Gliogenesis in the cerebral cortex begins as neurogenesis declines near birth and continues postnatally. In models like the Reeler mouse, which lacks Reelin, cortical neurons are mislayered—and astrocytes are also affected. Whether Reelin directly signals to astrocytes during development remains unknown. This project will use mouse genetics to investigate non-cell-autonomous effects of Reelin on astrocyte development. We will combine genetic tracing with spatial transcriptomic and single-cell RNA sequencing to study astrocyte lineage and diversity in the Reeler cortex. In utero electroporation of Magic Markers plasmids will allow clonal tracing based on astrocyte color and morphology. This project aims to uncover how neuron-astrocyte communication shapes cortical development and to define the molecular mechanisms involved.

1) Project description
Astrocytes are critical for proper brain development and functioning. During development, astrocytes actively shape neural circuits by regulating dendritic growth, synaptogenesis, and synaptic plasticity. At later stages, they provide structural and metabolic support, clear neurotransmitters, support myelin integrity, and respond to injury with inflammatory reactivity. Recent advances in imaging, spatial transcriptomics, and single-cell RNA sequencing have significantly expanded our understanding of astrocyte development. These tools have revealed astrocyte heterogeneity in morphology, reactivity, and molecular identity, reflecting their diverse embryonic origins and regional specialization in the forebrain.
Gliogenesis begins as neurogenesis declines near the end of gestation and continues postnatally. In the neocortex, astrocytes arise from different progenitor sources from both pallial and subpallial origins: the Radial glial cells, the oligodendrocyte progenitor cells (OPC) in the lateral ventricular zone (VZ) and the embryonic subpallial progenitors. Astrocyte proliferation occurs in two waves, before and after migration into the cortex. The first wave takes place in the VZ, and the subventricular zones (SVZ), where embryonic astrocyte progenitors are generated from RGCs and OPCs. These progenitors then migrate before birth into the cortex using either erratic paths or blood vessel-guided migration. The second wave of astrocyte proliferation occurs between cortical layers, where already differentiated astrocytes divide symmetrically to expand the population between postnatal days P4 and P10. These cells migrate away from their progenitors and intermix until spatial constraints limit further dispersion. In parallel, new astrocyte progenitors continue to arise from the ventricular zone, colonizing the cortex in a scattered pattern. Finally, astrocytes mature by expanding their volume and branching, establishing connections with neurons, other glial cells, and blood vessels. Similarly to neurons, astrocytes are also organized in layers with 3 laminae defined by the expression patterns of specific markers. Surprisingly, clonal analyses have revealed that single progenitors can give rise to astrocytes with diverse morphologies and molecular profiles, highlighting the critical role of the local environment in shaping their final identity and maturation.
However, little is known about how developing neurons communicate with astrocytes during cortical development. One candidate molecule is Reelin, a large secreted glycoprotein expressed during embryonic development by transient glutamatergic neurons Cajal-Retzius (CRs) cells in the marginal zone, and later by interneurons (INs) throughout the cortex. In mice, Reelin deletion results in the Reeler phenotype, characterized by severe cerebellar hypoplasia and an inversion of cortical neuronal layers. In Human, mutations are responsible for cortical malformations ranging from lissencephaly with cerebellar hypoplasia to polymicrogyria or pachygyria but also for multiples neuropsychiatric and neurodegenerative disorders without apparent brain malformations. Beside these well-known actions of Reelin in neurons mediated by the ApoeR2 and Vldlr receptors, very few studies have looked at its role in other brain cells, especially in macroglia.
Neonatal ablation of CRs results in an increased number of reactive astrocytes in the cortex. Astrocytes also show layering defects as for neurons in the Reeler mouse suggesting that layer-specific neuron-astrocyte interactions help establish astrocyte layer identity and function. Finally, the Reelin/Dab1 pathway also controls the astroglial ensheathment of synapses, modulating neuronal excitability, synaptic transmission and plasticity in postnatal brains.

2) Research Program
In this context, we propose to test in vivo whether Reelin is involved in the proliferation, migration and differentiation of astrocyte progenitors (or a subset of them) in the cerebral cortex during perinatal stages. To do so, we will use both non-cell-autonomous and cell-autonomous approaches to alter Reelin production in neurons and its signaling pathway in astrocytes, respectively. We will track distinct astrocyte lineages, by combining genetic tracing (Cre/Lox technology) and immunolabeling with the pan-astrocyte Sox9 marker, alongside markers specific to different astrocyte subpopulations. Potential regional differences across brain areas will be captured by examining three rostro-caudal brain sections along with the mediolateral axis. In addition, we will analyze astrocyte developmental trajectories using single-cell RNA sequencing and spatial transcriptomics.

Aim 1: Non-cell autonomous effects of Reelin on astrocytes
We will study the consequences of Reelin loss-of-function (lof) on astrocyte development. This is a challenging task since neurogenesis precedes gliogenesis in RGC, and overlap occur at late gestation. Since Reelin deletion disrupts neuronal layering before astrocyte migration begins, it is essential to distinguish between direct and indirect effects of Reelin loss on astrocyte development. As mentioned previously, the HRM model present normal neuronal lamination and has been successfully used to identify Reelin function on OPC proliferation and migration. In the laboratory, we have developed genetic models to specifically target the Reelin Lof in CRs versus INs using specific Cre lines. As for the hypomorph HRM model, the specific deletion of Reelin in CRs or INs has minor lamination defects, although their combined Lof result in lamination deficits reminiscent of the Reeler model (Unpublished data). We have preliminary results in the CRs-lof model supporting a 4 times reduction in Emx1-derived astrocytes in cortex at P10, suggesting that Reelin may control either the proliferation of their progenitors or their survival. Therefore, we will perform Brdu injections and immunolabeling experiments using activated caspase 3, Sox9 and Olig2 markers to monitor, the proliferation and death of the different astrocytes lineages, respectively. In parallel, we will perform IUE of Magic Markers plasmids at E15 in HRM to label a limited number of RGC clones and study the derived astrocyte clones. Pups will be analyzed at 3 critical stages according to the astrocyte development.
Altogether, these results will describe for the first time a role for Reelin in astrocyte development, and identify its specific functions in proliferation, migration or maturation.

Aim 2: Cell autonomous alteration of Reelin signaling in astrocytes
To directly assess the effect of Reelin pathway deficit in astrocytes, we will conditionally inactivate a key Reelin pathway target: the intracellular adaptor Dab1, which is phosphorylated upon Reelin receptors activation. We will specifically target Reelin Lof in astrocytes using a transgenic mouse expressing a temporally controlled Cre recombinase. Cre activity will be induced at E15 by Tamoxifen (Tx) injections to pregnant females to target the gliogenesis phase and brain samples of pups will be collected at P4, P10 et P21 stages. The consequence of Dab1 Lof in astrocytes will be assessed by immunolabeling and spatial transcriptomic on cortical sections to identify the different astrocyte sub-populations and scRNAseq experiments will be performed to search for altered trajectories in astrocyte lineages. Furthermore, we will inject Tx at later stages, P10 and P21 to study the role of Reelin in local proliferation and maturation, respectively. Finally, we will inject Tx in young pups (P4) to produce cohorts and study their behavior in adult focusing on tests associated with brain regions with astrocyte phenotypes.
Altogether, this new project on the regulation of astrocyte development by Reelin will highlight the importance of neurons to astrocytes communications during development.

Paramecium is a large 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 habituation to mechanical stimulation: when Paramecium is repeatedly touched, its reactions become weaker, then recover in the absence of stimulation. This phenomenon is known in animals as an elementary form of learning, and Paramecium offers a unique opportunity to decipher its mechanisms. The investigation will use a combination of modern techniques, including behavioral analysis, 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” [1]. 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 habituation in Paramecium. Habituation is considered the most elementary form of learning, in which the reaction to a repeated stimulus weakens, and then recovers in the absence of stimulation. This phenomenon was initially studied in animals, but it also appears in ciliates, where it has been characterized most exhaustively in the contraction behavior of Stentor, a trumpet-shaped ciliate that attaches to surfaces and contracts when mechanically stimulated [2]. However, while habituation behavior is easy to study in Stentor, electrophysiology is difficult. In contrast, Paramecium benefits from dozens of electrophysiological studies, including measurements of mechanotransduction currents [3]. As many ciliates, it is highly mechanosensitive. When it is touched on the anterior end, a calcium-based mechanoreceptor opens, depolarizes the membrane and may trigger an action potential leading to an avoiding reaction (backward swimming). When it is touched on the posterior end, a potassium-based current is triggered, hyperpolarizing the membrane and leading to an acceleration of swimming. The host lab has found both with behavior and electrophysiology that these mechanosensitive responses strongly habituate. This makes Paramecium a unique model organism to study habituation.

The project will tackle three complementary questions, focusing more specifically on the posterior mechanoresponse, which is stronger. First, what is the molecular basis of mechanosensitivity in Paramecium? The mechanoreceptors have been characterized electrophysiologically but not identified. Based on preliminary findings, our working hypothesis is that the posterior mechanosensitive current comes from either a Kv1 channel or a calcium-activated potassium channel, with the primary signal provided by a calcium-based mechanoreceptor. To explore it, the student will first use pharmacological blockers together with behavioral tests. The host lab has designed a simple vibration-based assay to test mechanosensitivity. Genes can then be identified with RNA interference by feeding (cells eat bacteria with the corresponding plasmids).

