responsible for the Reeler phenotype characterized by cerebellar atrophy and laminar disorganization in the cerebral cortex. In human, Reln mutations or gene deletion have been reported in patients with cortical malformations (MCDs: lissencephaly, pachygyria…), Autism Spectrum Disorder (ASD), temporal lobe epilepsy and also schizophrenia (1). Beside its function in the prenatal brain, Reln plays important roles in the postnatal and adult cortex, promoting the maturation of dendrites, synaptogenesis, synaptic transmission and plasticity, thus modulating the formation and function of synaptic circuits (2). These postnatal roles are also associated to a switch in gene expression from Cajal-Retzius cells (CRs) to inhibitory neurons. The CRs population is transitory and a subtype is known to completely disappear in the first two postnatal weeks in the mouse cortex by apoptosis in an activity-dependent manner (3,4). The kinetic of CRs elimination is delayed in the hippocampus until adulthood 5. By genetically maintaining them alive beyond these temporal windows, we found that CRs survival is associated with an increased sensitivity to kainate-induced seizures (5,6) together with enhanced dendritic spines of cortical and hippocampal pyramidal neurons. Whether these non-cell autonomous modifications are due to neurotransmission or paracrine effects remains unknown.
Heterozygous Reeler mutation (HRM) does not alter cortical lamination nor the cerebellum but leads to alterations in behavior (Paired Pulse Inhibition and Cued Fear Conditioning) (7) and in synaptic plasticity (8), a common phenotype found in ASD patients with various genetic conditions. In addition, rare variants in the Reln gene have been described in patients with autism (1). The most damaging variants according to bioinformatic predictions are all concentrated in the central domain important for binding to its main receptors, VLDLR and ApoER2. Previous work showed that Reln enhances LTP in hippocampal slices, an effect dependent on both VLDLR and ApoER2 (9). Therefore, we expect that Reln variants in ASD alter binding to its receptors and lead to enhancement or suppression of LTP. We have recently worked with functionally modeling Reln variants in patients with MCDs using in vitro and in vivo models (10).
Aim1: Consequences of Autism associated Reelin mutations on synaptic plasticity
In this context we will first study whether some of these mutations in the central domain also have effects on receptors bindings, on synaptic plasticity and model the ones found in ASD 1 (and unpublished data from Thomas Bourgeron (Coll.), Institut Pasteur). To do so, we will concentrate supernatants from cell culture expressing Reln variants and incubate hippocampal slices with those to monitor the LTP enhancement. In addition to these in vitro experiments, we will implement canulations in hippocampus and perfuse HRM with these supernatants to rescue or not the synaptic plasticity and learning deficits observed in those animals. The HRM mouse line is already available in our animal facility and the biochemistry facility at the IPNP Institute will produce the supernatants from transfected cell lines with Reln mutant plasmids. Canular implantations in these animals will be done under stereotaxic injection to perfuse the supernatant in adults.
Aim2: Contribution of Cajal-Retzius cells and inhibitory neurons to synaptic plasticity in hippocampus
In parallel we will implement a new conditional model by introducing a floxed allele of Reln in HRM and induce in postnatal the loss of the second allele. We expect to enhance the HRM phenotypes associated to developmental alterations by removing the second allele in juvenile or adult stages using tamoxifen injection in the Cre-ERT2 recombinase background. Since there is a time dependent switch of Reln expression after birth, this model will allow to study the contribution of CRs and interneurons to the Reln postnatal functions. Alternatively, we will perform stereotaxic injections of AAV-Cre virus in the hippocampus of HRM/lox mice using specific promoters to target one cell type or the other.
Aim3: Reln signaling in astrocytes and contribution to synaptic plasticity in hippocampus
Since the synapse is multi-partite with glia surrounding the pre and post-synaptic compartments (11), we will also investigate the responds of astrocyte to secreted Reln signaling in glia and its role in modulating synaptic transmission and plasticity. Mainly, we will specifically inactivate key components of the Reln pathway in astrocytes using Cre/Lox technologies.
Regardless of the three aims and thanks to the collaboration between our 2 laboratories, we will be able to analyze both the behavior and the synaptic plasticity in treated animals between 3 to 5 months Hippocampal slices will be recovered from controls and treated mice to record TBS (Theta-Burst Stimulation) induced LTP from CA1. Finally, we will inject biocytin in pyramidal neurons to reconstruct their neuronal morphologies.
Altogether, this project will bring new knowledge on postnatal Reln functions in various cell types and shed light on how various Reln mutations lead to distinct neurodevelopmental disorders (NDDs).

References:

1. Di Donato, N. et al. Monoallelic and biallelic mutations in RELN underlie a graded series of neurodevelopmental disorders. Brain 145, 3274–3287 (2022).
2. Jossin, Y. Reelin Functions, Mechanisms of Action and Signaling Pathways During Brain Development and Maturation. Biomolecules 10, 964 (2020).
3. Riva, M. et al. Activity-dependent death of transient Cajal-Retzius neurons is required for functional cortical wiring. eLife 8, e50503 (2019).
4. Ledonne, F. et al. Targeted Inactivation of Bax Reveals a Subtype-Specific Mechanism of Cajal-Retzius Neuron Death in the Postnatal Cerebral Cortex. Cell Reports 17, 3133–3141 (2016).
5. Riva, M. et al. Aberrant survival of hippocampal Cajal-Retzius cells leads to memory deficits, gamma rhythmopathies and susceptibility to seizures in adulthood. Nature Communications (2023).
6. Ramezanidoraki, N. et al. Activation of the PI3K/AKT/mTOR Pathway in Cajal–Retzius Cells Leads to Their Survival and Increases Susceptibility to Kainate-Induced Seizures. IJMS 24, 5376 (2023).
7. Rogers, J. T. et al. Reelin supplementation recovers sensorimotor gating, synaptic plasticity and associative learning deficits in the heterozygous reeler mouse. J Psychopharmacol 27, 386–395 (2013).
8. Qiu, S. et al. Cognitive disruption and altered hippocampus synaptic function in Reelin haploinsufficient mice. Neurobiology of Learning and Memory 85, 228–242 (2006).
9. Weeber, E. J. et al. Reelin and ApoE Receptors Cooperate to Enhance Hippocampal Synaptic Plasticity and Learning. Journal of Biological Chemistry 277, 39944–39952 (2002).
10. Riva, M. et al. Functional characterization of RELN missense mutations involved in recessive and dominant forms of Neuronal Migration Disorders. http://biorxiv.org/lookup/doi/10.1101/2021.05.25.445586 (2021) doi:10.1101/2021.05.25.445586.
11. Cresto, N., Pillet, L.-E., Billuart, P. & Rouach, N. Do Astrocytes Play a Role in Intellectual Disabilities? Trends in Neurosciences 42, 518–527 (2019).

We aim at developing a new type of neuro-inspired artificial neural network (NN), bio-inspired, for continual learning, based on the principle of Information Maximization proposed by Barlow. Information Maximization will serve to replicate the brain’s capabilities for large memory capacity, rapid acquisition, robust memory retrieval.
Recently, we successfully developed a NN that satisfies Information Maximization by exploiting random and unprecise neurons in order to encode information. The idea behind is to exploit randomness and quantization to generate orthogonal representations that do not overlap (sparse coding).
This PhD thesis aims at pursuing the recent works done to design new neural models, to explore new features and capabilities for continual learning. Some parallels and comparison will be made with current Machine Learning techniques and brain-inspired architectures . For instance, how the hippocampus acquire and store robustly information.

We aim at developing a new type of bio/neuro-inspired artificial neural network (ANN) for continual learning, based on the principle of Information Maximization proposed by Barlow (1969) to describe efficient coding in neurons. Information Maximization is a strong principle to understand how information is conveyed in the brain.

The hypothesis of efficient encoding states that neurons should encode information as efficiently as possible in order to maximize neural resource utilization. From an Information Theory viewpoint, compressing information is equivalent to a maximum entropy representation, suppressing redundant parts in signals, which can be reconstructed or approximated from the representation. Thus, efficient neural codes are more compact and can serve to encode more information for the same capacity of neurons, or reduce their number for the same size of data.

Achieving Information Maximization in ANNs will serve to better understand information processing in the brain. It will also serve to replicate the brain’s capabilities for large memory capacity, rapid acquisition, robust memory retrieval, as well as promote code compactness and low energy consumption, as fewer resources are required. These properties are difficult to obtain in current ANNs, as very large models like Transformers become the norm.

Recently, Pitti, Weidmann, and Quoy (PNAS 2022) successfully developed a NN that satisfies Information Maximization by exploiting random neurons in order to encode information. The idea behind is to exploit randomness — which implies high entropy — and quantization — which suppresses redundancy — to generate orthogonal representations that do not overlap (e.g., Olshausen 2004). This new associative memory has allowed the rapid learning of sparse and distributed memories with few neurons only, and has been shown to not forget catastrophically when acquiring new information (French 1999). Furthermore, experiments showed that this type of encoding achieves memory capacities close to Shannon’s information theoretic limit; hence following Barlow hypothesis of efficient encoding by Information Maximization.

This PhD thesis aims at pursuing the above-mentioned recent works to design new neural models based on Information Maximization and to explore new features and capabilities for continual learning (Annabi et al., NN 2022). Parallels and comparisons with current Machine Learning techniques and brain-inspired architectures will be made.

For instance, links can be developed on how the hippocampus robustly acquires and stores information without catastrophic forgetting (McClelland et al 1995, French 1999), and how complementary systems (e.g., the neocortex) can exchange and store information for large-scale memory capacity scaffolding and development (Pitti et al., TCDS 2022).

Other links can be made with recent Machine Learning neural networks like the modern Hopfield Network (mHN) (Krotov and Hopfield 2020) and the Sparse Distributed Memory system (SDM) (Bricken 2021), which are associative memory systems that have attracted a lot of attention recently. One line of research is to compare our NN with current state of the art methods and show how it can potentially learn and reconstruct information faster in the challenging one-shot learning scenario.

Furthermore, we expect to obtain models with higher memory capacity in comparison with mHN and SDM, since we will embed mechanisms that allow to convey maximum information. We believe that our neural architecture will expand its performance on larger databases during continual learning with a minimal computational cost.

Methodology to reach the scientific objectives of the project:

The thesis will be organized around three main research topics, focusing on improving the neural architecture for rapid acquisition and reconstruction using insights from Coding/Information Theory, comparison with other neural architectures in Machine Learning on known databases for continual learning, design of a complementary neural architecture for ‘fast’ encoding and ‘slow’ generalization.

In the first stage, we will develop an energy-efficient, high capacity and long-term memory network unaffected by catastrophic interference from new input. In the second stage, we will focus on coding efficiency and memory information capacity with respect to current Machine Learning algorithms. In the third stage, we will concentrate on continual learning with two complementary systems, modeling hippocampus-cortical interaction.

Location : ETIS laboratory, CNRS, ENSEA, CY Cergy-Paris University
Cergy is a mid-sized town north-west of Paris, at 40 minutes by train.
ETIS is the main computer science CNRS laboratory in the West of Paris.
The Neurocybernetics team (15 researchers and 15 PhD students/PostDoc) is an international team working in the field of Bio-inspiration for intelligent systems design. It is specialized in Cognitive Robotics and Brain-inspired models with many international projects and close to the international communities.

Contacts : alexandre.pitti@ensea.fr claudio.weidmann@cyu.fr
References :
Annabi, L. Pitti, A. Quoy, M. (2022) Continual Sequence Modeling With Predictive Coding Front. Neurorobot. 16:845955
Barlow, H.B. Possible principles underlying the transformation of sensory messages. Rosenblith, W. (Ed.), Sensory Communication. MIT Press, Cambridge, MA., 1961.
Bricken, T., & Pehlevan, C. (2021). Attention approximates sparse distributed memory. arXiv preprint arXiv:2111.05498.
French, Robert M. Catastrophic forgetting in connectionist networks Trends in Cognitive Sciences 3, 4 (1999): 128 35.
Krotov, D., & Hopfield, J. J. (2016). Dense associative memory for pattern recognition. Advances in Neural Information Processing Systems, 29, 1172–1180.
Krotov, D., & Hopfield, J. (2020). Large associative memory problem in neurobiology and machine learning. arXiv preprint arXiv:2008.06996.
McClelland, James & Mcnaughton, Bruce & O’Reilly, Randall. (1995). Why There are Complementary Learning Systems in the Hippocampus and Neocortex: Insights from the Successes and Failures of Connectionist Models of Learning and Memory. Psychological review. 102. 419-57. 10.1037/0033-295X.102.3.419.
McClelland James L., McNaughton Bruce L. and Lampinen Andrew K. (2020) Integration of new information in memory: new insights from a complementary learning systems perspective Phil. Trans. R. Soc. B37520190637. 20190637
Olshausen BA, Field DJ. Sparse coding of sensory inputs. Curr Opin Neurobiol. 2004 Aug;14(4):481-7.
Pitti, A. Weidmann, C. Quoy, M. (2022) Digital computing through randomness and order in neural networks, PNAS, 119 (33) e21153351
Pitti, A. Quoy, M. Lavandier, C. Boucenna, S. and Weidmann, C. (2022) In Search of a Neural Model for Serial Order: A Brain Theory for Memory Development and Higher Level Cognition, IEEE Transactions on Cognitive and Developmental Systems, 14, 2, 279-291.

The developing nervous system is highly vulnerable to environmental factors including chemical substances. Risk assessment of the potential adverse neural effects of real-life chemical mixtures is one of the major challenges faced by scientists and regulators. The team participated to the development of a new methodological approach to perform risk assessment of real-life contaminant mixtures. The theoretical methodology applied to a human breastmilk mixture suggested a high risk for infants for thyroid and related neurodevelopmental and cognitive processes. The project aims to:
i) Analyze the effects of exposure to human breastmilk mixture on thyroid function and related neurodevelopmental and cognitive processes, and underlying cellular and molecular mechanisms using a combination of mouse in vivo and human in vitro models.
ii) Use these data to challenge and improve the new methodological approach, and build adverse outcome pathways for chemical mixture assessment.

The nervous system is extremely sensitive to environmental factors including chemicals used in industrial and agricultural activities and released in the environment. This vulnerability is particularly important during developmental periods. Chemicals can cross the placental and blood brain barriers and act directly on the nervous system. They can also interfere with the endocrine systems, which then impact the neuroendocrine regulation of neural structures. This is the case of endocrine disrupting compounds (EDC) targeting for instance the thyroid system, which plays a key role in the regulation of early and late neural processes.
The majority of studies addressing the effects of EDC exposure focus on a single compound. When mixtures are studied, the substances are often derived from the same family. Humans and wildlife are, however, exposed to heterogenous mixtures of substances from different compound families and uses. Given the multiplicity and complexity of real-life mixtures, the host team contributed to the development of a new methodological approach to assess the chemical risk of heterogeneous contaminant mixtures (Crépet et al., 2022). This methodology includes i) selection of the mixture from combined human exposure by a statistical method, ii) selection of toxicological endpoints, iii) determination of the reference doses from literature, and iv) component-based risk assessment. This methodology was applied to the French ContaLait survey, which measured several contaminants in human breast milk collected in lactariums. The selected mixture was composed of 19 (or families of) substances including insecticides, fungicides, flame retardants and dioxins. The theoretical assessment using the developed method indicated that infants exposed to the 19-substances-mixture present in human breastmilk may have a high risk of thyroid disruption and related neurodevelopmental and cognitive processes. To challenge and validate this new model and characterize the neural effects and mechanisms of human breastmilk mixture, experimental data are required. In this context, the PhD project aims to conduct experimental studies to evaluate for the first time the effects of developmental exposure to the human breastmilk-mixture on the thyroid system and neurodevelopmental/cognitive processes. The project will combine the use of i) a mouse model of exposure for the neurodevelopmental, neuroendocrine and cognitive analyses, and ii) a human in vitro model of hippocampal neurons derived from induced pluripotent cells (iPSC).

i) Neurodevelopmental, neuroendocrine and behavioral analyses
Dams of the C57BL/6J background will be exposed two weeks before mating and throughout the gestational and lactational periods. They will be given once a day at the same hour a store-bought sugar cookie dough ball containing either the vehicle (control group) or the breastmilk mixture (Mix-exposed group) at an external dose recently determined by Ineris colleagues through conversion of the internal dose found in human breastmilk using physiologically based pharmacokinetic models. Dams will be weighed weekly to adjust the mixture dose.
Male and female offspring will be analyzed for developmental landmarks (body weight, age of eye opening and incisor eruption, age of sexual maturation, motor and sensory function, motor activity). After weaning, they will be analyzed for motor activity and anxiety-related behavior. At adulthood, females will be monitored for their estrous cycle two weeks before the beginning of the behavioural test battery and throughout the period of behavioural assessment to ensure proper data analysis as previously reported (Dombret et al., 2020; Naulé et al., 2015). Adult males and females will be evaluated for general (locomotor activity, motor coordination and anxiety-state level) and cognitive behaviors (spatial memory, temporal order memory, novel object recognition) as described by the team (Ducroq et al., 2023; Picot et al. 2016). At the end of the behavioural assays, blood, brain and thyroid will be collected. Hormonal levels (thyroxine, triiodothyronine and TSH) will be analyzed. The brains will be processed for dendritic spine density quantification (Golgi-Cox staining) and ultrastructural observations by electron microscopy in the hippocampus and cortex as recently conducted (Ducroq et al., 2023). Immunohistochemical analyses of GFAP, Iba-1 and inflammatory markers (COX, iNOS…) will be conducted (Ahmadpour et al., 2021). The mRNAs expression levels of presynaptic and postsynaptic proteins (synaptophysin, PSD-95, spinophilin), glutamate receptors (NR1, NR2A, NR2B…), thyroid hormone receptors and transcription factor recently identified as targets for EDC exposure (Aryl hydrocarbon receptor) (Ducroq et al., 2023) will be analyzed. In addition, a non-targeted approach using RNAseq will be performed to identify still unknown biomarkers in the hippocampus and cortex (ICM platform).

ii) Human hippocampal neurons for the in vitro study
For the translational relevance of the effects from mice to humans, a human in-vitro model of hippocampal neurons recently set up by the team using iPSC technology will be used. It will also allow to dissect the mechanisms of exposure by assessing separate substances, with a particular focus on those driving neurodevelopmental and thyroid effects.
Briefly, hippocampal differentiation of human iPSCs will be performed using established neural differentiation protocol developed in the host laboratory, and based on previous studies (Yu et al. 2014; Sarkar et al. 2018). After 6 weeks of differentiation, hippocampal neurons will be exposed to the whole mixture and separate substances for a long-term period of 14 days. Hippocampal neurons, that express thyroid hormone receptors alpha (THRA) and beta (THRB) and deodinases 2 and 3 (Sarkar et al. 2018), will be exposed to substances of the mixture. Key neurodevelopmental events including proliferation, and neuronal maturation and neuronal network formation/synaptogenesis will be analyzed. Proliferation will be evaluated using the luminescence-Bromodeoxyuridine assay and neuronal migration in hippocampal neurons will be tested by the scratch method. Dendritic and axonal (neurite) morphology will be characterised in hippocampal neurons. For this, the number of neurites per neuron, the length of the outgrowth and the number of branches will be measured in beta-III tubulin-positive cells. In addition, protein levels of presynaptic (synaptophysin, synapsin 1) and postsynaptic markers (PSD95, gephyrin, glutamate receptors) will be quantified by Western Blot to measure the impact of treatments on synaptic plasticity. Finally, the interesting biomarkers identified from the in vivo study will be also analyzed in this in vitro model.

The original data generated from the project will document for the first time the effects and underlying mechanisms of environmental doses of human breast milk chemical mixture on the developing brain. The relationships between key events ranging from molecular and cellular responses to the potential behavioral perturbation will be used to build adverse outcome of pathways, that can serve for risk assessment for human health. The whole project aiming at the promotion and validation of new methodological approach will have a great impact in the context of environmental health in general and the current regulation in particular. Indeed, under the European regulation and worldwide, chemical risk assessment for human health or environment is still performed compound by compound and at irrelevant high doses, far from the real environmental exposure.

The prefrontal cortex (PFC) is the brain region which covers the front part of the frontal lobe of the cerebral cortex and underlies cognitive processes. This brain area exhibits spontaneous neuronal activity, which is altered in several psychiatric disorders including, schizophrenia (SCZ). Evidence identified that two systems, cortical inhibitory circuits and cholinergic neurotransmission, are affected in SCZ, however, the actual interplay between these two pathophysiological mechanisms is unknown. A key hub structure that integrates these two systems is cortical layer 1 (L1), the most superficial layer of the cortical column. This project will use multiscale experimental approaches, combining genetics, behavior, imaging, electrophysiology and optogenetics to understand i) how cholinergic transmission affects specific L1 subnetworks and ii) the cellular, molecular and synaptic logic governing the modulation of L1 circuits by long-range cholinergic afferents in normal and SCZ brain.

The prefrontal cortex (PFC) is critical for cognitive processing with a central role in executive functions. Alterations in PFC circuits have been linked to psychiatric disorders, such as schizophrenia (SCZ). The PFC is organized in functional layers with concerted activity between these layers being crucial for cognition. In particular, the most superficial layer 1 (L1) has been attracting attention because it acts as a hub, integrating bottom-up and top-down information during cognitive behavior. L1 is a unique layer since it is composed of a sparse population of different types of GABAergic INs, the dendrites of pyramidal neurons and glutamatergic axons projecting from long-range cortical and subcortical areas (1). Among the many top-down afferents to L1, the basal forebrain provides a major source of cholinergic projections to this layer. Cholinergic transmission governs several arousal states and is essential for cognitive functions (2).
L1 INs exhibit diverse morpho-functional properties and were recently classified into three major subclasses based on their selective expression of either neuron-derived neurotrophic factor (NDNF), vasoactive intestinal peptide (VIP) or a7 nicotinic acetylcholine receptor (a7-nAChRs) (1). These distinct populations of L1 INs have been implicated in feed-forward and disinhibitory circuits (3, 4). NDNF INs account for the majority of L1 INs, and inhibit the apical dendrites of pyramidal neurons (1, 3). Yet, it is unknown how L1 INs decode information from projecting cholinergic inputs to the PFC and their ultimate functional role during cognitive processing.
Notably, dysfunctional changes in both inhibitory INs and cholinergic neuromodulation are thought to be core features for a range of psychiatric disorders, including SCZ (5). Recently, it was suggested that general network impairment within upper cortical layers is a core substrate associated with SCZ symptomatology and goes beyond the well characterized parvalbumin expressing INs, involving several types of GABAergic INs (6). Recent genome-wide association studies have identified human polymorphisms of AChRs linked to SCZ (7). Despite the mounting evidence for the role of INs and cholinergic transmission in SCZ, the actual interplay between these two pathophysiological mechanisms is completely unknown. Recent scientific and technological advances allow probing and manipulating specific cholinergic and local cortical GABAergic circuits. Here, we aim to address the following questions: i) Are PFC L1 INs recruited by cholinergic afferents during cognition? Which specific L1 GABAergic circuit is altered in SCZ? ii) What are the molecular, cellular and synaptic fingerprints underlying the cholinergic recruitment of the heterogeneous L1 IN population? How is this circuit logic altered in SCZ?

Task 1: Cholinergic recruitment of L1 INs in PFC during behavior in health and disease

Considering their distinct cellular properties, it is possible that L1 IN subtypes exert different circuit functions, therefore subtype specific approaches are required to precisely identify their function in PFC.
Cortical INs are differentially modulated by the animal’s behavior and behavioral state. Our preliminary data indicate that L1 INs in the prelimbic cortex of the PFC are highly recruited during learning in an operant conditioning task. Studies in humans showed SCZ patients consistently having deficits in tasks specific to several types of learning, including associative learning (8). We hypothesized that cholinergic modulation of L1 INs is crucial for their recruitment during learning and that this modulation is affected in SCZ. To address this hypothesis, we will perform 2-photon Ca2+ imaging and record the activity of distinct L1 IN types using relevant Cre mouse lines during behavior in control animals and mice expressing human single nucleotide polymorphisms (SNPs) of AChRs linked to SCZ. To quantify the relationship between cholinergic signaling, neuronal activity and behavior, we will use a genetically encoded reporter for ACh. To further elucidate how L1 INs modulate L2/3 pyramidal neuron computations, we will optogenetically stimulate the axons of L1 INs in vivo and record the activity of their output targets in L2/3. These experiments will be combined with extracellular recordings of pyramidal neurons in PFC of behaving mice.

Task 2: The functional architecture and connectivity profile of L1 INs and how it is changed in psychiatric disease

ACh release during the physiological recruitment of the basal forebrain, can rapidly influence the firing of L1 INs. Yet , this layer is composed of a diverse population of interneuron subtypes. We hypothesize that alterations of cholinergic recruitment of specific GABAergic L1 subnetworks are responsible for network impairments linked to SCZ. To test this hypothesis, we will inject the nucleus basalis of ChAT-Cre (choline acetyltransferase -Cre) mice with light-sensitive opsins (ChR2) and perform whole cell recordings on different subtypes of L1 INs while photo-activating ChR2-expressing fibers in acute slices of control and SNP mice. We will combine optogenetic interrogation of endogenous cholinergic recruitment of different L1 INs with local pressure application of ACh and relevant pharmacology to test the biophysical properties of L1 INs. To define the connectivity profiles of different L1 IN subclasses, we will perform paired recordings between different L1 INs and their postsynaptic targets in layers 2/3.
Overall, we will use multiscale experimental approaches, to provide fundamental insights into the cellular mechanisms of cortical function in health and SCZ.

1. Schuman et al. J. Neurosci. 39, 125–139 (2019).
2. Ballinger et al. Neuron. 91, 1199–1218 (2016).
3. Abs et al. Neuron. 100, 684-699.e6 (2018).
4. Letzkus et al. Nature. 480, 331–335 (2011).
5. Marín Nat. Rev. Neurosci. 13, 107–120 (2012).
6. Batiuk et al. Sci. Adv. 8 (2022), doi:10.1126/SCIADV.ABN8367.
7. Ripke et al. Nature. 511, 421–7 (2014).
8. Hall et al. Trends Neurosci. 32, 359–365 (2009).

Often, behaviors are not single actions but multiple actions organized into sequences that allow the animals to achieve their goals. Therefore, mechanisms must be in place to regulate the transition between the actions in a sequence. The network architecture underlying sequence generation is not well known This is primarily due to the challenges in mapping connectivity with synaptic resolution and establishing causality between neurons and sequences with cellular resolution. We propose to fill this knowledge gap by using a multidisciplinary approach that will combine functional imaging, neural manipulation during behavior, quantitative behavioral analysis and electron microscopy (EM) reconstruction of neuronal connectivity with synaptic resolution in the tractable model system, the Drosophila larva. We aim at determining with unprecedented resolution the circuit mechanisms underlying action sequences. This will lay the basis for future investigations in larger, more complex systems.

Neuronal control of action sequences

Background
A variety of behaviors that can vary from the essential, such as finding food, escaping predators or mating, to the more creative, such as playing music, dancing, or engaging in other forms of play are organized in sequences. This implies that the nervous system needs to be able to set the order of the different actions in the sequence and also regulate the transitions between these actions: determine when one action ends and when another starts. How the order of actions in a sequence is implemented and how the transitions to the subsequent actions are controlled to ensure the progress of the sequences at the neural circuit level remain an open question. To generate action sequences the nervous system needs to regulate transitions from one action to the next and at the same time prevent reversals from later actions back to earlier ones to ensure orderly progression within a sequence. There are two main hypotheses that try to explain how the serial organization of actions in a sequence is established. One hypothesis suggests that modules that promote earlier actions in a sequence provide the excitation for the module that promote the following action. Probabilistic sequences, such as human typing or fly grooming, are better described by models that propose all actions in a sequence are readied in parallel and the order is established through gradients of excitation and winner-take-all competition. However, it has been difficult to determine the circuits that would implement sequences consistent with either of the models. We propose to fill this gap in a powerful model organism for neural circuit analysis: the Drosophila larva. Drosophila larvae are ideally suited for combining comprehensive, synaptic-resolution circuit mapping across the nervous system with targeted manipulation of uniquely identified circuit motifs at the individual neuron level, which makes it possible to establish a causal relationship between circuit structure and function brain-wide (Eschbach and Zlatic, 2020; Jovanic, 2020; Jovanic et al., 2016).

Research program and organisation
We have previously identified the neural substrates underlying sequence transitions in the ventral nerve cord of the Drosophila larva (Jovanic et al., 2019, 2016; Masson et al., 2020). In order to determine neural circuit mechanisms underlying sequence tranisitons we will combine multiple approaches for neural circuit and behavioral analysis. We will use synaptic-resolution connectomics (reconstruction of neurons and synapses in electron microscopy) to determine the neural circuit architecture underlying sequence transition. Furthermore, we will perform neuronal manipulation (silencing/activation) of these neurons during behavior/imaging and quantify changes in sequences that occur upon these manipulations using automated algorithms and motor patterns analysis. Altogether this will allow us to determine the neural circuit mechanisms underlying action sequences. The research project will be organized in two main aims:
Aim. 1. Charcaterize neuromuscular patterns underlying sequence transition Ongoing work from in the lab has characterized the fictive behaviors: patterns of motor activity at the level of the motor neuron population characteristic of the different actions that larvae do in response to external stimuli : Hunch (head-retraction, Bend, Coral, Back-Up, Roll- a fast escape behavior, C-shape etc.). These behaviors are each characterized by specific patterns of left/right symmetric/asymmetric and synchronous/asynchronous activity along the different segment of the VNC. Using functional muscle calcium imaging we will further visualize individual muscle contraction pattern(s) underlying different stimulus induced behavioral transitions. This will provide insight into the neuromuscular mechanisms underlying sequence transitions at single cell (individual muscle fiber resolution). Linking the muscle and motor neuron activity for each action will provide the full characterization of the motor and muscle pattern of neuronal activity for the different transitions between actions
Aim 2. Determine the circuit mechanism underlying probabilistic sequences. Previous work previously shown that lateral disinhibition and feedback disinhibition implement a probabilistic sequence of two responses to a mechanical stimulus: lateral disinhibition promotes transitions from hunch to bend and feedback disinihibition prevents reversal from bend to hunch (Fig. 1). Whether this is a general mechanism used for longer sequences is unknown. We hypothesize that a chain of such circuit motifs could be a general mechanism of sequence generation and could implement longer behavioral sequences in the Drosophila larva and other organisms as well. Such a mechanism has similarities with the proposed synfire chain models but the use of disinhibition interneurons instead of direct excitation enables flexibility in the order of elements in the sequence. For example, context, or experience could modulate the activity of the disinhibitory interneurons to change the balance between reversals and forward transitions. To identify the circuit motifs that implement the probabilistic sequence and test the above hypotheses, we will determine the hierarchical order of the behavior by coactivating the neurons for different actions using optogenetics. We will then investigate how these neurons are structurally and functionally interconnected using EM reconstruction of neuronal connectivity, functional connectivity studies, modelling and manipulating specific circuit elements in freely behaving animals. Using dual color imaging we will corelate the activity these neurons with changes in motor pattern activity (sequence transitions). We expect to find lateral disinhibitory motifs between different action modules promoting transitions between different actions and complementary feedback disinhibitory motifs that prevent reversals. This would consist a novel, possibly general mechanism of circuit implementation of longer flexible sequences.
Finally incorporating connectomics and functional connectivity data with behavioral, motor and muscle pattern activity data will allow to make computational models in collaboration with the Masson team.

Project feasibility and collaborations
The project although ambitious and combining multiple approaches is feasible in the 3 year timeline for a PhD, thanks to:
-the extensive preliminary work on characterization of sensorimotor circuitry (Jovanic et al., 2019, 2016; Masson et al., 2020, Lehman et al., to be submitted) which allows to immediately start with candidate neurons for circuit mapping and functional testing
-Existing behaviroal set-ups and analysis pipeline for automated behavioral tracking, action classification and action sequence analysis (Masson et al., 2020, Lehman et al., to be submitted)
-Established preparations and analysis methods for motor pattern analysis
-a proven record of successful and synergistic collaborations with a Masson team (Jovanic et al., 2016; Masson et al., 2020, a co-supervision of A. Zhou, 3rd PhD student (Zhou et al, in preparation), as well as from other ongoing collaborations (e.g. Cardona, LMB-MRC, Cambrdidge for EM reconstruction and E.Heckscher, U. Chicago for muscle imaging)
- The Jovanic team’s research is supported by funding from CRCNS (DOE-ANR), ANR, FRC and FRM.
References
Eschbach C, Zlatic M. 2020 Curr Opin Neurobiol 65:129–137. doi:10.1016/j.conb.2020.09.008
Jovanic T. 2020.. J Neurogenet 55:1–9.
Jovanic T, et al., 2016. Cell 167:858-870.
Jovanic T, Winding M, Cardona A, Truman JW, Gershow M, Zlatic M. 2019. Curr Biol CB 29:554-566.
Masson J-B, …Jovanic T. 2020. Plos Genet 16:e1008589.