Second, what are the behavioral and electrophysiological characteristics of habituation in Paramecium? Previous studies in animals have identified several hallmarks of habituation (dependence on stimulation frequency and intensity, recovery, etc.). The student will examine them first with behavioral tests, then with electrophysiological measurements.

Third, what are the mechanisms of habituation? Previous work in Stentor suggests that the decay of mechanosensitive currents with repeated stimulation might be due to a change in voltage-sensitivity of the mechanoreceptor. This suggests that the phenomenon involves a phosphorylation pathway modulated by stimulation, possibly through the entry of calcium. The student will use pharmacology, RNA interference and behavioral analysis to decipher this pathway, and use the results to build a model.

The host lab has been working for many years on Paramecium with electrophysiology, modeling, behavior and molecular biology [1,4–7]. It is equipped with an electrophysiological setup to measure mechanosensitive responses (currents) with controlled mechanical stimuli, wide field microscopes and custom tracking software to measure behavioral responses, as well as vibrating devices (controlled by Arduino) to trigger mechanosensitive responses. The lab has experience in using pharmacological blockers as well as genetic tools (RNA interference to inhibit specific genes). It also has considerable experience in biophysical modeling of neurons as well as of ciliates specifically. In particular, the lab recently developed an electrophysiogical model of the action potential of Paramecium, coupled to its motion [5]. Finally, the lab is involved in close collaborations with Laboratoire Jean Perrin (Alexis Prevost and Léa-Laetitia Pontani), an experimental biophysics lab, and with Sandra Duharcourt, an expert in genetics and molecular biology of Paramecium in Institut Monod.



1. Brette R. Integrative Neuroscience of Paramecium, a “Swimming Neuron.” eNeuro. 2021;8: ENEURO.0018-21.2021. doi:10.1523/ENEURO.0018-21.2021
2. Wood DC. Parametric studies of the response decrement produced by mechanical stimuli in the protozoan, Stentor coeruleus. Journal of Neurobiology. 1969;1: 345–360. doi:https://doi.org/10.1002/neu.480010309
3. Machemer H, Deitmer JW. Mechanoreception in Ciliates. In: Autrum H, Ottoson D, Perl ER, Schmidt RF, Shimazu H, Willis WD, editors. Progress in Sensory Physiology. Berlin, Heidelberg: Springer Berlin Heidelberg; 1985. pp. 81–118. doi:10.1007/978-3-642-70408-6_2
4. Kulkarni A, Elices I, Escoubet N, Pontani L-L, Prevost AM, Brette R. A simple device to immobilize protists for electrophysiology and microinjection. Journal of Experimental Biology. 2020;223. doi:10.1242/jeb.219253
5. Elices I, Kulkarni A, Escoubet N, Pontani L-L, Prevost AM, Brette R. An electrophysiological and kinematic model of Paramecium, the “swimming neuron.” PLOS Computational Biology. 2023;19: e1010899. doi:10.1371/journal.pcbi.1010899
6. Escoubet N, Brette R, Pontani L-L, Prevost AM. Interaction of the mechanosensitive microswimmer Paramecium with obstacles. Royal Society Open Science. 2023;10: 221645. doi:10.1098/rsos.221645
7. Hosseini A, Fosse C, Awada M, Stimberg M, Brette R. Single camera estimation of microswimmer depth with a convolutional network. Journal of The Royal Society Interface. 2025;22: 20250428. doi:10.1098/rsif.2025.0428

CADASIL, the leading monogenic small-vessel disease, offers a unique model to understand how microvascular lesions drive cognitive, emotional, and motor decline. This PhD will leverage >2,500 longitudinal MRI scans with detailed clinical data to identify structural vulnerability networks and lesion thresholds predicting depression, cognitive slowing, memory impairment, and dementia. The candidate will use advanced image-processing and statistical pipelines, including automated lesion segmentation on T2/FLAIR, voxel-based morphometry (VBM), voxel-based lesion–symptom mapping (VLSM), lesion-network analysis, and longitudinal mixed-effects models. Noise reduction, MRI harmonization, and rigorous validation of lesion quantification will be key. Outcomes will advance mechanistic understanding and support predictive biomarkers for small-vessel disease.

Cerebral small-vessel disease (cSVD) is a major cause of cognitive decline, gait impairment, depression, and dementia. Yet the pathways linking microvascular injury to symptoms remain only partially understood, partly because sporadic cSVD develops slowly and presents substantial heterogeneity. CADASIL, the most frequent monogenic cSVD caused by NOTCH3 mutations, provides a unique model to study how microvascular damage accumulates and disrupts brain networks. With sensitive clinical and MRI measures, CADASIL allows precise tracking of lesion development and related functional consequences over time.

Our team has assembled the world’s largest longitudinal imaging dataset in genetically confirmed CADASIL: more than 2,500 structural MRI scans (baseline and follow-up) with standardized clinical and neuropsychological assessments. This unique resource enables fine-grained analysis of lesion topography, cumulative lesion burden, and clinical outcomes across cognition, mood, and motor function.

The objective of this doctoral project is to identify lesion patterns and structural brain networks whose damage predicts cognitive, emotional, and motor decline, and to determine combined volumetric and topographic thresholds associated with clinical progression.

The work will address four major domains:

Mood and emotional regulation
Evaluate whether injury to frontal–subcortical and limbic pathways, including cingulum, fronto-striatal tracts, and thalamo-cortical projections, increases risk of depression and emotional dysregulation.

Cognitive slowing and executive impairment
Identify white-matter lesion patterns affecting processing speed and executive functions, focusing on frontal networks, callosal fibers, and associative pathways.

Memory impairment
Investigate whether clustering of lesions in temporal and limbic pathways, or secondary network degeneration, contributes to episodic memory deficits in specific patients.

Global cognitive decline and dementia
Establish lesion-volume and topographical thresholds associated with dementia progression, and assess whether spatial metrics outperform total lesion burden in prediction.

The candidate will apply advanced image-processing and statistical modeling pipelines. Key methods include automated lesion segmentation on T2/FLAIR images, voxel-based morphometry (VBM), voxel-based lesion–symptom mapping (VLSM), lesion-network approaches when appropriate, and longitudinal mixed-effects models. Attention to noise reduction, harmonization of MRI protocols across time points, and rigorous validation of lesion segmentation will be essential. Analyses will integrate permutation-based statistical inference, multiple-comparison correction, and reproducibility checks. Where relevant, exploratory machine-learning components may be used to identify latent lesion vulnerability patterns.

Expected outcomes include identification of structural vulnerability networks in CADASIL, definition of lesion topographies associated with mood, cognitive, and motor decline, and creation of predictive imaging markers and thresholds for clinical progression. This work will advance mechanistic understanding of vascular neurodegeneration and support patient stratification for therapeutic trials. Insights may extend to sporadic cSVD, contributing to biomarker development and personalized treatment strategies.

The ideal candidate holds a Master’s degree in neuroscience, biomedical engineering, applied mathematics, or data science. Required competencies include foundational knowledge in statistics and linear modeling, experience with neuroimaging tools (e.g., SPM, FSL, ANTs), familiarity with Python, MATLAB, or R, and motivation to work with large datasets and quality-control procedures. Additional desirable skills include mixed-effects and survival modeling, machine-learning methods for imaging data, version control, and experience with high-performance computing. Training will be provided for specialized tools and advanced statistical methods.

The student will join a leading neurovascular research group within a major neuroscience institute, engaging with neurologists, imaging scientists, data analysts, and clinicians. The environment offers strong mentorship, structured doctoral training, interdisciplinary seminars, and opportunities for national and international collaborations.

Expected deliverables include high-quality scientific publications in neuroimaging and neurology journals, a validated lesion-vulnerability atlas for CADASIL, predictive models of clinical progression based on lesion topology, and presentations at major scientific meetings. Overall, this project offers a rare opportunity to develop advanced computational neuroimaging expertise in a clinically meaningful context, contributing to precision neurology and future treatment strategies for small-vessel disease.

Neuronal differentiation is critically regulated through alternative splicing and, reflecting this central role, splicing deregulations and mutations in splicing factors are strongly linked to neurodevelopmental disorders (NDDs).
The host group searched for in vivo partners for the essential splicing factor U2AF2 using proximity labeling, which validated known partners and identified novel ones. Intriguingly, the resulting list of 286 putative U2AF2 partners shows a dramatic enrichment in NDD-related genes (SFARI database, odd ratio 7.54). We hypothesize that misregulations of U2AF2 by its partners are broadly responsible for splicing defects in NDDs. The thesis project aims at characterizing the U2AF2-centered NDD-related network of interactions using cutting-edge bioinformatics tools coupled to biochemical and functional analyses to uncover previously unrecognized mechanisms of splicing regulation and their involvement in NDDs.