Narcolepsy, a rare neurological disorder marked by symptoms like daytime sleepiness, cataplexy, hallucinations & sleep paralysis, presents challenges. Cognitive issues affecting memory, learning and attention are often linked to daytime sleepiness. Narcolepsy results from a complex interplay of genetic and environmental factors. Recent research connects hereditary factors with narcolepsy and ADHD traits. Environmental risk factors like infections and vaccines are also implicated. Yet, their role in cognitive impairments is not fully understood. This study provides valuable insights into the genetic, socio-economic-demographic, and environmental factors that contribute to the understanding of cognitive impairment in narcolepsy. Cognitive assessments are conducted in a sample of 9-15-year-old children with narcolepsy and matched controls from narcolepsy reference centers. Findings hold significant potential for uncovering mechanisms, preventing and ultimately reducing the disease burden

Narcolepsy, a rare neurological disorder characterized by symptoms such as excessive daytime sleepiness, cataplexy (a sudden muscle weakness that occurs while a person is awake), hallucinations, and sleep paralysis, presents significant challenges for affected individuals. Many narcoleptic patients struggle with memory, learning, and concentration issues, which can impact their daily lives and even lead to serious accidents in daily life activities. Their cognitive impairments are often attributed to daytime sleepiness and sleep attacks. Narcolepsy therefore profoundly impacts the quality of life for those affected, with health-related quality of life reported as very poor.
Cognitive impairment is a common concern among children with narcolepsy, encompassing issues with memory, attention, and learning. While empirical evidence regarding cognitive impairments in narcolepsy has been somewhat inconsistent, it is well-established that patients often experience deficits in attention and executive function. These cognitive challenges may explain their reported difficulties in scholastic activities, such as learning, and are typically associated with slower reaction times and more variable performance compared to control groups. Recent research has also uncovered genetic risk factors associated with narcolepsy, showing a link between these genetic factors and attention deficit hyperactivity disorder (ADHD) traits in the Japanese population investigated.
Narcolepsy itself may arise from a complex interplay of genetic and environmental factors that result in the selective loss or dysfunction of orexin neurons in the lateral hypothalamus. Genetic studies have demonstrated the significance of the HLA association, particularly HLA-DQB1*06:02, in patients with narcolepsy. While recent genome-wide association studies have identified various genetic associations with narcolepsy, the genetic basis for cognitive impairments in narcolepsy remains largely unexplored. Nevertheless, there is evidence suggesting a shared genetic background between narcolepsy and ADHD, with hyperactivity and inattention traits showing significant associations with narcolepsy polygenic risk scores. Environmental risk factors, including infections such as streptococcus and the H1N1 influenza virus, as well as the use of the AS03-adjuvanted vaccine Pandemrix®, have been linked to narcolepsy. However, the relationship between these genetic and environmental factors, and cognitive impairments in narcolepsy has not been adequately studied.

This research project has two primary objectives. First, it aims to identify the risk factors associated with the type and severity of cognitive impairments in narcolepsy. Second, it seeks to assess the interactions between the genetic and environmental factors and their combined impact on cognitive impairments in narcolepsy. This work has the potential to provide invaluable insights into the complex and multifaceted nature of cognitive impairments in narcolepsy, contributing to improved care and management for individuals affected by this condition.
This study prospectively compares the cognitive performance of children with narcolepsy to a control group matched for age and sex within the 9-15-year age range. The study will recruit participants from the Beijing People's Hospital (China) and the Hospital Robert Debré (France) - Narcolepsy Reference Centers, where a significant number of subjects are receiving treatment for narcolepsy.
Children complete a computerized neurobehavioral test battery every two hours during wakefulness. This battery consists of different compositions, alternating with the following components: the pediatric daytime Sleepiness Scale a Likert-type subjective sleepiness scale; the pictorial sleepiness scale; a visual analog scale of sleepiness; the Child Behavior Checklist a questionnaire that assesses emotional and behavioral problems in children and adolescents; the Digit Symbol Substitution Task, cognitive test assessing processing speed, attention, and working memory; the Digit Span task assessing short-term memory, working memory, attention, and the ability to manipulate information in one's mind, given in forward and backward versions; the Psychomotor Vigilance Task, a sustained attention test utilizing reaction times as a behavioral alertness assay; the d2 test assessing sustained attention; Trail Making Test assesses processing speed, visual attention, and cognitive flexibility; the 15-words of Rey evaluates verbal learning and memory; the Rey Complex Figure test memory, visuospatial abilities, and organizational skills and Raven progressive matrices, a non-verbal intelligence test designed to measure abstract reasoning and problem-solving ability. This in addition to the assessment of the children's neurodevelopmental profile carried out using the NEPSY, a comprehensive neuropsychological test battery designed specifically for pediatric subjects. This battery of tests offers valuable insights into various aspects of neurodevelopmental functioning; i.e., attention/executive functions, language, memory/learning, sensorimotor functions, visuospatial and sensorimotor processing. Narcolepsy polygenic risk scores (PRS) will be calculated for each individual by previously collected genome-wide association examination of narcolepsy: TRA, TRB, CTSH, IFNAR1, ZNF365, TNFSF4, CD207, IKZF4-ERBB3, NAB1, CTSC, DENND1B, SIRPG, and PRF1. Logistic regression modeling is used to calculate odd ratios (ORs) and 95% confidence intervals for each potential risk factor, including demographics, socioeconomic, environmental-related variables, and PRS. Risk factors with an overall likelihood ratio test P-value of <0.1 in age-adjusted models are included in models adjusted for (i) socioeconomic-demographic factors, (ii) socioeconomic-demographic factors and PRS, and (iii) socioeconomic-demographic-environmental factors and PRS. Supervised machine learning algorithms will be used to predict cognitive performance based on the various significant risk factors.
This study will provide valuable insights into the genetic, socio-economic-demographic, and environmental factors that contribute to the understanding of cognitive impairment in narcolepsy. The implications of the study are significant in identifying mechanisms for such impairments, while also aiding in their prevention and reducing the burden of the disease.

We propose to study the brain organization in terms of functional and structural connectomes as well as cortical sulcation. Those features may be extracted using the phenotyping tools developed at NeuroSpin from neuro-images in two general population cohorts UK-Biobank and ABCD. These cohorts also include genotyping data which give unprecedented opportunities to study the relationships between these phenotypes and the genome or the environment.

We propose to consider alternative methods to conventional genetic approaches. We will study deep-learning approaches inspired by variational auto-encoders adapted to multiple data modalities. We will articulate these networks with graphical networks to model biological apriori such as gene-gene interactions or genome structure. This work will lead to new insights into the anatomical and functional organization of the brain during neurodevelopment or aging. It will provide networks and frameworks for their biological interpretation.

Human cognitive and behavioural activity is supported by the functional activity of the brain through a variety of interacting networks. Neural activity results from the flow of information from one brain region to another. These pathways of information between regions are supported by structural and functional connectomes and the regional specialization of the cortex and subcortical nuclei.
The advent of population imaging studies, in particular the UK Biobank general population cohort (500,000 individuals, including 100,000 with neuroimaging) or the ABCD cohorts (10,000 adolescents with multiple visits), has allowed detailed investigation of structural and functional connectomes as well as the cortical sulcal organization. Indeed, neuroimaging provides detailed evidence of the organization underlying human behavioral and cognitive activity, as well as the effects of neurodevelopment, environment, and aging on these structures. The simultaneous availability of the genotyping in these two cohorts (chip and whole exome sequence) offers a unique opportunity to study and characterize the links between genetics, neurodevelopment, environment and brain structure [1], [2].

Univariate genetic association analysis and heritability estimation tools applied to population cohorts have begun to reveal the genetic architecture of brain organization [3]. More recently, even more information about brain organization and development is being provided by tools derived from structural equation models applied to genomics (such as genomicSEM), which focus on pleiotropic phenomena and genetic correlation [4].

Deep learning is a relatively new and emerging technique as an alternative to these traditional genetic approaches in the field of basic life science research. Although still in its infancy, deep learning has attracted considerable attention due to remarkable advances, notably through the integration of transformers in AlphaFold networks for protein structure prediction and the application of variational autoencoders (VAEs) to the analysis of single-cell sequence RNA (c-seqRNA) data, enabling unsupervised categorization and subtyping of individual cells.

We believe that multimodal integration of omics and neuroimaging is a key approach to knowledge generation in fundamental neuroscience, and that deep learning is a productive framework for deploying multimodal integration approaches. Multi-channel VAEs provides a starting point for multimodal integration [5,6]. We propose to leverage such models to develop tools aimed at deriving knowledge from functional and structural connectomes associated with cortical sulcation data. The planned work includes the following tasks: i) determining the optimal dimensionality of the latent spaces within VAEs [7], ii) formulating cost functions that facilitate the separation of information encapsulated in the latent spaces of different data modalities, iii) designing training strategies that involve downstream tasks to focus the network learning, iv) exploring integration with attention-based hierarchical graph convolutional networks that can account for gene-gene interactions or genome structure (linkage disequilibrium) [8].

This work will be possible at NeuroSpin due to the wide range of phenotyping tools available and access to UK Biobank and ABCD. Other clinically focused cohorts could be used for replication or prediction purposes (IMAGEN-STATIFY, EU-AIMS, HBN, MEMENTO). This work will lead to new insights into the anatomical and functional organization of the brain during neurodevelopment or aging and produce original maps or trajectories. It will provide networks and frameworks for their biological interpretation.

Bibliography (lab contributions are marked with a *).
[1]* Y. Le Guen et al., ‘eQTL of KCNK2 regionally influences the brain sulcal widening: evidence from 15,597 UK Biobank participants with neuroimaging data’, Brain Structure and Function, vol. 224, no. 2, pp. 847–857, Dec. 2019, doi: 10.1007/s00429-018-1808-9.
[2]* Y. Mekki et al., ‘The genetic architecture of language functional connectivity’, NeuroImage, vol. 249, p. 118795, Apr. 2022, doi: 10.1016/j.neuroimage.2021.118795.
[3] L. Shen and P. M. Thompson, ‘Brain Imaging Genomics: Integrated Analysis and Machine Learning’, Proceedings of the IEEE, vol. 108, no. 1, pp. 125–162, Jan. 2020, doi: 10.1109/JPROC.2019.2947272.
[4] V. Warrier et al., ‘Genetic insights into human cortical organization and development through genome-wide analyses of 2,347 neuroimaging phenotypes’, Nat Genet, Aug. 2023, doi: 10.1038/s41588-023-01475-y.
[5] Aguila, Ana Lawry and Jayme, Alejandra, ‘Multi-view-AE’. [Online]. Available: https://multi-view-ae.readthedocs.io/en/latest/documentation/implemented_models.html
[6]* C. Ambroise, A. Grigis, E. Duchesnay, and V. Frouin, ‘Multi-View Variational Autoencoders Allow for Interpretability Leveraging Digital Avatars: Application to the HBN Cohort’, in 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023, Cartagena, Colombia, April 18-21, 2023, IEEE, 2023, pp. 1–5. doi: 10.1109/ISBI53787.2023.10230552.
[7] L. Antelmi, N. Ayache, P. Robert, and M. Lorenzi, ‘Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data’, PMLR, May 2019. [Online]. Available: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf
[8] S. Ghosal et al., ‘A Biologically Interpretable Graph Convolutional Network to Link Genetic Risk Pathways and Neuroimaging Markers of Disease’. bioRxiv, p. 2021.05.28.446066, Mar. 24, 2022. doi: 10.1101/2021.05.28.446066.

Microcephaly is a common cause of intellectual disability, with underlying mechanisms still poorly understood, and over 50% of patients with no known molecular basis. The project aims at dissecting the pathological mechanisms associated to a novel candidate gene of microcephaly encoding a protein with dual subcellular localization to centrosomes and nucleosomes suggesting a complex pathological mechanism involving microtubule-based structures (centrosome, primary cilium & mitotic spindle) and epigenetic-regulated processes (DNA replication & gene expression). To test this hypothesis, cutting-edge cell-imaging and multi-omics analyses will be carried out on brain organoids derived from patient IPSCs. Overall, this project will improve the diagnosis of microcephaly and our understanding of the underlying mechanisms. It will also broaden our knowledge of the emerging role of the primary cilium during corticogenesis, and help deciphering the epigenetic mechanisms regulating neocortex size.

The cerebral cortex, which computes the high-level sensory, motor and cognitive processes, has expanded dramatically during evolution. Corticogenesis is a complex and finely tuned developmental process which orchestrates the organization of many types of neurons through tight regulation of the survival, expansion, fate determination and differentiation of neural stem and progenitor cells (NSPC). Disruption of any of the corticogenesis steps may result in anomalies of cortical development, encompassing a large spectrum of disorders representing a major cause of intellectual disability and which classification remains challenging. Among them, microcephaly is characterized by prenatal reduced brain size and has been largely linked to mutations in genes encoding proteins involved in the proliferation of NSPC, especially many centrosomal proteins. Centrosomes function as microtubule-organizing centers which form the spindle poles during mitosis while they nucleate a primary cilium (PC) in G0/G1 phase. The impact of centrosomal dysfunction on the mitotic spindle has long been considered a mechanism underlying microcephaly, whereas the consequences on primary cilium dynamics and function have emerged only recently. Primary cilia are highly conserved microtubule-based organelles required for sensing and transducing various extracellular signals essential for coordinating diverse key processes of corticogenesis. In fact, they are present on all NSPC of the developing neocortex and are crucial for NSPC expansion, fate determination, neuronal migration and maturation, highlighting the need for further delineation of the roles of primary cilium during corticogenesis.
Whole exome sequencing analysis allowed us to identify a novel candidate gene for microcephaly, encoding a centrosomal protein, in which de novo truncating mutations were identified in 6 patients from 5 distinct families. Beyond the strong genetic arguments and the fact that it encodes a centrosomal protein with potential consequences on both mitotic spindle and primary cilium, this gene is an excellent candidate because of its high expression level in the ventricular zone of the developing human cerebral cortex. Moreover, this protein is part of a histone-acetyl-transferase complex, which raises a fundamental question about epigenetic mechanisms and their contribution to the regulation of neocortex development. To validate this gene as a microcephaly causing gene in human, the PhD candidate will generate 2D and 3D cell-based models of neocortical development, i.e. neural rosettes, isolated NSPCs and cerebral organoids, using described protocol established by the team (Boutaud et al. Jove 2022). For this purpose, blood cells from 3 patients have already been reprogrammed into induced pluripotent stem cells (IPSCs), which pluripotential capacities and genome integrity have been checked. In order to obtain isogenic controls, each IPSCs line will be edited by CRISPR/CAS9 to specifically correct each patient mutation. Such mutated and rescued IPSCs will be then used to generate complementary 2D and 3D cell-based models of cerebral development that will allow us to study the impact of the mutations of our novel candidate gene on NSPC proliferation, survival and differentiation. In addition, cell imaging analyses thanks to lightsheet, confocal, and expansion microscopy will allow us determine potential anomalies affecting mitotic spindle formation and/or orientation, primary cilium assembly/disassembly dynamics and/or function, including the transduction of signaling pathways dependent on the primary cilium. To analyze the consequences on histone acetylation and their potential impact on the regulation of gene expression, ATAC-seq and Cun&Run analyses will be coupled to RNA-seq studies. Finally, potential DNA replication defects will be tested through a DNA combing-based approach.
Thus, by combining cutting-edge cell imaging and multiomics analyses on relevant cellular models, this project aims at dissecting the pathophysiological mechanisms underlying microcephaly associated with mutations of this novel gene. Beyond the obvious clinical impact on the molecular diagnosis of microcephaly for affected individuals and their families, and on the understanding of the cellular mechanisms underlying microcephaly, this project will improve our knowledge concerning the emerging role of the primary cilium during normal and pathological cerebral cortical development, especially in sensing and transducing key extracellular signals. This project also provides the opportunity to further dissect the epigenetic mechanisms involved during neocortex development in human.

Head-direction (HD) cells are neurons whose activity is tuned to the orientation of the animal's head. These neurons have been found in many species and are central to spatial navigation as they act collectively as a neuronal compass. In this project, our aim is to study how the particular “ring attractor” architecture of HD cells circuits emerges during development, and to clarify how visual and vestibular sensory inputs together drive the dynamics of the HD circuit to ensure an accurate representation of the animal's orientation. In order to address these questions, we will use Danionella cerebrum, a novel model fish whose brain remains small and quasi-transparent throughout development. We will combine In vivo calcium imaging during fictive swimming and optogenetic circuit interrogation. This project will provide a unique view on the functional maturation of a complex neuronal circuit in the vertebrate brain.

Head direction (HD) cells are neurons that activate when an animal orients itself in a specific direction. Collectively, these neurons form a neural compass and thus play a central role in spatial representation. Although initially discovered in rodents [1], HD cells have been identified in various species, including flies [2] and more recently in zebrafish larvae [3].

The capacity of HD circuits to encode spatial orientations is thought to reflect their particular recurrent architecture that confers them the property of ring attractors: they thus possess a 1D continuum of stable configurations, each of them encoding a distinct orientation [4]. Furthermore, numerous models have been proposed to explain how visual stimuli and proprioceptive inputs concur to drive HD cells during navigation such as to ensure an accurate representation of the animal's orientation.
In the present project, our aim is to understand how the architecture and the associated functions of this circuit emerge and get refined during vertebrate brain development. In order to address this question, we will use Danionella cerebrum (DC), a recently introduced model vertebrate system [5,6]. DC is a small freshwater fish whose brain remains small and almost entirely transparent until adulthood. This exceptional trait allows one to monitor the brain's complete activity with cellular resolution across development using calcium imaging [6].

The PhD project is structured into several stages:

1) First, we will use a virtual reality setup developed in our lab to record brain activity as the fish fictively navigates within a virtual environment. Specific closed-loop protocols and analysis methods will be developed to functionally identify HD cells in the DC brain. This stage will be facilitated by the proximity of zebrafish and DC at larval stage and by the fact that HD cells have been recently identified in larval zebrafish [3].

2) Having defined a protocol for the functional identification of HD cells, we will then probe the changes in the circuit functioning during development, focusing on the emergence of stable (attractor) configurations and their directional tuning. This phase will elucidate how these neuronal populations progressively acquire the characteristics of a linear (ring) attractor. Analytical and computational models of circuit maturation will be derived from this data, revealing fundamental maturation principles.

3) We will further evaluate how this HD circuit integrates sensory stimuli to precisely keep track of the fish orientation. The most simplest modality is the visual system, as it is possible to induce animal’s reorientation using rotational visual patterns (through a reflex called OMR). Here we will also be able to test the contribution of the vestibular signals as our group has designed in recent years unique capacities to record brain activity during vestibular stimulation [7,8]. Interestingly, the semicircular canals, which provide information regarding the head rotation in the horizontal plane, become functional after a few weeks in zebrafish larvae. It is thus likely that this information is only available, and thus integrated only progressively during the development of the larva.

4) Finally, we will test these models by combining volumetric functional imaging with 2P targeted optogenetic activation, utilizing a unique setup recently developed [9]. This approach will provide a more detailed examination of the functional organization of HD circuits by specifically activating subpopulations of HD or HD-projecting neurons.

Context: The Laboratoire Jean Perrin (LJP) is a biophysics laboratory located on the P&M Curie campus. The lab mostly comprises physicists interested in biological systems. The LJP is part of the Institute Biology Paris Seine (IBPS), a research federation that gathers all Biology labs on campus. Our research group, currently consisting of three permanent researchers, four PhD students, and one post-doc, is dedicated to unraveling the neuronal basis of sensory-driven and spontaneous behavior using zebrafish and Danionella as model vertebrates [10]. We develop customized bio-imaging setups for whole-brain functional recording and optogenetic activation while conducting quantitative behavioral assays. Furthermore, we are heavily invested in developing computational methods inspired by statistical physics and machine learning for the modeling and analysis of these extensive datasets.

[1] J.S. Taube, R.U. Muller, J.B. Ranck Jr, J. Neurosci. 10(2):420-35 (1990)
[2] Seelig, J.D., and Jayaraman, Nature 521, 186–191 (2015)
[3] Petrucco et al., Nature neuroscience, volume 26, pages 765–773 (2023)
[4] Kim, S.S., Rouault, H., Druckmann, S., and Jayaraman, V. Science 356, 849–853 (2017)
[5] Schulze et al. Nature Methods volume 15, pages 977–983 (2018)
[6] G. Rajan, …, G. Debrégeas , C. Wyart, , F. Del Bene, Cell Reports, 38, 13 (2022)
[7] G. Migault , T.L. van der Plas , H. Trentesaux , T. Panier , R. Candelier , R. Proville , B. Englitz , G. Debrégeas, V. Bormuth, Current Biology, 28, 1-13 (2018)
[8] N. Beiza-Canelo, …, G. Debrégeas, V. Bormuth, Current Biology 33, 12, 2438-2448.e6 (2023)
[9] A. Hubert , M. Dommanget-Kott , S. Wolf , T. Panier , G. Debrégeas , V. Bormuth, Advances in Microscopic Imaging IV, 12630, 11-14 (2023)
[10] https://www.labojeanperrin.fr/?article6

Neurodevelopmental disorders (NDDs) affect 10% of children, leading to lifelong cognitive and behavioral handicaps. NDDs exhibit significant clinical variability, which hampers therapeutic solutions. Our project focuses on a shared aspect of NDDs: the disruption of stress response pathways involving Heat Shock Factors (HSFs).
Heat Shock Factors (HSF2), which govern proteostasis in stressful situations also contribute to brain development under physiological conditions. Therefore, using two NDD models derived from patient iPSCs - both demonstrating HSF2 dysregulations - alongside their HSF2-restored isogenic models, we aim to isolate the specific role of the HSF pathway in NDDs. We will use cutting-edge single-cell approaches, applied to cerebral organoids derived from these NDD models, to identify HSF-dependent candidate genes, pathways and cell populations. These findings will undergo validation through independent approaches and will serve as a 'readout' for future drug screening.

=>Background
- Hallmarks of stress in neurodevelopmental disorders (NDDs)
NDDs affect approximately 10% of children, leading to lifelong cognitive and neuropsychiatric disabilities (1). They can arise from both genetic and environmental factors, resulting in a broad clinical spectrum that challenges diagnosis and treatment. Regardless of their genetic or environmental origin, NDDs share common cellular stress-related dysregulations (2-5) in processes such as the endoplasmic reticulum stress response, the oxidative stress response, inflammation pathways, and mitochondrial activity (6-9). These pathways are likely interconnected and potentially linked (directly or indirectly) to Heat shock factors (HSFs,10-13), which belong to the family of stress-sensitive transcription factors. HSFs are important integrators of cell stress, development, and lifespan, bridging the gap previously seen as separate between cellular stress responses and normal physiology (14).
- Alteration of the HSF pathway in NDDs. HSFs, as heat shock-inducible transcriptional regulators, control the expression of a wide range of genes, including the one coding molecular chaperones (HSPs) and, those related to epigenetic processes. HSFs involvement in normal brain development is supported by the alteration of neural progenitor proliferation and radial neuronal migration upon HSF2 knock-down(15-18). Moreover, dysregulations of the HSF pathway are found in several pathological conditions (e.g cancer, inflammation, neurodegeneration) and aging. In various NDDs, an altered HSF activation was demonstrated: in Fetal Alcohol Syndrome (FAS) caused by alcohol exposure, and two NDDs of genetic origin, the Rubinstein-Taybi syndrome (RSTS,19) and an Angelman-like Syndrome (AGS-like,20). This strongly reinforces the hypothesis that HSFs could connect stress and neurodevelopment.
- HSF2 involvement in two NDDs, RSTS and Angelman-like syndrome:
*in RSTS (21), caused by mutations in CBP or EP300, we found a dysregulation of the HSF pathway, and notably decreased HSF2 protein levels(19). Using RSTS patient-derived cerebral models, generated from induced pluripotent stem cells (iPSC), we show that disruption of the CBP/EP300-HSF2-N-cadherin cascade results in impaired stress response and abnormal neurodevelopmental gene expression, both rescued by restoring HSF2 protein levels.
*in AGS (22), known to be caused mainly by UBE3A gene loss has, in some cases, an unknown genetic cause (described as AGS-like). Recently, an AGS-like patient was associated to a mutation in HSF2 (20). This first mutation in HSF2, identified as causative in an NDD, reinforce our hypothesis linking HSF deficiency to the neurodevelopmental defects observed in NDD.
Using these two NDD models as a paradigm will allow us to isolate the specific role of the HSF pathway in the context of NDD pathology with a particular focus on its potential interaction with other stress pathways.

=> RESULTS: Distinct phenotypic/molecular signature in RSTS related to HSF2 pathway dysregulation
To investigate the role of the HSF pathway in NDDs, we have generated RSTS-derived 2D (neural progenitors, iNPCs) or 3D (cerebral organoids, hCOs) neural models. In both models, in addition to decreased HSF2 protein levels, we identified the dysregulation of important proteins for stress and cell adhesion during neurogenesis, HSPs and N-cadherin, respectively. The loss of the N-cadherin network in the neurogenic niche alters cell adhesion and polarity, causing the loss of typical radial organization in RSTS iNPCs. As a result, the typical positioning of mitoses at the neuroepithelium apical surface is disrupted in RSTS hCOs. Furthermore, in line with the neurogenic niche disorganization (23), we have observed premature neuronal differentiation in RSTS hCOs as compared to healthy donor (HD), as revealed by the expression of the early cortical neuron marker TBR1. To fully characterize the consequences of HSF dysregulation on neural differentiation, we recently conducted a multi-ome single-cell (SC omics) analysis on hCOs, derived from RSTS patient and HD iPSCs at two differentiation stages (DIV25 and DIV45). While our analysis pipeline is still undergoing optimization, preliminary findings confirm the dysregulations in the HSPs and cadherin superfamily.
These promising results set the basis for the further development of HSF-rescue models that will allow to specifically isolate the HSF contribution in each NDD model, RSTS and AGS-like. To that goal, we will develop novel iPSC models to be derived in hCOs and iNPCs and expand our SC omics analysis to encompass these new isogenic models.

=>Impact of the restoration of the HSF pathway in RSTS and AGS-like
1- We employed Crispr/Cas tools to introduce genomic mutations at the HSF2 locus of RSTS iPSCs, producing a mimic of the acetylated HSF2 protein (HSF2-3KQ) that is less sensitive to proteasomal degradation and therefore “rescue” HSF2 levels. Following a similar analytical pipeline as used for HD vs. RSTS hCOs, we plan to conduct an unbiased analysis on RSTS isogenic clones, +/-rescue for HSF2. This analysis will focus on identifying: i) The most stress-susceptible neural population. ii) Altered HSF2-dependent pathways, as well as other interconnected stress pathways that may serve as potential read-outs. iii) Chromatin accessibility (correlated to gene expression) and transcription factor (TF) occupancy, especially focusing on TF involved in stress pathways (e.g. Nrf2, known to share overlapping targets with HSF).
2- To fully isolate HSFs, and specifically HSF2 contribution, we obtained blood cells of a patient with AGS-like syndrome, in which mutation in the HSF2 gene was discovered20 (thanks to Dr. Anna Ruiz (Hospital Parc Tauli, Spain)). We are currently generating AGS-like iPSC clones at the IMAGINE Platform and cells will be available soon (Oct2023). To rule out any possible influence of the patient’s genetic background, we will perform Crispr/Cas gene editing to restore the HSF2 mutation on the AGS-like iPSCs, and then use these two isogenic cell lines to further investigate neurodevelopment defects and HSF dysregulation.
Globally, the cross-analysis of data from both NDD models will allow us to identify common candidate pathways and gene targets. These candidates will be validated in hCOs, and eventually in iNPCs, through RT-QPCR, flow cytometry and/or western blotting, and we will characterize the associated phenotypic traits through immunofluorescence staining.

=> The student will join the various aspect of the project:
1- conduct CRISPR/Cas gene editing of HSF2. 2- Generate the isogenic 2D- (iNPC) and 3D- (hCOs) models for phenotyping and SC-omics. 3- Analyze phenotypic abnormalities and SC-omics data. 4- Validate the molecular and phenotypic target genes and pathways identified.

Collectively, these comparative analyses will enable us to isolate the HSF component of the RSTS/AGS phenotype, demonstrating its potential to drive pathological traits. This study will provide cellular and molecular tools for a large-scale drug screening campaign for NDDs (Ksilink company, ANR PRCE) that share stress-related characteristics.

References
1. Boyle CA, 2011 PMID: 21606152. 2. Drew 2015 PD PMID: 25703036 3. Zerbo O, 2014 PMID: 24951035 4. Smaga I, 2015 PMID: 25933971. 5. Schang AL, 2022 PMID: 36513635 6. Kawada K, 2018 PMID: 28770435 7. Liu X 2022 PMID: 35299821 8. Komada M, 2022 PMID: 35646933; 9-Hassan H 2021 PMID: 34940131 10. Dayalan N 2015 PMID: 25465722. 11. Himanen SV, 2022 PMID: 35687139. 12. Koike N, 2018 PMID: 29423676. 13. Löchli K, 2023 PMID: 36449150 14. Akerfelt M, 2010 PMID: 20628411 15. Duchateau A, 2020 PMID: 32147500. 16. Kallio M, 2002 PMID: 12032072 17. Chang Y 2006 PMID: 16600913 18. El Fatimy R 2014 PMID: 25027850 19. de Thonel A, 2022 PMID: 36385105 20. Aguilera C, 2021 PMID: 34653234. 21. Spena S, 2015 PMID: 27617129 22. Williams CA, 2010 PMID: 20445456. 23.Hakanen J, 2019 PMID: 31213986

Bi-directional, brain interfaced upper-limb prostheses may transform the autonomy of patients with severe neurological deficits. However, it is unclear the feedback side of these interfaces can be designed to carry touch and proprioceptive information. This feedback is essential for sensing, but also prosthesis embodiment and safety.
In our team, we study upper limb neuro-prosthetics in the mouse model, and we have built a number of innovative tools, that the PhD candidate will use to design spatiotemporally patterned optogenetic feedback targeted to the primary somatosensory cortex. It will, for the first time, provide simultaneously touch and proprioceptive feedback in the context of a brain-machine interfaced upper-limb prosthesis.
The PhD candidate will explore the physiological constraints on such feedback designs by probing their impact of the learning ability of mice that will be trained to solve behavioral tasks through the prosthesis.

Upper limb prosthesis that are directly controlled by the brain motor areas, including the primary motor cortex, are emerging as a therapeutic option for patients with major autonomy deficits, and in particular tetraplegic patients [1,2].

Recently, researchers have extended such prostheses to provide direct touch feedback by microstimulating the primary somatosensory cortex [3]. Indeed, touch feedback is critical to the safe use of the prosthesis. Further, tactile exploration is essential to most limb behaviors, including object manipulations. Beyond touch, other somatosensory inputs from the limbs are key, and in particular proprioception, which is the representation of limb posture. Notably, tactile and proprioceptive feedback are likely to have a major impact on the embodiment of the prosthesis [4].

However so far, feedback from the prosthesis to the cortex has been limited to 1 or 2 parallel channels of feedback, and there has been no successful representation of proprioceptive information.

These shortfalls are a directly reflection of fundamental limitations to our understanding of the integration of artificial stimulations by the cortex. In the context of this project, the PhD candidate will explore and optimize the integration of such artificial inputs into the somatosensory cortex in the mouse model, in order to provide both touch and proprioceptive information in the context of the control of an upper limb prosthesis.

To this aim, the project will take advantage of the unique mouse neuroprosthetics platform that we have developed in the laboratory. It includes:
- a unique bidirectional, cortical brain-machine interface [5] that already enabled us to study the ability of cortical tissue to integrate complex, distributed optogenetic inputs to the primary somatosensory cortex and shown that the spatiotemporal continuity and contiguity of cortical inputs is essential to their integration [6,7]
- a recently developed, one of a kind upper limb prosthesis for the mouse model, with 4 degrees of freedom as well as touch sensors, which is connected to th brain-machine interface.
- The ability to read and write the activity of the cortex at the mesoscopic scale — which is critical to design cortical feedback based on the activity it triggers.
- Experimental tools to probe the embodiment of the prosthesis, by evaluating the reaction of the mice to a menace to the prosthesis.

By combining these tools, the PhD candidate will be in a unique position to identify the physiological constraints that rest on distributed, direct cortical stimulation feedback patterns that feed back limb-relevant information. She/he will design the first distributed feedback that provides both proprioceptive and touch cortical feedback, and will test its impact on the behavioral performance of mice, as they control the upper limb prosthesis in order to solve simple rewarded tasks.

This PhD project is highly interdisciplinary and calls for an interest both in neuroscience and in engineering. I will take place in the context of a rich collaboration between our team and the robotics team of Maria Makarov at engineering school CentraleSupélec, located in the same Paris-Saclay campus, in the outskirts of Paris.

[1] Benabid, A. L., Costecalde, T., Eliseyev, A., Charvet, G., Verney, A., Karakas, S., ... & Chabardes, S. (2019). An exoskeleton controlled by an epidural wireless brain–machine interface in a tetraplegic patient: a proof-of-concept demonstration. The Lancet Neurology, 18(12), 1112-1122.