State of the art
Neurodevelopmental disorders (NDDs), including autism spectrum disorders (ASD), intellectual disabilities (ID), attention-deficit/hyperactivity disorder (ADHD), and developmental language disorders, are highly heterogeneous and genetically complex. They involve genes related to synaptogenesis, ion channels, cell adhesion, chromatin regulation, and cytoskeletal dynamics. Many of these genes are dosage-sensitive, with de novo mutations frequently contributing to pathology. Recent studies highlight that mutations impacting the RNA splicing machinery are also recurrent in NDDs.
RNA splicing is achieved by a megadalton machine, the spliceosome, comprising a diversity of proteins and small RNAs to ensure the “single nucleotide precision” of the reaction and avoid the translation of aberrant and possibly harmful proteins (1). The assembly of this machine begins with the recognition of small sequences splicing signals at the extremities of the introns. The ribonucleoprotein U2snRNP is recruited to the branchpoint sequence, an essential step that requires the U2 Auxiliary Factor (U2AF). This factor binds the pyrimidine tract and the 3' splice site, through its large subunit U2AF2 and its small subunit U2AF1 respectively. The host group has been involved in the characterization of interactions surfaces between splicing factors during these early steps of spliceosome assembly (2–5). In particular, they documented the interaction of SF1 with the branchpoint sequence and U2AF2 that prepares the recruitment of U2snRNP on the intron, through interaction with the U2snRNP subunit SF3B1 (2, 6). Interestingly these studies unraveled variations around a recurrent mode of interaction where folded U2AF-Homology Motifs (UHM) accommodate small tryptophan-containing peptides (UHM ligand motif, ULM) (5).

Still, the entire interaction network governing correct splice site choice remains incompletely understood, even though its therapeutic relevance is highlighted by mutations in multiple genetic diseases, including Vernej syndrome (mutation of the U2AF2-related factor PUF60), Nager syndrome (SF3B4 mutations), Fragile X Syndrome (indirect splicing defect through MBNL1 deficiency), RETT syndrome and Autism Spectrum Disorders (including U2AF2 mutations (7)).
Altogether, the mutations in NDDs often affect U2AF2 itself or related proteins. To identify U2AF2 partners, the host group recently performed proximity labeling in HEK293 cells, confirming known partners and identifying novel ones. Biochemical analyses such as co-immunoprecipitations could confirm selected novel interactions of U2AF2. Gene ontology analyses also supported the accuracy of the BioID approach, as they revealed that most U2AF2 partners belong either to four major classes of splicing factors, chromatin remodelers, 3' end processors, and RNA modification factors. BioID with a truncated mutant of U2AF2 lacking the RS dipeptide-rich low complexity domain (RS) revealed its global involvement for U2AF2 interactions and functional experiments showed the regulatory action of its phosphorylation. Motif searches revealed also the presence of 14 novel possible UHM-ULM contacts among U2AF2 putative interactions. Finally, RNA-seq analyses highlighted that U2AF2 and its RS domain are crucial for splicing exons flanked by short introns in a nuclear localization-dependent manner (Pankivskyi et al, Nucleic Acid Research, in revision).

Then, given the above mentioned connections of U2AF2 with NDDs, statistical analyses of the enrichment of disease-related genes among the list of 286 putative U2AF2 partners have been performed. This revealed a dramatic overlap with NDD-associated genes of the SFARI database combining high confidence genes and syndromic genes (38 genes, OR 7.54).

Hypotheses and objectives
Based on the literature and our statistical analyses, we hypothesize that deregulations of the U2AF2-centered NDD-related network are broadly responsible for splicing defects in NDDs. In this thesis project, to extend our knowledge of NDD-related U2AF2 partners, in vivo proximity labelling will be applied to the human neuronal-like cell line SHSY-5Y in proliferation and after RA-induced differentiation. Cutting edge sequence-based bioinformatics tools dedicated to the analyses of protein networks will boost the identification of interactions surfaces and the consequences of disease mutations on U2AF2 interactions. This work will guide biochemical analyses of this U2AF2-centered network and functional analyses of its impact on splicing and neuronal differentiation of neuronal cell lines and human iPSC as a closer model of human neuronal differentiation. This research project will reveal new pathways connecting genetic mutations to splicing alterations, providing insights for targeted therapeutic approaches aimed at modulating splicing to improve patient outcomes.

Methodology
The work plan divided into four work packages (WP) is detailed below in terms of work tasks (WT).
WP1) Extension of the U2AF2 interaction networks (BioID experiments in SHSY-5Y cells)
Deliverable 1: Identification of additional neuronal specific U2AF2 partners relevant for NDDs.
WP2) Bioinformatics and structural analyses of the protein networks centered on U2AF2
- WT2-1 In-depth data mining of diseases databases (ClinVar, SFARI, Gene4Denovo…) to extract all variants affecting the putative U2AF2 partners in NDDs.
- WT2-2 Bioinformatics-based prediction of interaction surfaces between proteins of the networks and the effect of mutations (collaboration with Elodie Laine, IBPS, Paris).
WP3) Biochemical and microscopy analyses of the novel interactions
- WT3-1 Protein-protein interactions will be characterized using co-immunoprecipitation, pulldown assays, and colocalization analyses and PLA.
- WT3-2 Site-directed mutagenesis will be used to experimentally validate predictions on interaction domains, and the impact of PTMs and disease-associated mutations.
Deliverable 2 (from WP2 and 3): A documented and validated U2AF2-centered NDD network.
WP4) Functional analyses (to be tailored to the current knowledge and emerging insights regarding each protein)
- WT4-1 Splicing analyses: knockdown of validated networks component followed by qPCR/RNA-seq and alternative splicing analyses. The functional link with U2AF2, splice sites sequences and gene architecture will be explored by correlation analyses.
- WT4-3 Impact of network components on neuronal differentiation (SABNP lab for SHSY-5Y and collaborative work for hiPSC).
Deliverable 3: Functional impact of novel U2AF2 partners on splicing and neuronal differentiation.

By mapping the NDD-specific U2AF2 interaction network, this project aims to reveal previously unrecognized mechanisms of splicing regulations and to identify new molecular targets for therapeutic development.

1. Wilkinson, M. E., Charenton, C. & Nagai, K. RNA Splicing by the Spliceosome. Annu. Rev. Biochem. 89, 359–388 (2020).
2. Loerch, S., Maucuer, A., Manceau, V., Green, M. R. & Kielkopf, C. L. Cancer-relevant splicing factor CAPER? engages the essential splicing factor SF3b155 in a specific ternary complex. J. Biol. Chem. 289, 17325–37 (2014).
3. Manceau, V. et al. Major phosphorylation of SF1 on adjacent Ser-Pro motifs enhances interaction with U2AF65. FEBS J. 273, 577–87 (2006).
4. Wang, W. et al. Structure of phosphorylated SF1 bound to U2AF65 in an essential splicing factor complex. Structure 21, 197–208 (2013).
5. Loerch, S. & Kielkopf, C. L. Unmasking the U2AF homology motif family: a bona fide protein-protein interaction motif in disguise. RNA 22, 1795–1807 (2016).
6. Tari, M. et al. U2AF65 assemblies drive sequence-specific splice site recognition. EMBO Rep. 20, e47604 (2019).
7. Li, D. et al. Spliceosome malfunction causes neurodevelopmental disorders with overlapping features. J Clin Invest 134, e171235 (2024)

Our research project aims to advance the understanding of how brains without a cerebral cortex can generate higher-order cognitive abilities, such as problem-solving and imitation. Outside of primates, these abilities are observed in a few species of birds, such as parrots and crows. By combining behavioral studies and MRI techniques, this thesis project will examine the brain organization underlying these high cognitive abilities, using medium-sized parrots as a model. Thanks to the collaboration of NeuroPSI and NeuroSpin, a parrot brain atlas is currently under construction. As a next step, we will perform functional MRI (MEMRI and BOLD fMRI) to identify the brain areas involved in puzzle-box opening (problem-solving with manipulation) and imitation. Our hypothesis is that mirror neurons-like activity supporting imitation can be detected in brain areas associated with the task performance in parrots, as has been suggested in primates.

Abilities of problem-solving and imitation form the foundation of complex behaviors such as tool-use and culture transmission. We are interested in understanding how these abilities have independently emerged multiple times during evolution. Higher-order cognitive functions underlying such behaviors have mostly been investigated in humans and other primates, which possess a large neocortex. However, some birds such as parrots also display remarkable problem-solving and imitation abilities, despite lacking a six-layered neocortex.
Comparative neuroanatomy between the avian and mammalian brains has revealed that the functions and neural circuitry of the avian pallium (dorsal telencephalon) are surprisingly similar to those of the mammalian cortex, although their morphology is very different (Stacho et al., 2020). Importantly, the functional similarities arose independently in mammals and birds through convergent evolution. While the basic organization of the avian pallium is conserved across bird species, certain pallial areas – particularly associative areas – are especially elaborated in parrots and crows (von Eugen et al., 2020; Ströckens et al., 2022), much like in primates. Comparing parrot and primate brains may therefore reveal shared organizational principles underlying complex cognition.

By combining anatomical and behavioral analyses, the thesis project aims to characterize brain areas involved in problem-solving and imitation in medium-sized parrots. We chose the green-cheeked conure (Pyrrhura molinae) as a parrot model species, which offers an optimal balance for both anatomical and behavioral investigations. As a behavioral paradigm, we have developed multistep puzzle-box tasks that require sequential manipulations to solve, thereby engaging complex problem-solving skills. These behavioral tests will be combined with two different functional MRI studies: Blood-Oxygenation-Level-Dependent (BOLD-fMRI) and Manganese-Enhanced MRI (MEMRI), to visualize the brain areas activated during task performance and its imitation.
The behavioral tasks will be performed in NeuroPSI under supervision of Kei Yamamoto, and MR imaging will be performed in NeuroSpin under supervision of Luisa Ciobanu.