[2] Collinger, J. L., Wodlinger, B., Downey, J. E., Wang, W., Tyler-Kabara, E. C., Weber, D. J., ... & Schwartz, A. B. (2013). High-performance neuroprosthetic control by an individual with tetraplegia. The Lancet, 381(9866), 557-564.

[3] Flesher, S. N., Downey, J. E., Weiss, J. M., Hughes, C. L., Herrera, A. J., Tyler-Kabara, E. C., ... & Gaunt, R. A. (2021). A brain-computer interface that evokes tactile sensations improves robotic arm control. Science, 372(6544), 831-836. [4] Btovinick

[5] Abbasi, A., Goueytes, D., Shulz, D. E., Ego-Stengel, V., & Estebanez, L. (2018). A fast intracortical brain–machine interface with patterned optogenetic feedback. Journal of neural engineering, 15(4), 046011.

[6] Lassagne, H., Goueytes, D., Shulz, D. E., Estebanez, L., & Ego-Stengel, V. (2022). Continuity within the somatosensory cortical map facilitates learning. Cell Reports, 39(1).

[7] Abbasi, A., Lassagne, H., Estebanez, L., Goueytes, D., Shulz, D. E., & Ego-Stengel, V. (2023). Brain-machine interface learning is facilitated by specific patterning of distributed cortical feedback. Science Advances, 9(38), eadh1328.

Orientation in space is crucial for our survival, for navigating an environment and for finding food and shelter. The head direction (HD) system functions as the brain's compass system. Vestibular based HD signals combine with visual signals to permit orientation coding with respect to external landmarks. But how is the visual scene informative for our inner compass? Here we will investigate how the brain integrates and represents visual stimuli, as a function of the speed of movement of the visual surround. We will focus on the visual cortex and on the pre- and postsubicular cortex, where neurons coding for HD are present.

We ask:
1. Does slow optic flow influence the orientation that is represented by head direction cells, in mice?
2. Does 'naturalistic' optic flow (similar to what the animal sees when it actually turns its head) induce changes in HD signals?
3. Is there a role of stationary phases before and after the movement?

Background.
This project arises out of an interaction between the research of the two directors. The research group of Desdemona Fricker specializes in the neurophysiology of spatial orientation, using the mouse brain as a model. The presubiculum is located in the brain’s temporal lobe, just below visual cortex. It contains head direction cells that code the animal’s orientation; these cells are internally organized and constrained by attractor dynamics that are based on excitatory-inhibitory interactions (Simonnet et al, 2017). Computational models postulate that the population of HD cells reacts coherently to changes in landmarks (Simonnet & Fricker, 2018). It is however unclear how optic flow input onto these neurons may improve spatial orientation or if HD cells may escape attractor dynamics. Mark Wexler and his collaborators have been studying how motion perception at ‘naturalistic speeds’ (typical speeds of stationary objects on the retina during saccades) depends on the details of the motion trajectory, and in particular on whether the moving object is seen stationary before and/or after the motion. In humans, even very brief (20 ms) stationary phases have been found to have a profound effect on perceived motion: trajectories with stationary phases are perceived as having smaller amplitude, slower motion, are more clear and less smeared (Wexler & Cavanagh, 2019). These results imply that visual stabilization during eye movement may rely much more on purely visual, rather than visuo-motor, mechanisms than has been supposed in the past. Other recent results indicate that motion of the visual background can have profound effects on visual localization in human subjects (Özkan, Anstis, Hart, Wexler & Cavanagh, 2021). It would be very interesting to check if similar results apply to how background optic flow affects spatial orientation in mice, as measured through HD cells.

Methods.
Desdemona Fricker’s group has built a setup for head-fixed in vivo recordings in mice, where Neuropixels probes can be inserted in the mouse brain via a 4-axis manipulator to carry out acute recordings of populations of neurons (Figure 1; Jun et al., 2011). The probe trajectory is calculated to record simultaneously from both the visual cortex and the pre- and postsubiculum. We use electrophysiological data acquisition software (SpikeGLX), and a data analysis pipeline with kilosort3 for spike sorting and visualization in phy for semi-automatic curation and for merging and splitting of clusters. We can combine vestibular rotational stimuli with visual projections at high frame rate (1440 Hz) on a dome surrounding the animal, to deliver precisely matched or mismatched sensory stimuli.

Scientific program.
We will examine the encoding of visual signals in populations of single neurons in the visual and in the presubicular cortex.
TASK 1. Pseudorandom back-and-forth movements of the motorized rotating stage will be used. We will develop algorithms to create motion trajectories that sample stochastically but uniformly the space of orientations, angular velocities, or a combined two-dimensional space of orientations and velocities. We will record extracellular spiking activity, and quantify the angular tuning of individual HD cells by calculating the Rayleigh vector (R). Initial experiments indicate that more than half of well-identified units are angularly tuned in these conditions in the presubiculum (Figure 2). Further, we will use multivariate analysis and machine learning techniques to decode orientation and or velocity from populations of neurons (Figure 3). We will compare neurons in presubiculum and visual cortex and ask whether their spiking patterns are dominated by HD and/or visual input. We will train decoders on physical motion in the dark, and test them on the visual simulation of the same motion, in order to determine if self-motion from vestibular input and from optic flow are represented in the same networks.
TASK 2. We wish to use trajectories of head movements that have been recorded in freely behaving animals in Berlin (Simonnet & Brecht, 2019). These trajectories can be replayed as a command to our rotating stage, in order to mimic natural head movements that contain high angular velocities. Alternatively, we will replay these trajectories as naturalistic optic flow, by equivalent rotation of the visual surround. We will examine if the firing of presubicular HD cells or visual cortical neurons reflects such fast changes of the visual scene. We will carefully look if the typically coherent population coding of HD in the presubiculum remains stable, and if decoding of the population angularly tuned cells will reflect the fast movements of visual stimuli.
TASK 3. We will check whether stationary phases of the visual stimulus before and after the trajectory modulate the reaction of HD cells to optic flow, as they do in human perception. Because in the naturalistic situation head movements are preceded and followed by phases in which the head is nearly still, we predict that optic flow with stationary phases will lead to the closest approximation of actual movement. Optic flow without stationary phases should then lead to abnormally large reactions in HD cells, if the effect of stationary phases is the same in mice as in humans.

References:
- Simonnet J, Nassar M, Stella F, Cohen I, Mathon B, Boccara CN, Miles R, Fricker D. (2017) Activity dependent feedback inhibition may maintain head direction signals in mouse presubiculum. Nat Commun. 20;8:16032.
- Simonnet J, Fricker D. (2018). Cellular components and circuitry of the presubiculum and its functional role in the head direction system. Cell Tissue Res. 373:541–556.
- Özkan, M., Anstis, S., ’t Hart, B. M., Wexler, M., & Cavanagh, P. (2021). Paradoxical stabilization of relative position in moving frames. PNAS, 118(25), e2102167118.
- Wexler, M., & Cavanagh, P. (2019). Fast motion drags shape. Journal of Vision, 19(10), 288c.
- Jun JJ, Steinmetz NA, et al. (2017). Fully integrated silicon probes for high-density recording of neural activity. Nature. 551:232–236.
- Simonnet J, Brecht M. (2019). Burst firing and spatial coding in subicular principal cells. Journal of Neuroscience. 1656–18.

In a continuous stream of auditory signals, predicting when certain events will occur (a sound indicating danger or a phoneme important for understanding a sentence) is essential for adaptive behavior. Temporal predictions help us orient attention in time, which is crucial to protect a cognitive system with limited capacity from overload. Here, we will investigate how temporal predictions benefit hearing, using a psychophysical approach in healthy human adults. Furthermore, we will test the interaction of temporal and sensory predictions in audition by investigating their neural signatures in specialized auditory regions of the brain. To achieve this goal, we will take advantage of the excellent spatial and temporal resolution of magnetoencephalography. This project addresses fundamental research questions in human cognition at the interface of timing and hearing, with the aim to provide a better understanding of the mechanisms supporting the temporal efficiency of auditory processing.

Temporal regularities are naturally present in auditory signals, like the sound of a running stream, a chirping bird, or a speaking human, and the human auditory system has evolved to efficiently integrate signals over a range of co-existing temporal scales (Giraud and Poeppel 2012). Extracting temporal regularities from auditory inputs allows the listener to form temporal predictions, that is to predict when a relevant event will occur, allowing to focus capacity-limited resources on the predicted moments in time and interact efficiently with dynamic sensory environments.
Contrary to sensory predictions ('what'), temporal predictions ('when') cannot draw on a dedicated sensory system, but interact with the respective sensory analysis to enhance perceptual processing at the most relevant moments (Arnal and Giraud 2012; Feldman and Friston 2010): temporal predictions thus guide the orienting of attention in time (Nobre and van Ede 2018). Given the particular relevance of timing for audition, a better understanding of the mechanisms supporting the temporal efficiency of auditory processing is critical. Numerous studies have shown that temporal predictions improve auditory perception, for instance gap detection or pitch discrimination. Yet, there are important controversies across studies, and it is unclear how the orienting of attention in time interacts with sensory processes along the auditory hierarchy. To resolve these heterogeneities, we will map out the functional benefits of temporal prediction for auditory perception in an extensive psychophysical approach, and test in healthy human listeners which aspects of auditory perception (e.g. detection, discrimination) benefit from temporal predictability. Temporal predictions will be induced implicitly by variation of the foreperiod distribution (Herbst and Obleser 2019), i.e. the interval separating a cue and a target stimulus. Bayesian observer models (developed by the PI) will be used to formalize and quantify temporal predictions derived from the previously encountered temporal statistics of the auditory inputs.
Second, we will assess how temporal and sensory predictions of auditory events interact on the behavioural and neural level, taking advantage of the high spatial and temporal resolution of magnetoencephalography (MEG). Animal studies have shown that temporal attention specifically enhances the precision of frequency-tuned neural populations in early auditory areas (Jaramillo and Zador 2011), but this remains to be verified in humans. To arbitrate between general versus specific effects of temporal predictions in audition, we will study the neural dynamics underlying temporal and spectral orienting of attention, derived from temporal and sensory predictions. Temporal orienting has been associated with oscillatory entrainment in auditory and motor areas, but also modulates transient target-evoked activity.
In sum, these experiments address the fundamental research question of how timing interacts with hearing. The knowledge gained in this project can eventually contribute to the development of training strategies for compensation and rehabilitation of auditory deficits, and the engineering of neuro-prosthetic devices capitalizing more on the temporal structure of auditory inputs.

Expected student profile: The ideal candidate has a background in Background in Psychology / Cognitive Science / Neuroscience, or a related field. He/ she should bring a strong interest in the outlined scientific questions, and readiness to tackle technical challenges and find solutions to novel problems. Prior experience with programming (e.g. Python, Matlab or R) and basic statistical knowledge is required, as well as oral and written communication skills in English. (French language skills are not mandatory, but courses are available at the University). Prior experience with EEG/MEG or auditory psychophysics is a plus.

Team: The candidate will carry out their work in the Cognition and Brain Dynamics Team, embedded in the UNICOG (INSERM U992) research unit at Neurospin, situated on the CEA-Saclay site. Neurospin is an internationally renowned Neuroimaging institute with excellent imaging facilities (fMRI, MEG, EEG, psychophysics), research support staff, and in-house pipelines for neuroimaging analyses. The Cognition and Brain Dynamics Team fosters diversity, curiosity, enthusiasm, motivation, and learning in an international and collegial atmosphere.


References:
Arnal, Luc H., and Anne-Lise Giraud. 2012. “Cortical Oscillations and Sensory Predictions.” Trends in Cognitive Sciences 16 (7): 390–98. https://doi.org/10.1016/j.tics.2012.05.003.

Feldman, Harriet, and Karl J Friston. 2010. “Attention, Uncertainty, and Free-Energy.” Frontiers in Human Neuroscience 4 (December). https://doi.org/10.3389/fnhum.2010.00215.

Giraud, Anne-Lise, and David Poeppel. 2012. “Cortical Oscillations and Speech Processing: Emerging Computational Principles and Operations.” Nature Neuroscience 15 (4): 511–17. https://doi.org/10.1038/nn.3063.

Herbst, Sophie K., and Jonas Obleser. 2019. “Implicit Temporal Predictability Enhances Pitch Discrimination Sensitivity and Biases the Phase of Delta Oscillations in Auditory Cortex.” NeuroImage 203 (September): 116198. https://doi.org/10.1016/j.neuroimage.2019.116198.

Jaramillo, Santiago, and Anthony M. Zador. 2011. “The Auditory Cortex Mediates the Perceptual Effects of Acoustic Temporal Expectation.” Nature Neuroscience 14 (2): 246–51. https://doi.org/10.1038/nn.2688.

Nobre, Anna C., and Freek van Ede. 2018. “Anticipated Moments: Temporal Structure in Attention.” Nature Reviews. Neuroscience 19 (1): 34–48. https://doi.org/10.1038/nrn.2017.141.

The cerebellum plays a major role in sensory motor adaptation and non-motor tasks. Initially, the cerebellar cortex was considered as a homogeneous structure with a “crystal-like” organization at the level of the cytoarchitecture and neuronal connectivity. However, it is now apparent that the cerebellar cortex is organized in modules found at the functional and molecular levels and that correspond to a topographic organization of cerebellar functions. A high heterogeneity in the neuronal populations forming this network is starting to be deciphered. Granule cells are the most numerus neurons in the brain, however their diversity has been very poorly studied. Our goal is to decipher how this diversity is generated and its relevance for the connectivity and function of the cerebellum.

The cerebellum plays a major role in sensory motor adaptation and non-motor tasks. Initially, the cerebellar cortex was considered as a homogeneous structure with a “crystal-like” organization at the level of the cytoarchitecture and neuronal connectivity. However, it is now apparent that the cerebellar cortex is organized in modules found at the functional and molecular levels and that correspond to a topographic organization of cerebellar functions. A high heterogeneity in the neuronal populations forming this network is starting to be deciphered. But, while neurochemical and functional heterogeneity in Purkinje cells (PC) have been well described and its relationship to the diversity of functional properties for PCs is better understood, diversity of other neuronal populations in the cerebellar cortex is very poorly studied. Some markers have been found to be expressed heterogeneously in granule cells (GC), especially along the antero-posterior axis and single cell transcriptomic analysis has defined at least five types of GCs. The goal of this project is therefore to determine whether the molecular heterogeneity of GCs underlie a functional and computational heterogeneity in the cerebellar cortex.
We will assess the spatial organization of two different GC lineages that are organized in two different antero-posterior gradients. Thanks to two genetically modified mouse lines already available and characterized as labeling these two GC lineages, we will describe their relationships to specific cerebellar modules defined by Purkinje cells markers, by combining anatomical and imaging techniques. We will also compare how they relate to the four subpopulations of GCs identified previously by analysis of single-cell transcriptomics data. We will then determine the specific surfaceome for anterior versus posterior GCs using proteomic techniques and identify relevant markers for their specific connectivity and functional properties. Several families of surface proteins have been identified as candidates. Functional analysis of some of those markers will be performed using the Crispr/Cas9 technology to assess the relevance of GC molecular heterogeneity for their network integration and their functional diversity.
Second, using an intersectional approach, we will test how two parameters, GC genetic lineage and GC date of birth, interact and influence local molecular heterogeneity. For this we will combine our two genetically modified mouse lines with electroporation of Cre dependent reporter constructs at different postnatal timepoints. Using sophisticated imaging techniques, we will describe how these two parameters interact to influence GC network integration and the diversity of their synaptic properties.
This project will involve collaboration with the team of Dr. Isope (INCI, Strasbourg) and has already received funding from the Agence Nationale de la Recherche.
Altogether our project will combine various technologies to decipher fundamental mechanisms regulating neuronal diversity, network integration and function of the most numerous neuronal population of the brain. The results will impact our understanding of brain development and function. Our results will also be of interest for the treatment of brain diseases since the cerebellum is also involved in several types of brain disorders such as autism spectrum disorders.

Lower-grade gliomas are malignant primary brain tumors of the adult that are characterized by somatic IDH1 mutation (IDH1R132H), the earliest event in lower-grade glioma development. Although the genetics of these tumors is well established, the cellular and molecular mechanisms of tumor initiation and development are less understood. No study has reported the effects of physiological IDH1 mutation in human oligodendrocyte precursor cells (OPC), the potential cells of origin for lower-grade gliomas. The goal of the project is to understand how IDH1R132H alters human OPC development and differentiation, and how this modulates their interaction with neurons. To this aim, we will generate and characterize a novel model of IDH1R132H oligocortical organoids that will allow to probe the effects of IDH1 in the closest system recapitulating features of the human brain. This project will involved a tight collaboration with organoid experts and technological platforms of the institute.

Context
Diffuse gliomas are the most malignant primary brain tumors in the adult. This family of tumors regroup heterogenous diseases that differ in histology, molecular alterations and prognoses (Louis et al., 2021). The most common subtypes include the highly aggressive glioblastomas, as well as astrocytomas and oligodendrogliomas, collectively known as lower-grade gliomas. Lower-grade gliomas are notably characterized by heterozygous somatic mutation in the isocitrate dehydrogenase gene IDH1, with IDH1R312H being the most frequent mutation. Somatic IDH1 mutation is an early, clonal event, that can even be found in glial cells of the white matter of the prefrontal cortex in nondiseased human brain (Ganz 2022). Mutated IDH1 enzyme, through the production of 2-hydroxyglutarate (2-HG) metabolite, induces global changes in DNA and histone methylation, conferring a proliferative advantage in human astrocytes (Koivunen et al., 2012; Turcan et al., 2012).
Although oligodendrogliomas and astrocytomas are well characterized at the genomic level, the cellular and molecular mechanisms of tumor initiation and development are still poorly defined, especially for oligodendrogliomas. The presumed cells of origin for lower-grade gliomas are oligodendrocyte precursor cells (OPC), which are glial cells derived from neural stem/progenitor cells and serving as precursors for oligodendrocytes, the myelin-forming cells of the central nervous system. The effect of IDH1 mutation on OPC is not known. In our group, we have generated a genetically-engineered mouse model in which we expressed IDH1R132H at endogenous levels and specifically in OPC. We observed an induction of white matter OPC proliferation and expansion at late time points, as well as activation of microglia, the resident immune cells of the brain (Joppé, Pottier et al., unpublished). We are currently fully characterizing this model to identify the cellular effects and molecular pathways affected by IDH1 mutation in an in vivo context.

Project objectives
IDH1 mutation represents the earliest mutation in lower-grade gliomas. No study has reported the effects of physiological IDH1 mutation in human OPC, the potential cells of origin for lower-grade gliomas. The goal of the project is to understand how IDH1R132H alters human OPC development and differentiation, and how this modulates their interaction with neurons. To this aim, we will generate and characterize a novel model of IDH1R132H oligocortical organoids that will allow to probe the effects of IDH1 in the closest system recapitulating features of the human brain.

Description of the research program
1/Generation of IDH1R132H-iPSC from lower-grade glioma patients. Organoids are self-organizing 3D cellular models that are commonly derived from human iPSCs. They contain multiple cellular lineages and preserve complex cell-cell interactions. Our collaborator Emeline Tabouret (AP-HM and INP Marseille) has generated a biobank of peripheral blood mononuclear cells from 20 patients with high grade diffuse gliomas, including 7 patients with lower-grade gliomas (GLIOMANOID, NCT03971812 clinical trial). She has reprogrammed these cells into iPSCs. We will use CRISPR/Cas9 genome editing to knock-in IDH1R132H heterozygous mutation in iPSC derived from lower-grade gliomas patients (n=4 astrocytomas, n=3 oligodendrogliomas). We will first design CRISPR tools, proceed to genome editing and analyze the resulting clones by sequencing and verification of off-target sites and quality control. This aim will be done in close collaboration with the ICV-iPS platform in our institute (manager: Stéphanie Bigou). Control iPSC will be unmodified cells.

2/Generation and characterization of IDH1R312H-iPS derived oligocortical organoids. Our goal is to generate cerebral organoids from IDH-mutant iPSC. To enrich these organoids in oligodendroglial cells, we will implement protocols allowing efficient production of OPC and mature oligodendrocytes from iPSC (Madhavan et al., 2018; Marton et al., 2019). The resulting oligocortical IDH1-mutant and IDH1-wild type oligocortico organoids will be characterized and compared, using a combination of immunofluorescence, electronic microscopy and transcriptomic approaches, for: 1) the production of the 2-HG oncometabolite; 2) proliferation, survival and differentiation status oligodendroglial cells; 3) changes in the interactions of mutated cells with the neurons present in the organoids; 4) genomic changes for genes and loci recurrently altered in oligodendrogliomas and alterations in DNA and histone methylation profiles. All the data will be compared with our transcriptomic, genomic and proteomic data on patients oligodendrogliomas.
For this aim we will beneficiate from the expertise of Dr Philippe Ravassard and his team, who are experts in cerebral organoid engineering and characterization in our institute, as well as from the histology, electrophysiology and imaging platforms of the institute.

Relevance and originality. The project will for the first time determine the function of IDH1mutation in the most likely cell of origin for adult diffuse lower-grade gliomas. We will use iPSCs generated from patients carrying lower-grade gliomas. This is important as it is known that inherited risk loci have been identified for IDH-mutant gliomas (Yanchus et al., 2022): for example, a single-nucleotide polymorphism at 8q24 confers a six-fold greater risk of IDH-mutant glioma and was shown to accelerate tumor development in an Idh1R132H-driven lower-grade glioma mouse model. It is therefore important to consider the patient’s genomic background when modeling the earliest steps of tumorigenesis. In perspective, this model will be complemented with other oligodendrogliomas alterations (1p19q loss, pTERT mutation, CIC mutation) that can be xenografted in mice to assess tumorigenicity.

References
Koivunen, P., Lee, S., Duncan, C. G., Lopez, G., Lu, G., Ramkissoon, S., Losman, J. A., Joensuu, P., Bergmann, U., Gross, S., et al. (2012). Transformation by the (R)-enantiomer of 2-hydroxyglutarate linked to EGLN activation. Nature 483, 484–8.
Louis, D. N., Perry, A., Wesseling, P., Brat, D. J., Cree, I. A., Figarella-Branger, D., Hawkins, C., Ng, H. K., Pfister, S. M., Reifenberger, G., et al. (2021). The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol 23, 1231–1251.
Madhavan, M., Nevin, Z. S., Shick, H. E., Garrison, E., Clarkson-Paredes, C., Karl, M., Clayton, B. L. L., Factor, D. C., Allan, K. C., Barbar, L., et al. (2018). Induction of myelinating oligodendrocytes in human cortical spheroids. Nat Methods 15, 700–706.
Marton, R. M., Miura, Y., Sloan, S. A., Li, Q., Revah, O., Levy, R. J., Huguenard, J. R. and Pa?ca, S. P. (2019). Differentiation and maturation of oligodendrocytes in human three-dimensional neural cultures. Nat Neurosci 22, 484–491.
Turcan, S., Rohle, D., Goenka, A., Walsh, L. A., Fang, F., Yilmaz, E., Campos, C., Fabius, A. W. M., Lu, C., Ward, P. S., et al. (2012). IDH1 mutation is sufficient to establish the glioma hypermethylator phenotype. Nature 483, 479–83.
Yanchus, C., Drucker, K. L., Kollmeyer, T. M., Tsai, R., Winick-Ng, W., Liang, M., Malik, A., Pawling, J., De Lorenzo, S. B., Ali, A., et al. (2022). A noncoding single-nucleotide polymorphism at 8q24 drives IDH1-mutant glioma formation. Science 378, 68–78.

A major open question in neuroscience is how perception is generated in the brain. Perception allows us to make meaning of our sensations, and is drastically effected in psychiatric disorders like schizophrenia. Auditory disturbances are prevailing, suggesting that the auditory system could be particularly affected. An influencing hypothesis states that perceptual disturbances are produced when the brain overweights priors, learnt “top-down” information that is necessary to evaluate incoming sensory signals. Yet, we have little knowledge of where and how top-down information is processed in sensory systems. In this project, we will combine in vivo calcium imaging with optogenetics in mice to elucidate how top-down information is causally encoded in auditory cortex during learning and memory to mediate physiological and pathological perception. Our results will illuminate our understanding of how sensations are transformed into perceptions in thalamo-cortical circuits.

A central goal of neuroscience is to understand the mechanisms that enable sensory perception, an active process that depends on our previous experiences and behavioural demands. In this framework, information that is captured by our sensory organs, that flows to sensory cortex in a “bottom-up” stream, needs to integrate with internal “top-down” information, to form an accurate representation of the world 1.

This fundamental brain function is affected in major neuropsychiatric disorders. While a main open question is how accurate and disturbed perception are determined, a central hypothesis states that it depends on the balance between “bottom-up” and “top-down” information. Directly addressing this question has remained challenging, largely because we have little knowledge of where and how “top-down” information is encoded in the brain. Therefore, this project aims to elucidate the function of “top-down” inputs to sensory cortex.

As opposed to the first-order sensory thalamus that conveys bottom-up information to sensory cortex, recent evidence from others and us indicate that the little explored higher-order (HO) thalamo-cortical pathways are relevant for internal top-down information across sensory modalities 1. We have recently found that the HO auditory thalamus conveys information related to aversive memories to secondary auditory cortex 2.

Conversely, a dysfunction of HO sensory thalamic nuclei has been related to cognitive deficits such as inattention and schizophrenia. Schizophrenia manifests in most cases with auditory perceptual disturbances, suggesting that alterations in the auditory HO system may be a cause.

In this research project, we will test the hypothesis that the HO-auditory thalamus sends to auditory cortex top-down signals that are critical for learning a sensory percept, and affected in pathological conditions related to schizophrenia.

To test our hypothesis, our first aim will be to define the diversity of signals that the HO-auditory thalamus transmits to auditory cortex during a discriminative perceptual decision making task. For this, we will use our lab’s development of in vivo 2-photon calcium imaging of long-range axons to investigate what top-down information is conveyed by the HO-auditory thalamus to auditory cortex during learning and memory in this task.

Our second aim is to test whether and how these top-down signals affect discriminative associative learning and perceptual discrimination. To do this, we will inhibit the thalamic synaptic activity in cortex with optogenetics at specific instances during the task and quantify how behaviour is affected.

Our third aim is to determine how these top-down signals affect encoding in auditory cortex. For this we will combine in vivo optogenetic manipulations with calcium imaging recordings of cortical neurons.

We will compare physiological and pathological conditions by employing mouse models of neuropsychiatric disorders.

Together, this research will uncover how top-down information is encoded during complex behaviours to generate perception. We will in turn determine how HO thalamic inputs to cortex affect cortical function and behaviour, as well as identify features affected in pathological conditions associated to neuropsychiatric disorders.

1. Pardi, M. B., Schroeder, A. & Letzkus, J. J. Probing top-down information in neocortical layer 1. Trends Neurosci. (2022). doi:https://doi.org/10.1016/j.tins.2022.11.001

2. Pardi, M. B. et al. A thalamocortical top-down circuit for associative memory. Science. 370, 844–848 (2020).

Individual brains vary in both anatomy and functional organization, even within a given species. Inter-individual variability is a major impediment when trying to draw generalizable conclusions from neuroimaging data collected on groups of subjects. Current co-registration procedures rely on limited data, and thus lead to very coarse inter-subject alignments.
In this PhD project, we propose to develop a novel method for inter-subject alignment based on Optimal Transport, using Gromov Wasserstein type of loss. Such a method aims at aligning cortical surfaces based on the similarity of their functional signatures in response to a variety of stimulation settings, while penalizing large deformations of individual topographic organization.
We will perform technical developments to scale up such transport solvers to large scale (> 10^5 nodes), as well as various validation experiments with both intre- and inter-species cortical alignment procedures.

Objective
The availability of millimeter or sub-millimeter anatomical or functional brain images has opened new horizons
to neuroscience, namely that of mapping cognition in the human brain and detecting markers of diseases. Yet
this endeavor has stumbled on the roadblock of inter-individual variability: while the overall organization of
the human brain is largely invariant, two different brains (even from monozygotic twins [1]) may differ at the
scale of centimeters in shape, folding pattern, and functional responses. The problem is further complicated
by the fact that functional images are noisy, due to imaging limitations and behavioral differences across
individuals that cannot be easily overcome. The status quo of the field is thus to rely on anatomy-based
inter-individual alignment that approximately matches the outline of the brain [2, 3] as well as its large-scale
cortical folding patterns [4]. Existing algorithms thus coarsely match anatomical features with diffeomorphic
transformations, by warping individual data to a simplified template brain. Such methods lose much of the
original individual detail and blur the functional information that can be measured in brain regions.

In order to improve upon the current situation, a number of challenges have to be addressed: (i) There
exists no template brain with functional information, which by construction renders any cortical matching
method blind to function. This is unfortunate, since functional information is arguably the most accessible
marker to identify cortical regions and their boundaries [5]. (ii) When comparing two brains – coming from
individuals or from a template – it is unclear what regularity should be imposed on the matching [6]. While
it is traditional in medical imaging to impose diffeomorphicity [2], such a constrain does not match the
frequent observation that brain regions vary across individuals in their fine-grained functional organization
[5, 7]. (iii) Beyond the problem of aligning human brains, it is an even greater challenge to systematically
compare functional brain organization in two different species, such as humans and macaques [8, 9]. Such
inter-species comparisons introduce a more extreme form of variability in the correspondence model.

Context: different flavors of brain-feature alignment
Several attempts have been made to constrain the brain alignment process by using functional information.
The first one consists in introducing functional maps into the diffeomorphic framework and search for a
smooth transformation that matches functional information [11, 12, 13], the most popular framework being
arguably Multimodal Surface Matching (MSM) [13, 5].
A second family of less constrained functional alignment approaches have been proposed, based on heuristics,
by matching information in small, possibly overlapping, cortical patches [14, 15, 16]. This popular framework
has been called hyperalignment [14, 17], or shared response models [18]. Yet these approaches lack a principled
framework and cannot be considered to solve the matching problem at scale. Neither do they allow to estimate
a group-level template properly [19].
An alternative functional alignment framework has followed another path [20], considering functional signal
as a three-dimensional distribution, and minimizing the transport cost. However, this framework imposes
unnatural constraints of non-negativity of the signal and only works for one-dimensional contrasts, so that
it cannot be used to learn multi-dimensional anatomo-functional structures. An important limitation of the
latter two families of methods is that they operate on a fixed spatial context (mesh or voxel grid), and thus
cannot be used on heterogeneous meshes such as between two individual human anatomies or, worse, between
a monkey brain and a human brain.

Proposed work
Following [21], we use the Wasserstein distance between source and target functional signals – consisting
of contrast maps acquired with fMRI – to compute brain alignments. We have already contributed two
notable extensions of this framework [22]: (i) a Gromov-Wasserstein (GW) term to preserve global anatomical
structure – this term introduces an anatomical penalization against improbably distant anatomical matches,
yet without imposing diffeomorphic regularity – as well as (ii) an unbalanced correspondence that allows
mappings from one brain to another to be incomplete, for instance because some functional areas are larger
in some individuals than in others, or may simply be absent. We have shown that this approach successfully
addresses the challenging case of different cortical meshes, and that derived brain activity templates are
sharper than those obtained with standard anatomical alignment approaches.
With the present proposal, we would like to further develop this approach, and to further validates it with a
so-called decoding framework.
Technical developments An important need is to scale up the Gromov-Wassertsein plan to very high
resolution, leading to very costly implementations. We propose to consider coarse-to-fine strategies as well
as quantized approaches to allow the solver to scale with reasonable time and memory requirements.
Decoding-based validation the classification of brain states or decoding, when performed across individ-
uals, provides the most credible metric to measure the gain brought by inter-individual alignment. Indeed,
across-individuals generalization of brain state classification is generally hampered by inter-individual vari-
ability, but is likely to improve after alignment. We want to assess such gain using high-resolution data and
a large array of decoding problems [16].

Implementation
The work will be based on the work described in [22] and available in the fugw package. It will leverage state-of-the art
optimal transport methods. Imaging data will be used from the publicly available Individual Brain Charting
and Human Connectome Project datasets, that are well mastered by the supervisors of this proposal.
Required skills:
The successful candidate will be interested in applications of machine learning and in the understanding of hu-
man cognition. Note that the work will take place in a multi-disciplinary environment (physics, neuroscience,
computer science, modeling, psychology).
Prior experience on deep model is a major asset, as it makes it easier for the candidate to understand the
concepts and tools involved. Knowledge of scientific computing in Python (Numpy, Scipy) is required. All
the work will be done in Python based on standard machine learning libraries and the Nilearn library for
neuroimaging aspects. The candidate will benefit from the numerous development of the Parietal team
for computational facilities and expertise in the various domains involved (machine learning, optimization,
statistics, neuroscience, psychology).