Aim 1) Behavioral trainings of problem-solving with sequential manipulation in parrots:
It has been clearly shown that different species of parrots are capable of performing problem solving tasks (Chen et al., 2019; O’Neill et al., 2021). We classify manipulation abilities hierarchically: i) pushing ii) grasping iii) pulling apart, and iv) twisting. For example, pushing, involving a single contact with an object, is widespread across animals. In contrast, twisting such as unscrewing is extremely rare, requiring strong grip, precise coordination, and persistent goal-directed action, and it can be observed only in large-brain species like primates and parrots. To assess these capacities in various species, we have developed multistep puzzle-box tasks that require animals to remove a series of locks in sequences in order to obtain a food reward. This paradigm can be used to evaluate problem-solving and cognitive flexibility in animals (Estienne et al., 2025). The originality of our puzzle-box design lies in its modularity and adaptability: 1) the number of “pushing locks” can be adjusted to suit different species and 2) additional screw-like “twisting locks” can be inserted, which must be rotated to remove. Using this paradigm, we can compare sequential manipulation ability across vertebrates, including humans. Based on our preliminary studies, our parrots can perform up to four-step puzzle-box openings. Furthermore, unlike most birds other than corvids, the parrots can be trained performing twist-opening movements.
Using this puzzle-box paradigm, our first goal is to test the parrot’s problem-solving abilities involving sequential manipulation. The second goal is to establish these behaviors as a bases for the functional MRI studies in Aim 2 (MEMRI) and Aim 3 (BOLD-fMRI).

Aim 2) Identification of brain areas activated during problem-solving tasks in parrots using MEMRI:
In collaboration with NeuroSpin, we have previously established a MEMRI protocol revealing brain regions involved in object manipulation in cichlid fish (Estienne et al., 2025). A modified version of this protocol will be applied in parrots. Birds will receive an intraperitoneal MnCl? injection (50 mg/kg) 24 hours before the behavioral task, allowing manganese uptake in active neurons during task performance. Birds will interact with different types of puzzle-boxes, including tasks requiring screw removal, which demands sustained, goal-directed manipulation suitable for manganese accumulation. After repeating the behavioral tasks, birds will be anesthetized and scanned using a 17.2T preclinical MRI to obtain high-resolution (100 µm isotropic) T1-weighted images. Neural activity will be compared between the puzzle-box performing group and a control group to identify brain regions associated with problem-solving manipulation behavior.

Aim 3) Brain activation during action observation in parrots using BOLD-fMRI:
In humans and non-human primates, “mirror neurons” are activated both when an individual performs an action and when observing another individual performing the same action. These neural circuits are thought to contribute to imitation (Gallese et al., 1996; Rizzolatti and Arbib, 1998). In monkeys, these neurons are located in the premotor cortex (area F5), which is associated with actions like grasping and object manipulation. In humans, a comparable system within Broca’s area has been linked to imitation processes involved in language production. Mirror neurons have been studied almost exclusively in primates, but it remains unclear how such systems evolved.
An intriguing hypothesis is that the neural circuits underlying language originated as an extension of the premotor networks for hand manipulation in ancestral primates (Thibault et al, 2021). Parrots demonstrate high imitation ability – both in speech and actions –, nonetheless, the presence of mirror-neurons have not yet examined. Thus, parrot studies offer an interesting model of the evolution of imitation.
To this end, we will develop BOLD-fMRI for awake parrots. Bird’s head will be fixed in the 17.2 T MRI scanner, while watching videos of conspecifics performing puzzle-box tasks (Aim 1).
We expect activation of “parrot mirror-neuron” during action observation. These activation sites will be mapped and compared with those identified in Aim 2 to assess the relationship between areas involved in puzzle-box manipulation and its observation. This approach will provide a unique opportunity to decipher the brain circuitry responsible for imitation, which has so far only been studied in primates.

Overall, this project bridges comparative neurobiology and advanced imaging, deepening our understanding of how complex cognition evolved across distant vertebrate lineages.

The view of the meninges as ‘simple’ envelopes of the central nervous system has recently been challenged by the discovery of several previously unknown functions, thanks to the study of specific cell populations, such as resident macrophages and meningeal fibroblasts, and specific circulatory networks, such as meningeal lymphatic vessels. Meninges are now known to play a role in various diseases and a better understanding of their structure and function could therefore be useful in improving the treatment of these conditions. The aim of this project is to identify the progenitor cell of meningiomas, the most common brain tumours, using an atlas of somatic mosaicism and normal meningeal cells within the adult human meninges. This project could inform both the origin of meningiomas and the function of meninges.

Meninges were once considered as purely structural support for the CNS and protection from trauma. Recently, new functions of the meninges have been uncovered, most notably in brain development and homeostasis, and they have been involved in a variety of diseases. The recent « rediscovery » of the meningeal lymphatic vessels also adds to the structural complexity of the meninges, unravels new interactions between resident immune cells and the peripheral immune system, and questions the mechanisms of immune surveillance of the CNS. The meninges are a complex structure composed of three structurally and cellularly distinct layers: the pia, arachnoid, and dura mater, with different cell populations. Most notably, meningeal fibroblasts which were first considered as purely structural cells, are now known to execute a variety of functions, such as the regulation of neighboring blood vessels, immune cells, and lymphatic vessels, through the production of growth factors, cytokines, and extracellular matrix remolding.

Meningiomas are the most common central nervous system tumors in the population of age 35 and older. They originate from the meningeal coverings of brain and spinal cord. The two most frequent genes involved in meningioma tumorigenesis are NF2 (chromosome 22q) and TRAF7 (chromosome 16p). Meningiomas with either NF2 gene mutation and/or loss of Chromosome 22q represent 50% of all benign sporadic meningiomas, are overrepresented at the convexity, while meningiomas with TRAF7 mutations, either alone or in association with mutations of the PI3K pathway (PIK3CA, AKT1) or KLF4, represent ~25 % of all benign sporadic meningiomas, are predominantly found at the skull base.

Several authors have suggested that those two mutually exclusive meningioma tumorigenesis pathways might correspond to two different developmental origins of the meninges. As mentioned, those two meningioma subgroups are characterized by different locations and different histological subtypes, but also different immune environments and different risks of tumor progression.

According to the 2021 WHO classification, about 80% of meningiomas are slow-growing grade I benign tumors. Atypical grade II meningiomas represent 15% to 20% of meningiomas, and 1-3% percent of all meningiomas are grade III tumors and behave as true malignant neoplasms. Up to 80-90% of WHO Grade II-III meningiomas harbor NF2 mutations while WHO Grade II-III meningiomas with TRAF7-PI3K mutations are almost inexistant. NF2-mutant and TRAF7-mutant meningiomas also differ in terms of immune environment, as NF2-mutant tumors are characterized by an increased number of immune cells, essentially macrophages, compared to TRAF7-mutant tumors. Interestingly, the number of macrophages significantly decreases in higher-grade meningiomas on a similar NF2-mutant background, but no additional genetic or epigenetic events appear to drive higher-grade meningioma tumorigenesis. One hypothesis would be that aggressive meningiomas are not linked to a specific mutational landscape but to a specific cell of origin.

Finally, we have shown that normal meninges already harbored mutations in meningioma driver genes, NF2 and TRAF7, and that those mutations were mainly found in the dura mater. These observations suggest the existence of different spatially-restricted progenitor cell populations in the meninges capable of generating different meningioma subtypes with different immune interactions and specific tumorigenesis mechanisms, influencing tumor progression.

We believe that identifying such populations, which would constitute the project main objective, would not only inform the mechanisms of meningioma initiation but also tumor progression towards malignancy and therefore help for the development of new therapeutic strategies for aggressive meningiomas, which represent an understudied disease and an unmet medical need, without any available medical treatment for surgery and radiation-refractory cases.

This project could be relevant to inform meninges structure and function and therefore open research perspectives for our understanding of the immune microenvironment but also the neurovascular interfaces regulating this tissue. The discovery of a progenitor cell population in the meninges could also open perspectives to treat different related diseases such as brain trauma.

The first step of our project will be to draw an extensive map of these putative progenitor cell populations. We already performed our first exploratory study using targeted sequencing in 90 meningeal samples of 5 individuals. To complete this first analysis, we will therefore perform an extensive map of somatic mosaicism in human meninges. With the collaboration of the NeuroCEB network in Pitié-Salpêtrière hospital, we will perform whole exome-sequencing (WES) (500X) of 300 meningeal samples (iGenSeq Sequencing platform, Paris Brain Institute), from 15 individuals, spanning all types of meninges and locations, including cranial and spinal meninges, with both arachnoid and dura mater. In cranial meninges, we will sample skull base and convexity meninges independently to look for a specific mutational landscape that could mirror the repartition of meningiomas along the cranio-spinal axis in patients. We will therefore be able to demonstrate the frequency of such clones, types of mutations, and preferential location among meninges.

To map those newly discovered mutant clones to specific cell populations, we will need to perform an extensive spatial single cell atlas of human meninges, illustrating the complexity of this structure. We plan to collect 30 human meninges samples from 6 donors with specific spatial locations for the five samples analyzed per patient (perisinusal dura mater along the sagittal sinus / convexity dura mater / skull base dura mater / arachnoid mater in the perimesencephalic cisterns /arachnoid and pia mater from convexity regions).