For details and references, please see https://team.inria.fr/mind/files/2022/11/fugw.pdf

The signal generated by the senses is very high-dimensional and strongly fluctuating in time. Yet, we perceive the world in terms of stable objects. That must be because the cortex builds invariant representations of the sensory signal. But how does the cortex build these representations? According to a popular idea, they emerge from hierarchical, feed-forward processing. Anatomical and physiological observations, however, severely challenge this notion. In this research project, we investigate an alternative hypothesis: Invariant representations emerge as a result of attractor dynamics due to recurrent, feed-back processing. The proposed research is interdisciplinary, drawing on concepts and methods used in the physics of disordered systems, dynamical system theory, machine learning, neuronal and synaptic physiology. The results obtained will elucidate the network mechanisms underlying the emergence of stimulus-evoked responses in the cortex and their representational content.

The cortex is informed about the world by the signal originating from peripheral sensory receptors. This signal is high-dimensional and strongly changing from moment to moment due to extrinsic (e.g., changes in the world) and intrinsic factors (e.g., moving the eyes). Yet, we perceive the world in terms of stable objects that can be quickly recognized in practically any situation. How do we achieve this feat? That must be because the cortex builds invariant representations of the sensory signal.

Simply stated, building an invariant representation consists in disregarding the irrelevant information carried by a signal, where “relevant” and “irrelevant” is defined by the task at hand. For instance, we can recognize the face of a friend despite changes in viewing angle, distance and condition of illumination. For this to be possible, the sensory signal, highly variable across the different viewing conditions, has to activate, at some level, the same neuronal representation (that of our friend). From a computational point of view, hence, building an invariant representation is tantamount to partition the high-dimensional space of all possible inputs into a comparatively small number of clusters.

But how does the cortex build invariant representations? A popular idea is that these emerge from the hierarchical, feed-forward processing of the sensory signal [1]. The canonical example is the increase in size and complexity of the receptive fields along the ventral visual stream, a set of cortical areas involved in object recognition. Convolutional neuronal networks (CNNs) have been proposed as a model of the required processing [2]. In CNNs, the activity of the neurons in one layer are linearly combined, via adjustable synaptic efficacies, to generate the receptive fields of the neurons in the next layer. In fact, very complex receptive fields can be generated in these networks by suitably adjusting the synaptic efficacies and the number of layers.

Despite the undeniable successes of CNNs at image recognition, they are unsatisfactory from a neuroscientific point of view. In these networks, the connectivity is purely feed-forward while the cortex is strongly recurrent. About 50% of the synaptic connections onto a cortical neuron originate locally, that is, from neurons located within a few hundred micrometers from the receiving neuron [3]. Moreover, reciprocal connections among nearby neurons are stronger than average and over-represented, as compared to a random network [4].

Furthermore, in CNNs, the neurons in the same layer operate as almost-independent “feature extractors”, because they do not share synaptic connections. Weak local correlations also appear inconsistent with many experimental observations [5-7]. Stimulus-evoked responses of simultaneously-recorded neurons occupy only a subspace of all possible patterns, implying strongly correlated network states [5-6]. A case in point is the study by Bathellier et al. [7]. They presented naïve mice with different auditory stimuli while monitoring the activity of neuronal populations (~100 neurons) in the primary auditory cortex. Despite the large number of stimuli, the evoked responses clustered around a few “response modes” (2-5 modes).

Altogether, these experimental observations severely challenge the notion that invariant representations emerge from purely feed-forward processing. Rather, they suggest that categorical, and hence invariant, representations of stimuli already emerge in the primary sensory regions of the cortex, independently of any prior training [7].

In this research project, we will investigate the hypothesis that the clustering of stimulus-evoked responses observed in primary sensory cortices result from local, stimulus-driven attractor dynamics.

An attractor is a state, or a set of states, toward which a dynamical system evolves starting from a large collection of initial conditions. Hence, the system dynamically maps a large set of potential inputs (e.g., the different auditory stimuli in [7]) onto a small number of possible outputs (e.g., the response modes). In recurrent neuronal networks, the number and the dynamical nature (e.g., fixed points, limit cycles) of the attractor states depend mainly on the structure of the synaptic connectivity. In Theoretical Neuroscience, there is a long tradition of using attractor networks as a model of cortical dynamics for high-order associative cortices. By contrast, the pertinence of attractors to describe the dynamics, and the computations, of primary sensory cortices remains to be demonstrated.

The proposed research is interdisciplinary as it lies at the interface of theoretical physics, applied mathematics and system neuroscience. The concepts and methods that will be used will draw on the physics of disordered systems, dynamical system theory, machine learning, single neuron and synaptic physiology and neurophysiological studies in behaving animals. The research plan is articulated in 3 sub-projects:

(1) Dynamics in networks with clustered attractors. We will build on early work (e.g. [8]) with hierarchically-correlated attractors, extending it in the direction of biological realism. In particular, (i) we will use model networks that operate in the balanced excitation-inhibition regime and (ii) we will adapt/design synaptic learning rule(s) for the case in which synapses are either excitatory or inhibitory, consistently with Dale’s law. In this class of models, we will investigate under which conditions the network is able to reproduce the relevant experimental phenomenology. The theory will make predictions about the patterns of synaptic connectivity required to reproduce the response modes as observed in the experiment.

(2) Learning clustered attractors in a recurrent network. We will use techniques from machine learning to investigate, in a less constrained setting (i.e., without committing to a specific learning rule) which synaptic structure(s) are compatible with the observed experimental phenomenology. Nevertheless, learning will be constrained by Dale’s law, i.e., the efficacies of the synaptic connections cannot change sign as a result of the learning process. Both excitatory and inhibitory synapses can be “learned”. The results will allow us to validate and, possibly, expand upon the theory developed in sub-project (1).

(3) Synaptic signatures of clustered attractors in visual cortex. We will compare the neuronal dynamics and the synaptic connectivity obtained from the theoretical and computational studies in the sub-projects (1) and (2) with experimental data from mouse visual cortex. For this, we will employ the publicly available IARPA MICrONS dataset (https://www.iarpa.gov/research-programs/microns). The dataset contains both the functional imaging and the anatomical reconstruction of the primary visual cortex of a mouse. The functional imaging dataset contains about 75.000 pyramidal neurons with single-cell responses to a variety of visual stimuli. The anatomical dataset contains about 120.000 neurons (including the 75.000 for which the responses properties have been characterized) and more than 500 million synapses.

The research proposed will deepen the current understanding of the representational content of stimulus-evoked responses in primary sensory areas and elucidate the synaptic mechanisms underlying their emergence.

[1] DiCarlo JJ, et al. Neuron 73:415-434 (2012).
[2] Yamins DL, et al. PNAS 111:8619-8624 (2014).
[3] Harris KD, Mrsic-Flogel TD. Nature 503:51-58 (2013).
[4] Brunel N. Nat. Neurosci. 19:749-755 (2016).
[5] Luczak A, et al. Neuron 62:413-425 (2009).
[6] Schneidman E, et al. Nature 440:1007-1012 (2006).
[7] Bathellier B, et al. Neuron 76:435-449 (2012).
[8] Tsodyks, M. Mod. Phys. Lett. B 4:259-265 (1990).

Alzheimer's disease (AD) is characterized by a long asymptomatic phase that precedes the apparition of cognitive impairments by over a decade. Human brain imaging studies indicate that during this asymptomatic phase there is progressive build-up of amyloid beta (Aß) deposits, derived by enzymatic cleavage from the Amyloid Precursor Protein (APP) protein. The mechanisms initiating AD remain unclear. Using Drosophila, our Energy & Memory team has recently discovered an astrocyte-to-neuron H2O2 signaling (ANHOS) cascade essential for long-term memory (LTM) formation. Strikingly, we show that ANHOS requires a permissive role of APP, and is inhibited by human Aß. The objectives of the PhD project is to elucidate how H2O2 signaling mediates LTM formation in neurons of the olfactory memory center, and to investigate how this pathway is deregulated in genetic models of AD. This will project will be based on a combination of approaches, and in particular in vivo brain imaging.

Role of the Alzheimer's disease-related pathway APP in regulating H2O2 signaling during memory formation

Alzheimer's disease (AD) is a neurodegenerative disease that causes cognitive deficits, progressive loss of autonomy, and eventually death. About 35 million people are affected worldwide, and due to the aging of the population the number will rise to 66 million by 2030. Unfortunately there is currently no treatment for AD, and the initial causes of this pathology remain largely unknown. However, it is crucial to focus on the etiology of AD in order to identify early biomarkers and develop new avenues for treatment. Our project is to study in drosophila a switch from a physiological stage into an early pathological stage leading to AD, in relation to brain energy metabolism.
At the neuropathological level, AD is characterized by the progressive formation in the brain of neuritic plaques which correspond to the extracellular accumulation of amyloid beta (Aß) peptide, generated by cleavage of the transmembrane Amyloid Precursor Protein (APP), and by the formation of neurofibrillary tangles consisting of hyperphosphorylated TAU protein. Cerebral energy metabolism defects have been correlated with the onset of AD, and the disease is linked to an increased in brain oxidative stress. Our working hypothesis is that energy metabolism defects are among the initial causes of AD. Importantly, most of the proteins involved in humans in AD have homologs in the drosophila fly. In particular the APP protein, precursor of amyloid peptide Aß42 that accumulates in the brain of AD patients, has a unique homolog called APP-like (APPL). APPL is strongly expressed in the mushroom body, the center of the olfactory memory of drosophila, and we showed that this protein plays an important physiological role in associative memory (Goguel et al., J Neurosci 2011).
We recently engaged into an integrated study of the interplay between memory and energy metabolism. We showed that long-term memory formation critically relies on an acute increase of the metabolism of glucose-derived metabolites in the Drosophila mushroom body (Plaçais et al., Nat Commun 2017). Strikingly, our results demonstrate for the first time that memory formation involves glucose oxidation in the pentose phosphate pathway, which is a major regulator of oxidative stress. We also demonstrated the crucial role of glial cells in fueling the mushroom body during memory formation (de Tredern et al., Cell Rep 2021; Silva et al. Nat Metab 2022; Rabah et al. Nat Metab 2023).
Our project is to study the links between the APPL pathway and the regulation of energy metabolism during memory formation. The project is based on a strong set of preliminary results on signaling cascades involved in memory formation and energy metabolism. In particular, we recently showed that a gradient of H2O2 is required in the drosophila olfactory memory center for long-term memory formation (https://www.biorxiv.org/content/10.1101/2023.07.11.548505v1). The formation of this H2O2 gradient involves an interaction between cholinergic neurons and astrocytes, and the copper-binding domain of APP. We propose to perform an integrated study how APP transmembrane protein and its toxic derivative amyloid beta interact with energy metabolism at the molecular, circuit and behavioral levels. In practice, the project will involve the use of powerful drosophila genetics tools and behavioral assays that we master, along with functional imaging of energy metabolism by multi-photon fluorescence microscopy, for which our lab is a world leader.

Understanding the complexity of behaviours as well as their underlying neurobiological processes in normal and pathological conditions has become a major goal in Neurosciences. For that purpose, the study of behaviour in rodent models has become crucial but conventional procedures, often performed in non-adaptable commercial systems, are limited to test complex behavioural dimension. In this respect, we have developed the “Behavioural and Autonomous operant Box” where the animal can live and be exposed to complex behavioural tasks. These automated chambers ensure high quality results thanks to the massive amount of data collected in a more naturalistic way. The aim of this PhD project will be to: 1) optimize the hardware and software to make the BEATBox easily accessible to the scientific community, 2) standardize the processing of the data acquisition and formatting to allow easier data sharing and 3) adapt machine-learning based tools for analysis of complex behavioural datasets.

Understanding the complexity of behaviours as well as their underlying neurobiological processes in normal and pathological conditions has become a major goal in Neurosciences. For that purpose, the study of behaviour in rodent models has become crucial but conventional procedures, often performed in non-adaptable commercial systems, are limited to test complex behavioural dimension relevant for translational aspects. In this respect, we have developed the “Behavioural and Autonomous operant Box” (https://nerb.team/index.php/beatbox/) where the animal can live and be exposed to complex behavioural tasks for several weeks. These automated chambers ensure high quality results thanks to the massive amount of data collected in a more naturalistic way. The aim of this PhD project will be to: 1) optimize the hardware and software to make the BEATBox easily accessible to the scientific community, 2) standardize the processing of the data acquisition and formatting to allow easier data sharing and 3) adapt machine-learning based analysis tools for automatic analysis of complex behavioural datasets.

Proposal:
Behavioural neurosciences is aiming at understanding the complexity of behaviours and their underlying neurobiological processes at stake (Mainen et al., 2016; Niv, 2020), in both normal and pathological conditions. In this purpose, animal models have become essential over the past years as there are multiple powerful tools to monitor (Siegle et al., 2017; Steinmetz et al., 2018) and/or manipulate their neural activity (Whissell et al., 2016) while they are performing behavioural tasks. Nevertheless, several challenges remain to refine and fully exploit animal models behavioural data. First, we need to accurately reflect the behavioural outputs by collecting enough data per subject. To ensure the good quality of these data, the animal should be tested in appropriate conditions where its naturalistic rhythm is preserved. Second, current neuroscientific questions call for the design of novel complex tasks that can reflect, as much as possible, some sensorimotor or cognitive process observed in humans (Nithianantharajah et al., 2015). Thus, we need to translate challenging behavioural tasks for mice that require highly flexible design of the experimental apparatus and sometime long-lasting training of the animals.
For this purpose, the development of operant chambers to implement complex behavioural tasks like the human paradigms have been extensively used. Recently, the diversity and complexity of learning procedures have increased thanks to the use of visual touchscreens (Bussey et al., 2008). Yet, these kinds of commercial devices remain overly expensive with little flexibility. As a result, some teams have developed open-source and low-cost systems with the same purpose (Devarakonda et al., 2016; O’Leary et al., 2018).
Although these systems have drastically improved the quality of the implemented behavioural tasks, some limitations remain. Despite a large progress in automation (Schaefer and Claridge-Chang, 2012), the experimenter is still involved in numerous steps along the experimental process since most of the studies rely on a discontinuous protocol during which the animals perform the task every day for 30 minutes to several hours (O’Leary et al., 2018). Not only this limits the amount of acquired data but it also introduces a great amount of variability as multiple factors can alter the results of animal studies (Balcombe et al., 2004), such as handling, water/food restriction and disturbance of circadian rhythms.
To palliate these limitations, we have developed the Behavioural and Autonomous operant Box (BEATBox, https://nerb.team/index.php/beatbox/). This is an operant chamber highly customizable by design and entirely produced from Computer-Aided Design with manufactuting technologies such as 3D printing and and laser cutting of acrylic panels. It is operated by the open-source electronics platform Arduino and is entirely piloted by an in-house software run on a general-purposecomputer. The BEATBox system allows the administration of a various range of complex and long-lasting behavioural procedures to multiple animals at the same time. The main interest of this autonomous system is that mice can live and freely move in the operant chamber. Thus, the animals self-pace their activity to engage in the tasks with very few interventions of the experimenter. The automation of the system also guarantees a high throughput data acquisition while drastically reducing experimenter’s time.
Although our first versions of the BEATBOX met its original aim, our current project aims to go further in its implementation and development in three aspects:
1) Enhace Hardware and Software Components: The objective is to refine both the hardware and sofware aspects of the experimental box, ensuring it is fully prepared for widespread adoption within the scientific community. This refinement will pave the way for the comprenhensive utilization and valorisation of this technology.
2) Implement data standardization. Data standarization, or the process of establishing consistent methods and practices for collecting and deport behavioral data is essential to ensure data comparability, research reproductibility, cross-study analysis and longitudinal studies. Although it is of great importance, the standardization of behavioural data is an aspect that, compared to electrophysiological or imaging data, is currently not very well developed or adapted across scientific groups. This is why we plan to propose a standardize data collection process by developing a ready-to-use data management software application and prove its use and advantages in different behavioral tasks using our experimental box.
3) Develop Dynamic Behavioral Characterization. A crucial aspect of our project involves the creation of machine-learning based algorithms to assess the changes in behavioral parameters, such as learning processes and shifts in strategies, throughout various tasks that we have previously implemented and validated in the laboratory.

Deciphering how the dazzling diversity of neuronal types of the nervous system (CNS) emerges from seemingly identical neural progenitors during development constitutes a long-standing question in Neuroscience. We proposed to address this question in the neural retina, a network of >70 types of neurons providing easy experimental access compared to the brain. The project will combine state-of-the art technologies developed by two research teams at Institut de la Vision for cell lineage tracing with combinatorial labels and functional neuron characterization with multiphoton Calcium imaging and optogenetics. Using these approaches, we will mark clones of retinal neurons originating from individual retinal progenitor cells during development, and functionally characterize their cell-type composition. This will enable us to understand the logic of neuronal type production by retinal progenitors and probe how the clonal origin of retinal neurons influences their connectivity.

The central nervous system (CNS) comprises a dazzling diversity of neuronal types whose specific morphology, connectivity and functional properties are key to proper circuit operation. Deciphering the origin of this diversity constitutes a long-standing question in Neuroscience, whose answer will have profound fundamental and translational implications. The neural retina is a model of choice to address this question: it comprises a vast diversity of neurons which can be grouped in 6 main classes occupying specific retinal layers, and >70 neuronal types interspaced across its surface according to semi-regular mosaic patterns. Embryologically, the retina derives from the diencephalon and is thus part of the CNS. As in the brain, all retinal neurons are generated during development from seemingly identical neural progenitors. Lineage tracing studies conducted in vivo and in vitro have demonstrated that early retinal progenitor cells (RPCs) are multipotent and produce the 6 classes of retinal neurons in sequential yet partially overlapping waves: ganglion cells (RGCs), followed by cone photoreceptors (PRs), horizontal cells (HCs), amacrine cells (ACs), followed by rod PRs and bipolar cells (BCs). Clones generated by individual RPCs typically form columns comprising cells of all retinal classes.

According to the prevalent model, RPCs may transition through different states of competence to generate these different classes of neurons. However, this model does not explain how the full diversity of retinal neuron types is generated. An intriguing possibility is that cell lineage could play a role in the process of cell type determination in the retina. Indeed, certain types of amacrine cells have been shown to be already specified around their birth, and direction-selective RGCs may originate from a molecularly defined RPC subpopulation. Testing whether retinal type identity may be encoded in their lineage requires tracing back their developmental origin with both single-cell precision and high throughput. Yet, clonal analysis techniques based on monochrome markers require very sparse labeling to achieve clonal resolution. The Brainbow strategy efficiently resolves cells in a population using combinations of fluorescent protein (FPs) which are faithfully transmitted to their progeny through cell division. Based on this idea, the Livet team at Institut de la Vision (IDV) has developed optimized Brainbow transgenes and mouse lines for multiplex clonal tracking in the developing nervous system, using FP color labels activated by Cre/lox recombination [1, 2]. These transgenes efficiently delineate the columns of retinal neurons generated by individual RPCs. The team has also introduced a breakthrough scheme termed “iOn” to experimentally manipulate neural stem cells and their output using exogenous DNA transgenes activated by genomic integration, thanks to a novel type of genetic switch [3]. iOn vectors open the way to complex genetic manipulations and provide a new avenue to rapidly probe gene function in the intact retina through simple electroporation of RPCs. The Emiliani and Marre teams (also at IDV), who will collaborate on the project, have established methods to finely characterize retinal neuron types by recording their response to different light stimuli using Calcium imaging [4].

The proposed project will aim at probing the relation between neuronal identity and linage in the retina, by applying the above methodologies to characterize the cell type composition of clones of retinal neurons originating from individual mouse RPCs, in normal and experimentally perturbed contexts.
The first objective of the PhD will be to establish the conditions for combined functional and lineage characterization of mouse retinal neurons. The student will generate retinas with multicolor labeling of retinal clones by triggering Brainbow clonal labels in RPCs at the beginning of retinal neurogenesis. Functional analysis will take place in mature retinas ex vivo, using a genetically encoded fluorescent Calcium sensor expressed by AAV vectors enabling restricted labeling of neuronal classes (BCs, ACs or RGCs). Calcium activity will be recorded by two-photon imaging in response to an array of light stimulation patterns, enabling to functionally classify cell types according to their response. Following this, retinal clones will be mapped at high resolution with multichannel confocal microscopy.
Based on this approach, the second objective of the project will be to probe the cell-type composition of retinal clones. We will first seek to characterize biases in the composition of retinal clones within individual classes of neurons, starting with RGCs and followed by ACs and BCs. For instance, we will test whether clonally-related RGCs more frequently comprise (or exclude) certain cell types or associations of cell types (such as OFF, ON-OFF, fast-ON…). We may then try to extend our analysis to clonal biases between distinct classes of retinal neurons, such as RGCs and BCs. Results from this part of the project will provide unprecedented data on the links between the clonal and functional organization of the retina.
In a third objective, we will test to which extent clonally-related neurons are functionally connected. Using the approach established in Objective 1, we aim at defining maps of functional connections between distinct classes of retinal neurons clonally-related in a column. We will employ two-photons optogenetics to activate presynaptic cells and to monitor post-synaptic cells activity with Calcium imaging. We will use AAVs to genetically encode a protein sensitive to light (opsin) in a presynaptic class of neurons, i.e. BCs, and a Calcium indicator in a different class of neuron, i.e. RGCs [4]. Emiliani and Marre teams demonstrated previously that it is possible to locally map local connections between neurons in the retina and in cortical layer2/3, by activating one or more pre-synaptic cells with targeted two-photons illumination and by simultaneously recording the activity in post synaptic cells with Calcium imaging or single cell electrophysiological recordings [4,5].

The project will take place at IDV in the center of Paris. The student will be co-mentored by Jean Livet (neurodevelopment, genetic engineering) and Valeria Zampini (functional imaging and optogenetics), with the collaboration of Olivier Marre (neurophysiology of the retina). Overall, the project will provide a quantitative exploration at the single-cell level of the links between cell lineage and functional organization in the retina. It will enable to tackle major questions, such as the composition of retinal clonal units. It will also provide a rich multidisciplinary training for the PhD student concerning state-of-the art techniques to analyze and link neuronal function and development. The approaches generated during this project will have general applicability in Neuroscience.

References :
1. K. Loulier et al. Neuron 2014
2. S. Clavreul et al. Nat Commun 2019
3. T. Kumamoto et al. Neuron 2020
4. G. Spampinato et al., Cell Rep Methods, 2022
5. Shemesh et al., Nat Neurosci. 2017

Animal behavior, analogous to language, unfolds as a series of organized motifs in time and space, modulated by sensory inputs. As an example, we can think of the actions sequence required to grab and drink a cup of coffee or to execute complex tasks such as hunting or grooming. Concurrent to this functional organisation, brain-scale spontaneous neuronal activity exhibits highly-structured spatio-temporal dynamics, characterised by the stochastic exploration of discrete metastable states. However, how large neuronal networks composed of neurons, operating on the millisecond time scale, can yield slow structured dynamics that in turn govern spontaneous behavioural sequences is not well understood. The objective of this PhD is to combine brain-wide recordings, optogenetic circuit interrogations, and machine learning tools to characterise and model internal brain dynamics that govern spontaneous behaviour in zebrafish larvae.

Behaviors, like a human drinking coffee, a fish hunting, or a fly grooming, are a series of self-initiated actions whose timing and sequence rely on inherent or learned programs and sensory feedback. Significantly, in all these instances, the succession of various elementary actions transpires over timescales ranging from a fraction of a second to several seconds. This duration is orders of magnitude larger than the typical timescale controlling the internal dynamics of spiking neurons 1,2.

This observation suggests that comprehending the neuronal processes underlying spontaneous behaviors necessitates understanding the collective dynamics of extensive neuronal circuits. These circuits can be viewed as high-dimensional dynamic systems, facilitating emergent, persistent timescales in spontaneous dynamics originating behavioral sequences and other essential mechanisms like episodic memory or sensory response modulation.

In this PhD project, we will address the following question: How can large neuronal networks, comprised of neurons operating on a millisecond timescale, yield slow, structured dynamics that subsequently govern spontaneous behavior?

Using functional light-sheet microscopy, we will record brain-wide cell-resolved brain dynamics 3,4. From these recordings and using data-driven network models, we will then infer the relevant functional cell assemblies and the most likely brain states. Using holographic two-photon mediated targeted optogenetic stimulation5 we will then inhibit or excite selected neuronal populations while simultaneously recording the brain-wide network response. These will be directly compared to similar assays performed in silico using the trained network model.

The model organism: Experiments will employ zebrafish larvae, uniquely suited due to its small, transparent brain which accommodates brain scale optical recording and perturbation techniques

The model: We will model brain-scale dynamics using data-driven neural network models such as restricted Boltzmanm Machines (RBMs). In these energy-based models, network dynamics is represented as a stochastic exploration of a high-dimensional energy landscape, possessing multiple metastable states 2. Consequently, spontaneous behavior is mirrored in the transition probabilities between these discrete states.

First, we will systematically stimulate each identified cell assembly, observing both the average brain-wide response and the resultant behavior. Two primary outcomes are anticipated:

(i) Derivation of the inter-cell assembly connectivity matrix, elucidating how cell assemblies interact to form metastable brain states. This matrix will be quantitatively compared to the model prediction and, additionally, to structural connectivity data extractable from a publicly accessible database of anatomical projection patterns.
(ii) Identification of cell assemblies that maintain a causal relationship with behavioral outputs.

Second, cell assemblies and metastable brain states inferred by the model are thought to be network attractor states. A hallmark of attractor states is pattern completion. Stimulating a subpopulation of neurons that form a cell assembly should lead to the activation of the remaining neurons of the assembly. Stimulating a subpopulation of cell assemblies that form a metastable brain state should activate the full brain state. We will test these hypotheses using controlled two-photon mediated targeted optogenetic stimulations. We will vary the percentage of stimulated subparts and will compare the pattern completion capacity with the model predictions.

Third, being able to optogenetically bring the brain repeatedly into a specific state will allow us to study the subsequent temporal evolution of the network starting from this specific initial point and thus to probe locally around this metastable brain state the energy landscape that drives the network dynamics. We will characterise the average dwell time in the forced brain states, map the subspace of states into which the system can transition once in a specific state, and characterise the corresponding transition probabilities. We will compare our results to in silico optogenetic data in which the Monte Carlo dynamics for sampling brain states from our model is initiated with a similar activity pattern as in the in vivo experiment.

As a long-term aim, we will follow the network state using voltage imaging to increase the temporal resolution. We might resolve how a brain state is activated starting from a seed stimulus and we hope to resolve phase sequences of hidden unit activity along which the brain transitions between metastable states.

1. Wolf, S., Goc, G. L., Debrégeas, G., Cocco, S. & Monasson, R. Emergence of time persistence in a data-driven neural network model. eLife 12, e79541 (2023).
2. Plas, T. L. van der et al. Neural assemblies uncovered by generative modeling explain whole-brain activity statistics and reflect structural connectivity. eLife 12, e83139 (2023).
3. Panier, T. et al. A Versatile and Open Source One- and Two-Photon Light-Sheet Microscope Design. bioRxiv 2023.07.10.548107 (2023) doi:10.1101/2023.07.10.548107.
4. Panier, T. et al. Fast functional imaging of multiple brain regions in intact zebrafish larvae using Selective Plane Illumination Microscopy. Frontiers in neural circuits 7, 65 (2013).
5. Hubert, A. et al. Random-access two-photon holographic optogenetic stimulation combined with brain-wide functional light-sheet imaging in larval zebrafish. Adv. Microsc. Imaging IV 12630, 1263007-1263007–4 (2023).

Down-syndrome is associated with language delays whose origins remain largely unknown and underspecified. A fascinating study recently showed that the administration of Gonadotropin-releasing hormone (GnRH) improved cognition, language in particular, in this population (Manfredi-Lozano et al., 2022). In this thesis project we will characterize and precisely quantify the language disorder in down-syndrome kids using the measurement of speech-related neural oscillations (Wang et al. 2023) and follow-up the language improvement after GnRH treatment. The project is partly funded by the Swiss National Fund and the EEG recordings will take place in Lausanne with the mobile set-up of the Institut de l’Audition.

At present, there is no viable treatment for cognitive deficits in Down syndrome. In the Down syndrome Ts65Dn mouse model, neurological symptoms 1) follow the postpubertal decrease in the expression of a master molecule that controls gonadotropin-releasing hormone (GnRH) and 2) are related to an imbalance in a microRNA-gene network known to regulate GnRH neuron maturation, ultimately leading to altered hippocampal synaptic transmission. In the mouse model, epigenetic, chemo-genetic, cellular, and pharmacological interventions aiming at restoring physiological GnRH levels have been shown to rescue olfactory and cognitive functions. Consistently, pulsatile GnRH therapy improves cognition and brain connectivity in adult patients with Down syndrome. Interestingly, relatives of treated subjects reported significant language improvement along GnRH pulsatile therapy, reaching marginal statistical significance (Manfredi-Lozano et al., 2022). However, language improvement has so far not been deeply characterized and quantified. In this project we hypothesize that the language improvement is paralleled by a resynchronization of speech and language neural networks. The PhD project consists in deploying proper tools to follow-up the language functions during GnRH treatment both behaviorally and neurophysiologically with high-density EEG.
The PhD will be part of the SNF-funded PulseUp project, where our contribution will be to measure language levels in DS patients before, during and after therapy, while following up the reorganization of speech and language neural synchrony using protocols already established and validated in the team in children with autism (Wang et al., 2023; Vinurel et al., ongoing).
The PhD student will be part of the NeuroSpeech team at the Institut de l’Audition, benefitting from its intellectual and technical environment, and will carry our part of their research at the CHUV in Lausanne with Prs. Nelly Pitteloud and Bogdan Draganski.

Many animals are capable of learning complex, rapid movements. Because feedback is often too delayed for real-time control, many such movements can only be learnt by iteration. The cerebellum is thought to play an integral role in such learning, and a theory for how this learning is implemented by synaptic plasticity within the cerebellar network has been developed over several decades. Key to this implementation is the instructive role played by the climbing fibre, which is thought to convey information about movement errors and instruct plasticity of the abundant granule cell--Purkinje cell synapses. In recent work we have analysed this theory and shown that it has severe limitations. We have proposed a more capable algorithm and suggested a detailed implementation, in which the climbing fibre plays a dual role. We seek a student to join our program to test the competing theories of cerebellar function using multidisciplinary in vivo, in vitro and theoretical approaches.

We have proposed a novel algorithm and specific implementation for iterative motor learning by the cerebellum [1]. We show that this algorithm is much more capable than the classical Marr-Albus-Ito theory. The new theory, a form of "stochastic gradient descent", incorporates a dual role for climbing fibres: not only do they signal movement error, as in the classical theory, but their spontaneous activity represents network probes of modified motor commands that are consolidated or not depending on whether the movement error is decreased or increased by the trial modification. The theory predicted a set of synaptic plasticity rules that contradicted the existing literature but that we were able to verify by performing in vitro synaptic plasticity experiments under more physiological conditions than previously employed [1]. These novel plasticity rules have since been largely replicated by an independent group [2]. Another key prediction of our theory is that spontaneous climbing fibre activity can, through the complex spikes induced in Purkinje cells, perturb movements. This is crucial because trial modifications of the movement must be generated. Recent work by another independent group has tested and verified this aspect of our theory during ocular saccades [3].

Since publishing our theory, we have carried out experiments to test the competing theories in vivo. We chose to perform chronic multielectrode recordings from Purkinje cells in a classical model cerebellar learning behaviour: eyeblink conditioning. We have performed the first chronic recordings during this behaviour and are currently analysing how the patterns and histories of complex spike activity control firing modifications of individual cells tracked throughout the learning, with the aim of distinguishing the two competing theories. The analysis of these data is being finalised. As part of this work we have developed a novel spike-sorting post-processing pipeline, initially to improve analysis of Purkinje cells, but which offers benefits to multielectrode spike-sorting generally [4].

The student will be able to pick and choose from in vivo, in vitro and theoretical approaches. In addition to the established techniques of in vivo multielectrode recordings, slice electrophysiology and computational approaches, the project will involve optogenetic interventions and can be extended to include imaging experiments during the establishment and investigation of a richer learning behaviour that challenges the limitations of current theory.

[1] Bouvier, Aljadeff, Clopath, Bimbard, Ranft, Blot, Nadal, Brunel, Hakim and Barbour (2018) Cerebellar learning using perturbations. eLife. https://doi.org/10.7554/eLife.31599

[2] Titley, Kislin, Simmons, Wang and Hansel (2019) Complex spike clusters and false-positive rejection in a cerebellar supervised learning rule. J Physiol. https://doi.org/10.1113/JP278502.

[3] Muller, Pi, Hage, Fakharian, Sedaghat-Nejad and Shadmehr (2023) Complex spikes perturb movements, revealing the sensorimotor map of Purkinje cells. bioRxiv. https://doi.org/10.1101/2023.04.16.537034

[4] Llobet, Wyngaard and Barbour (2022) Automatic post-processing and merging of multiple spike-sorting analyses with Lussac. bioRxiv. https://doi.org/10.1101/2022.02.08.479192