The creation of such an atlas would be facilitated by two kinds of local resources: high-quality meningeal tissues and pre-existing bio-informatics data. The meninges are characterized by a very low number of cells and the isolation of a significant amount of living cells to perform single-cell transcriptomic analysis represents a first challenge. With our first study, we already demonstrated our ability to obtain sufficient DNA from fresh-frozen meninges to perform ultra-deep targeted sequencing and therefore believe we have all the required skills to obtain a sufficient amount of living meningeal cells in order to perform the single cell sequencing, owing to the good quality of tissues. We benefit from the collaboration with the tissue bank NeuroCeb at Pitié-Salpêtrière hospital, which will help us gather large amounts of freshly resected post-mortem pia, arachnoid and dura-mater. To conduct the bio-informatic analysis, we will take advantage of the publication of single cell data on mouse meninges which may be of interest despite inter-species variations, but also fetal human meninges to perform a first sorting of the different cell populations in human meninges.

Our first objective will be then to characterize the different populations of meningeal fibroblasts, their transcriptional differences, then to determine if there is a common progeny inside the meninges consisting of progenitor cells and finally if there are regional differences between those fibroblasts that might reflect the differences of embryologic origin of the meninges, which may themselves mirror the regionalization of mutational profiles in meningiomas.

We will then perform cell-type-informed genotyping of somatic mutations using an approach similar to GO-TEN (Genotyping Of Transcriptome Enhanced with Nanopore sequencing), which was recently published by S. Bizzotto. This allows to map single-cells genotyped for the mutant loci of interest onto single-cell transcriptomic data and obtain the identity of the mutant cells.

Multiple sclerosis (MS) is an autoimmune demyelinating disease of the central nervous system (CNS). The role of innate immune mechanisms in MS onset and progression remains unclear. We hypothesize that trained immunity—an enhanced, memory-like state of myeloid cells—contributes to the persistent inflammation underlying MS. This project will investigate how trained immunity shapes innate immune activation and influences disease course. Using multi-omics analyses of patient-derived monocytes and macrophages, combined with functional assays, we will determine whether epigenetic and metabolic reprogramming typical of trained immunity occurs in MS. In vitro and in vivo models will assess how these reprogrammed cells affect CNS inflammation and degeneration. Our goal is to reveal how trained immunity connects environmental triggers to chronic immune activation in MS, potentially identifying novel innate immune–targeted therapeutic strategies.

Background and Rationale:
Multiple sclerosis (MS) is a complex autoimmune disease characterized by inflammation, demyelination, and neurodegeneration. While the adaptive immune system has traditionally been viewed as the primary driver of MS pathology, emerging evidence suggests that innate immune memory—or trained immunity—may significantly contribute to the chronic inflammatory environment that sustains lesion activity and neurodegeneration in MS. (1) Trained immunity is mediated by long-term metabolic and epigenetic reprogramming of innate immune cells, leading to persistent pro-inflammatory responsiveness (2).
Our preliminary data (3) and recent literature indicate that monocytes and macrophages from MS patients exhibit metabolic shifts, altered chromatin accessibility, and cytokine hyperresponsiveness—hallmarks of a trained state, as reviewed in our recent article (4). These features may arise from environmental exposures such as viral infections (e.g., Epstein-Barr virus) or metabolic stress (e.g., Western diet), potentially perpetuating the inflammatory imbalance that impairs remyelination.

Project Foundation:
We have access to a unique biobank of peripheral blood mononuclear cells (PBMCs) from MS patients with diverse clinical profiles. These samples have already been characterized, providing a robust foundation for our mechanistic and translational studies. Leveraging this resource, we aim to elucidate the role of trained immunity in MS and explore its therapeutic reversibility.

Objectives:
This project will employ an integrative approach combining multi-omics, functional assays, in vitro modeling, and in vivo validation to:
Reconstruct and manipulate trained immunity states in vitro using co-culture systems.
Investigate the in vivo relevance of trained immunity and its therapeutic reprogramming in models of demyelination.

Specific Aims:

Aim 1: Reconstruct and Manipulate Trained Immunity States In Vitro
We will develop human in vitro co-culture models incorporating monocytes/macrophages, lymphocytes, and glial cells to mimic the inflammatory and regenerative environments of MS lesions. Using environmental triggers such as ?-glucan, EBV antigens, or metabolic stressors, we will induce and monitor trained immunity-like states. Functional outcomes—including cytokine production, phagocytosis, and oligodendrocyte precursor cell (OPC) differentiation support—will be assessed. Epigenetic and metabolic inhibitors will be employed to test the reversibility of these states.
Expected Outcome: Experimental evidence demonstrating that trained immunity-like reprogramming in monocytes/macrophages can be induced, modulated, or reversed, and that these states directly influence glial repair mechanisms.

Aim 2: Investigate the in vivo Relevance of Trained Immunity and Its Therapeutic Reprogramming
We will use focal demyelination models to validate trained immunity signatures identified in human cells. Through adoptive transfer of reprogrammed macrophages or pharmacological modulation of metabolism and epigenetics (e.g., HDAC inhibitors, itaconate pathway activators), we will test whether reprogramming trained immunity promotes remyelination and mitigates neuroinflammation.
Expected Outcome: In vivo evidence that targeting trained immunity mechanisms can restore immune homeostasis and enhance lesion repair in MS models.

Methods
We hypothesize that trained immunity contributes to the establishment and maintenance of the chronic inflammatory state driving MS progression. To test this, we will combine systems-level and experimental approaches to:
Identify molecular and epigenetic signatures of trained immunity in monocytes and macrophages derived from our biobank of MS patient PBMCs.
Determine how these reprogrammed cells influence neuroinflammatory processes using in vitro co-culture systems and in vivo models of demyelination.
Explore how environmental factors associated with MS risk—such as infections or metabolic stress—may trigger or amplify trained immune responses.

Our multidisciplinary strategy integrates immunology, neurobiology, and computational modeling to bridge patient-derived data with mechanistic experimentation. By elucidating how innate immune memory contributes to MS pathophysiology, this project will uncover new links between peripheral immune activation and central nervous system damage. Ultimately, these insights will pave the way for innovative therapeutic strategies aimed at reprogramming or preventing maladaptive trained immune states, offering new avenues to halt or reverse disease progression in MS.

Conclusion
This project is uniquely positioned to advance our understanding of trained immunity in MS, thanks to our access to a well-characterized PBMC biobank from MS patients. By leveraging this resource, we aim to translate mechanistic insights into therapeutic opportunities, ultimately improving outcomes for individuals living with MS.

(1) Franklin RJM, Simons M. CNS remyelination and inflammation: From basic mechanisms to therapeutic opportunities. Neuron. 2022 Nov 2;110(21):3549-3565.
(2) Netea MG, Quintin J, van der Meer JW. Trained immunity: a memory for innate host defense. Cell Host Microbe. 2011 May 19;9(5):355-61.
(3) Fransson J, Bachelin C, Ichou F, Guillot-Noël L, Ponnaiah M, Gloaguen A, Maillart E, Stankoff B, Tenenhaus A, Fontaine B, Mochel F, Louapre C, Zujovic V. Multiple Sclerosis Patient Macrophages Impaired Metabolism Leads to an Altered Response to Activation Stimuli. Neurol Neuroimmunol Neuroinflamm. 2024 Nov;11(6):e200312.
(4) Popa E, Cheval H., Zujovic V. Cues of Trained Immunity in Multiple Sclerosis Macrophages. Cells, 2025 Jul 10;14(14):1054.

The Bridge project will be to develop the next-generation hybrid Brain-Machine Interface (BMI) combining functional ultrasound imaging (fUSI) with electrical and behavioral signals to achieve high performance and minimal invasiveness. BMIs and Deep Brain Stimulation (DBS) have advanced treatments for neurological disorders like Parkinson’s and epilepsy, but face issues of invasiveness and long-term stability. fUSI offers high spatial resolution and low invasiveness by measuring brain activity through neurovascular coupling. By integrating fUSI with EEG, ECoG, and behavioral markers (pupil size, saccades, heartbeat), the project aims to build a hybrid BMI (HBMI) with improved decoding precision. Its goals include creating novel fUSI devices for real-time 3D brain imaging and developing multimodal computational frameworks that merge neural and behavioral data, paving the way for safer, more robust neurotechnologies.

Invasive brain-machine interfaces (BMIs) provide the highest information bandwidth and temporal precision by recording extracellular activity from or near individual neurons. Penetrating microelectrode arrays (MEAs), such as the Utah arrays, are considered the gold standard for clinical and experimental single-unit recordings. These devices offer sub-millisecond temporal resolution and sub-millimeter spatial selectivity, allowing for precise decoding of movement intentions and speech motor commands. Contemporary research systems can now record from thousands of neurons simultaneously using high-density silicon probes like Neuropixels. These probes feature thousands of recording sites on a flexible shank, enabling stable chronic recordings in behaving animals (Steinmetz et al., 2021). The landscape for invasive BMIs now includes several complementary modalities. Subdural electrocorticography (ECoG) captures mesoscale field potentials with greater spatial and temporal resolution than electroencephalography (EEG), facilitating robust decoding of speech and motor functions while improving biostability. Depth electrodes and deep-brain stimulation (DBS) leads expand BMI control to deep brain structures, supporting closed-loop stimulation approaches. A novel hybrid method, the endovascular (stent-electrode) array, utilizes cortical blood vessels to place recording contacts intravascularly, potentially offering a lower-risk option for chronic implantation (Oxley et al., 2023).

The goal of this PhD will be to develop a novel hybrid and multimodal system that offers partial mitigation. This system will combine behavior metrics such as pupil size and saccade movements with EEG signals (including SU, MU, and LFPs) and cerebral blood volume (CBV) measurements. By merging electrical and hemodynamic information, we can enhance the system's robustness. Additionally, we will integrate peripheral bio-signals, including pupil size, eye tracking, and heartbeat, to increase control bandwidth.

Altogether, the project is focused on exploring and developing a next-generation hybrid Brain Machine Interface (BMI) framework in a preclinical setting, focusing on high performance and minimal invasiveness (Deffieux et al. 2021). This will be done by combining high-resolution functional ultrasound imaging (fUSI) with electrical and behavioral signals. BMIs are transforming the possibilities for patients with severe neurological disorders, enabling direct high-resolution communication with the brain and closed-loop modulations. These technologies hold immense promise for restoring lost functions - such as movement, speech, and sensory perception - and treating conditions like Parkinson’s disease, epilepsy, paralysis, and neuropsychiatric disorders. However, their clinical success depends on addressing key challenges, including minimal invasiveness, long-term stability, and the accuracy of neural signal decoding (Griggs et al. 2024).

This PhD project aims to elucidate the novel molecular and cellular mechanisms underlying secretory ferroptosis in human neurons and glial cells. It focuses on the elimination of iron and peroxidized lipids via VAMP7-dependent endolysosomal secretion and its role in Parkinson’s disease (PD) pathogenesis. Using human iPSC-derived dopaminergic neurons, astrocytes, and microglia with CRISPR-Cas9-mediated VAMP7 knockout, the project will dissect how impaired secretion affects ferroptosis, alpha-synuclein aggregation, and intercellular communication. By integrating multi-omics (proteomics, lipidomics, metabolomics), advanced cell biology, and functional assays, this work aims to uncover new therapeutic targets for PD and advance understanding of neuronal-glial interactions in neurodegeneration. Raphaël Rodriguez (Institut Curie), an expert in ferroptosis, will contribute as a collaborator.

I. Context, positioning, and objectives
This project seeks to uncover the molecular and cellular mechanisms underlying the elimination of iron and peroxidized lipids via endolysosomal secretion in neurons and glia. Iron accumulation and lipid peroxidation are key contributors to neuronal vulnerability in Parkinson’s disease (PD) (Wise et al., 2022). We hypothesize that dysregulated iron clearance via late endosomal secretion contributes to neurodegeneration, and that modulating this process could yield novel therapeutic strategies.
Classically, proteins are secreted through the ER–Golgi pathway, while unconventional protein secretion (UPS) involves autophagosomes, late endosomes, or lysosomes (Vats and Galli, 2022). The vesicular SNARE protein VAMP7 mediates endolysosomal secretion in several cell types, including astrocytes, and has been implicated in the release of autophagy- and ER-related proteins. Our previous proteomic studies identified several LC3-interacting and ER-phagy-associated proteins (e.g., RTN3, pro-VGF) secreted via this pathway in neuronal models (Filippini et al., 2023; Wojnacki et al., 2020), suggesting a connection between VAMP7-dependent secretion, cellular stress responses, and Parkinsonism.
Recent findings indicate that neurons export peroxidized lipids and iron through a VAMP7-dependent mechanism (Ralhan et al., 2023). Impairment of this secretory pathway leads to iron and lipid peroxide accumulation, sensitizing cells to ferroptosis—a regulated form of necrotic death driven by iron-dependent phospholipid oxidation (Liang et al., 2022). VAMP7 also mediates alpha-synuclein secretion (Xie et al., 2022), linking vesicular trafficking, oxidative stress, and PD pathology. Intracellular iron and alpha-syn are tightly interconnected, jointly contributing to dopaminergic neuron vulnerability and potentially driving a vicious cycle of toxicity in PD pathology (Wise et al., 2022). We therefore propose that a VAMP7-dependent secretory ferroptosis mechanism in neurons and glia regulates cellular iron and lipid balance, and its impairment promotes neurodegeneration.
________________________________________
II. Methodology and work packages
This interdisciplinary project combines cell biology, neurochemistry, functional genomics, and molecular neuroscience using human iPSC-derived models and mouse genetics.
WP#1 – Ferroptosis biomarkers in VAMP7-KO dopaminergic neurons and glia
1.a. Generation of VAMP7-KO iPSC models (year 1)
We will use the KOLF2.J1 iPSC line, known for robust dopaminergic differentiation, to generate VAMP7 double-knockout cells via CRISPR-Cas9 editing (in progress at the SCeNT platform in Marseille). These cell lines will be differentiated into dopaminergic neurons (DAn), astrocytes (AC), and microglia (MG) using established protocols (Nolbrant et al., 2017; Santos et al., 2017). As a backup, we will derive DAn, AC, and MG from conditional or full VAMP7 knockout mice. This will provide robust cellular models to study the impact of impaired endolysosomal secretion and autophagy on neuronal and glial ferroptosis.
1.b. Phenotyping of VAMP7-KO iPSC models (years 1–2)
Mutant and control DAn, AC, and MG will be characterized using specific markers by immunostaining (IF), RT-qPCR, and Western blotting (WB) as described in the literature (Nolbrant et al., 2017; Santos et al., 2017). Cell viability will be assessed using LDH in the medium as a marker. Further characterization will include assessing the endoplasmic reticulum (ER), Golgi apparatus, autophagosomes, endosomes, and mitochondria using IF with classical markers, as well as secreted unconventional proteins: RTN3, VDAC, UQCRC2, and VGF.
We will characterize alpha-synuclein expression and aggregation by WB and IF, and measure the presence of monomers, multimers, and fibrils in our iPSC models (Stojkovska and Mazzulli, 2021). Inflammation responses in AC and MG will be assessed to determine whether VAMP7-KO impairs cytokine and chemokine secretion. To this end, AC will be treated with IL-1beta, and MG will be treated with lipopolysaccharide (LPS). This detailed characterization will allow us to define the function of late endosomal secretion in neurons and glial cells deficient for VAMP7. Exploration of alpha-syn will provide insights into how these processes might alter the aggregation of this pathological protein.
1.c. Ferroptosis characterization (years 1–2)
Little is known about how dopaminergic neurons regulate iron homeostasis and limit the accumulation of peroxidized lipids, suggesting that dysregulation of these processes may contribute to their selective vulnerability in PD. Based on our lipidomics analysis in catecholaminergic PC12 cells (Wojnacki et al., 2020) and recent literature (Ralhan et al., 2023), we hypothesize that late endosomal secretion regulates neuronal ferroptosis.
We will quantify lipid peroxidation using BODIPY-C11 fluorescence by FACS and fluorescence microscopy. We will identify lipid membrane classes and the extent of their oxidation using mass spectrometry, and assess iron levels using ICP-MS (Institut Curie platform). We will evaluate the effects of various ferroptosis inhibitors, including radical-trapping agents (?-tocopherol), iron chelators (deferoxamine-DFO), and synthetic inhibitors such as ferrostatin-1 and liproxstatin-1, to test causality. GPX4 expression and activity will be measured to assess antioxidant defense.
These analyses will precisely define how VAMP7 loss sensitizes cells to ferroptosis. Raphaël Rodriguez will contribute his expertise in ferroptosis assays, lysosomal iron biology, and mechanistic studies.
________________________________________
WP#2 – Secretome analysis of VAMP7-KO neurons and glia
2.a. Multi-omics characterization (years 2–3)
We aim to delineate the contribution of the secretory pathway to secretion using VAMP7-KO DAn, AC, and MG. We will molecularly characterize secretomes using proteomics, lipidomics, and metabolomics on specialized platforms (Institut Curie, SFR Necker, and GenoToul, respectively).
While awaiting results of this multi-omics analysis, we will measure the amounts of RTN3, VDAC, UQCRC2, VGF (propeptide and VGF-derived small peptides), iron, and peroxidized lipids secreted by our neurons and glial cells. Additionally, we will characterize ?-syn expression by WB and IF to assess the presence of monomers, multimers, and secreted fibrils in the culture medium. Inflammatory cytokines and chemokines will be measured following IL-1beta or LPS stimulation to assess glial inflammatory responses to impaired VAMP7-dependent secretion.
2.b. Functional effects of secretomes (years 3)
We will investigate the potential toxicity of secretomes on control iPSC-derived DAn. We will collect culture media conditioned for 24 h by our DAn, AC, and MG. After removing cell debris by centrifugation or isolating exosomes, we will treat control DAn with secretomes and exosomes derived from WT, VAMP7-KO, and ATG5-KO DAn, AC, and MG.
We will then assess neuronal survival, dopaminergic marker expression, and mitochondrial function using Seahorse respirometry, as well as neuronal activity using multi-electrode arrays. To specifically evaluate the role of iron, we will deplete it from secretomes using established chelators (e.g., DFO, deferasirox-DFX, deferiprone). Parallel experiments will be conducted after stimulating or inhibiting ferroptosis in the donor-secreting cells. To this end, we will knock down ferroptosis suppressors, such as GPX4, FSP1, and xCT.
Finally, we will test whether specific cell types display particular vulnerability to ferroptosis inducers, including in-house lysosomal iron activators developed by Rodriguez's laboratory (e.g., ironomycin, fentomycin), as well as conventional inhibitors of ferroptosis suppressors (e.g., ML210, RSL3, Erastin, iFSP1).

This PhD project explores the emerging view of the cerebellum as a key structure at the interface between motor control and cognition. It investigates its domain-general role in learning through predictive signals, i.e., anticipating the outcomes of actions based on experience. We hypothesize that the cerebellum integrates multiple types of outcome information (e.g., performance and motivational feedback) to guide and progressively improve both motor and non-motor behaviors. Using a unique combination of high-density electroencephalography (EEG), transcranial magnetic stimulation (TMS), and rare intracranial recordings (iEEG), the project will provide the first direct evidence of cerebello–basal ganglia–cortical network dynamics during learning in humans. Ultimately, it aims to uncover the physiological mechanisms by which the cerebellum enables individuals to make predictions and learn from experience.

Whether predicting where a tennis ball will land or deciding whether to rest or go out the night before an exam, human behavior relies on learned expectations. Through experience, the brain refines these expectations by comparing what is predicted with what actually occurs. Discrepancies between predicted and observed outcomes drive adaptive adjustments that optimize future behavior. As our understanding of the environment improves, better predictions lead to more successful actions. Motivation is closely intertwined with this process, as learning is shaped by our ability to predict positive or negative consequences, guiding our actions.
Traditionally, predictive processing has been attributed to the prefrontal cortex and dopaminergic circuits. The cerebellum, by contrast, has long been considered a structure dedicated to the fine-tuning of movements, comparing predicted and actual outcomes to improve precision in real time. However, growing evidence challenges this narrow view. Several studies reveal extensive cerebellar connections with cortical and subcortical regions, including the prefrontal cortex and basal ganglia, which are key hubs for motivation and cognitive control [1]. Functional imaging studies, including some from our lab [2], along with recent animal work [3], indicate that the cerebellum contributes to predictive processes, generating anticipatory signals not only for movement but also for non-motor events such as reward. Yet, most studies have investigated motor and cognitive tasks in isolation, which makes the contribution of the cerebellum in predictive process across domains unclear. Furthermore, progress in understanding the precise mechanisms and temporal dynamics by which the cerebellum contributes to such processes in humans has been limited by both conceptual and methodological barriers: conceptually, the cerebellum has often been excluded from models of cognition and motivation; technically, traditional tools such as EEG and magnetoencephalography (MEG) struggle to capture cerebellar activity due to its deep and folded anatomy. With recent advances, including several developed in our lab [2,4], we can obtain direct neurophysiological evidence for cerebellar contributions to human learning through motivation.
In this PhD project, we hypothesize that the cerebellum predicts future motor and non-motor outcomes based on experience. Using a multimodal approach combining electroencephalography (EEG), intracranial EEG (iEEG) recordings and transcranial magnetic stimulation (TMS), we aim to show: (1) the cerebellum’s role in motivation-based motor learning, (2) its interaction with the basal ganglia, and (3) its contribution to predictive processes during reinforcement learning. Ultimately, this project will identify the physiological properties of cerebellar signals preceding action, revealing how they predict future outcomes and refine behavior.

Project Aims

Aim 1: To characterize cerebellar-to-cortec connectivity changes after motivation-based motor learning.
We will examine how the cerebellum processes motivational feedback (reward and punishment) to improve motor performance. While both reward and punishment influence the way individuals learn, previous work suggests that punishment drives rapid, trial-by-trial motor adjustments [5,6]. As evidence indicates that the cerebellum encodes aversive signals [7], we hypothesize that the cerebellum may process punishment-related feedback to update motor plans and refine movements. Specifically, we expect that the magnitude of oscillatory potentials evoked by cerebellar TMS over the cortex (cerebello-cortical connectivity) will increase following punishment-based motor learning, specifically for faster learners. The PhD student will analyze a pre-existing dataset combining EEG with concurrent cerebellar TMS, which will provide direct physiological evidence linking changes in cerebello-cortical communication to behavioral efficiency associated with punishment-driven learning.

Aim 2: To examine cerebello-basal ganglia communication during motivation-based motor learning.
While cerebellum and basal ganglia are traditionally part of separate motor loops, previous work challenges this view, suggesting functional interconnections that may underlie motivation-based learning [1]. We hypothesize that direct evidence for this communication will manifest as increased connectivity (coherence) between the cerebellum and a part of the basal ganglia, the globus pallidus (GP). Certain patients with Tourette's syndrome, characterized by dysfunction within cortico-basal ganglia loops, have implanted electrodes in the GP, offering a rare opportunity to record human basal ganglia activity. We will recruit these patients from our reference centre. The PhD candidate will lead concurrent intracranial and high-density EEG recordings while patients will perform a motivation-based motor learning task. This multimodal approach will provide direct evidence of cerebello–basal ganglia communication supporting motivation-based motor learning in humans.

Aim 3: To determine the role of the cerebellum in predictive processes supporting learning beyond the motor domain.
We will test whether the cerebellum’s predictive function is domain-general, extending beyond motor control to cognitive and motivational contexts. Instrumental learning, the process of associating actions with their outcomes, is traditionally attributed to cortico-striatal circuits. Cerebellar involvement, while observed, is not well understood. Using high-density EEG, participants will perform an instrumental learning task in which they learn to favor actions leading to reward and avoid those leading to punishment. We hypothesize that as participants learn, cerebello-prefrontal dynamics will generate a predictive signal that guides choices. Specifically, their communication preceding actions will progressively increase for the optimal choice as learning progresses. Identifying such a signal will broaden the cerebellum's contribution to predictive processes beyond the motor domain.

References
[1] Pavel, Gallea, et al 2017, Mov Disord; [2] Bracco […] Gallea. (aip), Curr Biol; [3] Gao et al, 2018, Nature; [4] Rohira […] Gallea, Bracco, 2025, Brain Stim; [5] Bracco […] Gallea, (in prep); [6] Galea et al, 2015, Nat Neurosci; [7] Moulton et al, 2011, J Neurosci.

Environment and supervision
The PhD will be conducted at the Paris Brain Institute (ICM), within the world-renowned Pitié-Salpêtrière Hospital. This environment provides an exceptional research ecosystem, integrating cutting-edge neuroimaging, electrophysiology, and stimulation facilities with clinical access. The student will benefit from state-of-the-art platforms for EEG, MEG, TMS, and MRI, supported by expert staff and access to relevant patient populations. The institute fosters a vibrant international community through the “Ajites” student association, promoting social and scientific exchange.
Supervision will be provided jointly by Dr Cécile Gallea and Dr Martina Bracco, whose complementary expertise will ensure full support to the student. Dr Gallea is specialized in cerebello–basal ganglia interactions using neuroimaging, while Dr Bracco is an expert in TMS, EEG, and human learning. The student will receive comprehensive training in experimental design, multimodal data acquisition, advanced signal processing, and scientific dissemination.

Animals and humans process vast amounts of information but retain only the most relevant experiences, with emotions playing a key role in memory consolidation. Emotions, particularly stress, enhancing attention and arousal, influence brain regions like the hippocampus. In neuroscience, it is known that hippocampal reactivations help recall and consolidate experiences. In artificial intelligence (AI), strategies inspired by these reactivations, especially in reinforcement learning, have been shown to accelerate learning. When it comes to emotionally charged circumstances, the relationship between stress, its most prominent hormone, cortisol, and memory functions in the hippocampus, is complex. This project aims to explore the link between stress and hippocampal reactivations via a computational model, and test it in artificial agents. This bio-inspired approach could (a) reveal new insights into the generation of hippocampal replay mechanisms and (b) improve AI's learning efficiency.

Most animals and humans encounter an immense amount of information and experiences throughout their lives, yet they manage to retain and recall only the most relevant, or a partial, but meaningful representation of them. Not everything experienced is stored in long-term memory; emotions play a key role in this memory consolidation process [Bower (1983) Phil. Trans. Royal Soc. London. B, Bio. Sci. 302.1110:387-402]. Emotions strongly influence cognitive processes, by enhancing our level of arousal and attentional resources, modulating the activity of brain regions such as the amygdala, the prefrontal cortex, and the hippocampus [Tyng et al., (2017) Front. Psychol]. Since the work of Scoville and Milner [(1957) Journ. Neurosc. 9.8:2907-2918], the role of the hippocampus as a cognitive mapping center has been studied in neuroscience, particularly through neurophysiological studies on spatial tasks in rodents. This research highlighted the fact that certain patterns of sequential activation of hippocampal neurons, observed during task execution, are then replayed during sleep or periods of calm wakefulness in the animal. These activities have been called hippocampal reactivations and are recognized as a powerful mechanism used, particularly by place cells, to recall, organize, and consolidate past experiences and infer future ones.
From an artificial intelligence (AI) perspective, it is also known that offline replay and updating of values associated with an agent’s actions can accelerate learning after a small number of real interactions with rewarding or punishing events (for example, Lin [(1992) Mach. Learn. 8-293-321]). The great interest in implementing computational strategies inspired by hippocampal reactivations in AI lies in tasks where past experiences and acquired knowledge must be re-evaluated and refined to perform better in future decision-making steps. This is the case with reinforcement learning (RL) paradigms [Sutton and Barto (2018) MIT press], where initially, when no prior knowledge is usually available, the best strategy is to interact with an environment through trial and error, and only when the level of experience increases the agent can exploit its previous knowledge to reach a sequence of actions approaching optimal behavior. In mammals and rodents, this consolidation of knowledge does not depend solely on the animal performing the same actions in the same situations: memories, particularly targeted recalls of experiences, are fundamental for effective learning from a small set of accumulated real experiences. Since the first RL algorithms which exploit strategies inspired by hippocampal reactivations [Sutton (1990) Mach. learn. proceed.], several researchers have proposed RL-based computational models inspired by these neuroscientific findings that are capable of reproducing the major experimental results regarding the quantity and type of reactivations generated [Khamassi and Girard (2020) Bio. Cybern. 114.2:231-248].
One of the major forces that impact the emotional state is stress, induced by the interaction between the animal, other animals, and the environment, both perceived through the animal’s senses. The influences from an external stressful environment and positive social interactions have been recently modeled by Khan and Cañamero [(2022) Front. Rob. AI 9] as the interaction between two hormones, cortisol and oxytocin and their homeostatic balance. Concerning the role of emotional states related to stress and cortisol, Cañamero’s research has for example modeled their influence on learning and adaptation [Hiolle, Lewis and Cañamero (2014) Front. Neurorob. 8], behavior development [Lones, Lewis and Cañamero (2017) IEEE Trans. Cogn. Devel. Sys. 10.2-445-454], pain perception [L’Haridon and Cañamero (2023) ACII], and the appearance of compulsive behavior in Obsessive-Compulsive Disorder [Lewis, Cañamero and Fineberg (2019) Comp. Psych.; Lewis and Cañamero (2019) ACII]. Regarding the role of anxiety and stress on the hippocampus memory mechanisms, many key issues still need to be investigated: it has been known from various animal and human studies that stress impairs many memory functions at the level of the hippocampus [Kim, Pellman, and Kim (2015) Learn. Mem. 22.9:411-416] but, recently, Sherman et al. [(2023) Journ. Neurosc. 43.43:7198-7212] found that cortisol could also enhance the hippocampal associative memory functions.
The thesis project aim at a better understanding of the relationship between stress and the generation of hippocampal reactivations by means of a computational model that could be systematically tested in simulation or eventually on a real robot. This will bring the great advantage of testing functional hypotheses we have about the relationship between stress and hippocampal replay, with our computational model, in a very controlled experimental set-up, when we could repetitively simulate different emotional profiles and sensitivities on our agents. An additional point to this research will be to observe and analyzed the proposed model embodied on a robotic platform with the aim to look systematically at what are the effects of such a model in a very controlled context that still present elements of stochasticity and unpredictability that make a robot interacting with the real world a step closer to animal experiments, compared to pure simulations.
As analyzed in Massi et al. [Front.Neurorob. 16], the adoption of RL techniques inspired by hippocampal reactivations for AI has just begun. After validating a strategy that combines offline reactivations generation through a model-based agent with reactivations generated by a model-free method, the question remains of how to optimize the timing of this offline reactivation generation and its quantity. So, the proposed project aims to link the generation of offline reactivations to the internal emotional state of an agent. The idea is that, by following the concept of homeostasis and the regulation of key emotion-related hormones, such as oxytocin and cortisol [Avila-Garcia and Cañamero (2004) SAB], in a learning task, the agent will change its internal emotional state in relation to (a) its performance in task completion (i.e., in a spatial navigation task, effectively avoiding punishments to quickly reach a reward state), (b) egocentric external stimuli (i.e., in a spatial navigation task, proximity to walls or other agents approaching), and (c) a combination of the above two elements. This emotional internal state will allow the agent to trigger reactivations during moments of intense stress, for example, and not at just any moment in the task, where they may prove unnecessary. With this bio-inspired approach to AI, we will test and validate optimal strategies for offline reactivation generation within RL algorithms, with the aim of improving and accelerating artificial learning and disclose new possible driving mechanisms for hippocampal reactivations in neuroscience.
So far, in RL, many strategies inspired by neuroscientific evidence on the hippocampus have been proposed and tested to have a spontaneous and optimal generation of different types of replay-like activities [Mattar and Daw (2018) Nat. Neuro. 21.11:1609-1617; Diekmann and Cheng (2023) ELife 12]. Still, a model that bases the generation of such reactivations on emotions and more specifically on the internal emotional state of an agent is lacking. That’s why the proposed project aims to theorize and test such a model which could spontaneously enable the generation of RL-based replay to improve memory consolidation and learning processing over different tasks and help shedding light on our understanding on the functional relationship between emotional states (stress in particular) and hippocampal replay. This research objective can be accomplished also thanks to the possibility of validate our results against experimental data from rodents that could be provided by our collaborators.

Our laboratory investigates how neuronal circuits control, memorize, and adapt movements in response to changing environmental conditions. Using in vivo two-photon imaging, electrophysiology, and optogenetics, we study how populations of neurons in the cerebellum and related structures contribute to rapid motor adaptation. Animals must constantly adjust their movements to external changes: from humans unconsciously adapting gait to heavy clothing to a gazelle evading a predator on slippery terrain. In a recent study (Nguyen et al., Nature Neuroscience, 2025), we recorded large neuronal populations while mice adapted their movements in a novel joystick-based task (link to video; https://bit.ly/3GmXWeX). This PhD project extends those findings and offers training in systems neuroscience, including advanced imaging and Neuropixels recordings in vivo to record in real-time how neurons in the brain control movement.

In rodents, motor and premotor cortical neurons send projections to the pontine nuclei, which relay information to the cerebellum via mossy fibers and to the deep cerebellar nuclei (DCN). DCN neurons, in turn, project to the thalamus, which feeds signals back to the frontal motor cortex, forming a bidirectional cerebello-thalamo-cortical loop (Guo et al., 2021). This architecture allows cortical motor areas to integrate multiple streams of movement-related information. While basal ganglia circuits are involved in action selection, cerebellar loops are thought to fine-tune the initiation, timing, and precision of movement execution.
Recent work has revealed that these loops are active throughout all stages of movement. During motor planning, cerebello-cortical interactions help sustain preparatory activity (Chabrol et al., 2019; Gao et al., 2018). At initiation, cerebellar output to the thalamus provides a temporally precise “go” signal that triggers cortical activity (Dacre et al., 2021). During ongoing movement, feedback from the cerebellum continuously refines trajectories, smooths kinematics, and corrects errors. With repetition, these interactions evolve into long-term adaptations, as error-driven plasticity reshapes both cerebellar and cortical networks. In short, the cortex and cerebellum operate as partners: the cortex encodes movement goals and learned programs, while the cerebellum handles prediction, precision, and error correction.
Despite this framework, key mechanistic questions remain open. How does the cerebellum encode “expected” sensory feedback from movement (expected reafference)? Where and how are mismatches between expected and actual sensory feedback detected? And how are these error signals used to update cortical motor commands?
Our team recently developed a novel behavioral paradigm that allows us to study these questions in mice performing voluntary forelimb movements with precisely controlled perturbations (Nguyen et al., Nature Neuroscience, 2025). The task is conceptually similar to classical “prism goggle” experiments in humans (Martin et al., 1996), where visual feedback is displaced and the brain must rapidly recalibrate its motor output. In our setup, mice manipulate a joystick under head-fixed conditions while visual or mechanical feedback is systematically altered. This design allows real-time measurement of adaptation while we record neural population activity with single-cell precision.
The PhD project will combine in vivo two-photon calcium imaging, Neuropixels electrophysiology, and optogenetic perturbations to dissect how cerebellar and cortical circuits interact during motor adaptation. The student will analyze how populations of neurons in both structures represent perceived versus expected sensory feedback, and how these representations evolve across adaptation.


Objectives and hypotheses
- Identify how cerebellar neurons encode expected sensory consequences of movement.
The hypothesis is that specific cerebellar cell types learn and store “expected reafference” through synaptic plasticity mechanisms that are activated by efference copies of motor commands.

- Determine how discrepancies between expected and actual sensory feedback drive adaptation.
When perturbations are introduced, prediction errors detected in the cerebellum are hypothesized to trigger changes both locally and in downstream motor cortical circuits.

- Characterize the bidirectional communication underlying forward and inverse models.
In this framework, the cerebellum implements a forward model predicting the sensory outcome of ongoing movements, while the cortex maintains an inverse model updating future motor commands. The interplay between these models ensures rapid error correction and long-term skill refinement.


Experimental approach
The project will involve three integrated components:

- Behavior: Training mice in the joystick-based adaptation task with controlled sensory or mechanical perturbations.

- Imaging and electrophysiology: Recording cerebellar and cortical activity using two-photon imaging and Neuropixels probes to capture cellular and network dynamics during adaptation.

- Manipulation: Using optogenetic activation or inhibition of specific cerebellar or thalamic populations to test causal contributions to motor learning.


Training environment
The PhD student will join a multidisciplinary team combining expertise in cerebellar physiology, motor cortex dynamics, and systems neuroscience. The project offers full training in advanced experimental techniques (surgery, imaging, electrophysiology, optogenetics, behavior design) and quantitative data analysis using Python and state-of-the-art pipelines.
Expected outcomes
This work will provide new insight into how cerebello-cortical interactions support rapid motor learning and adaptation at the cellular level. By elucidating the mechanisms linking error signals to cortical motor adjustments, the project aims to bridge the gap between theoretical models of internal forward/inverse computation and the actual dynamics of neuronal populations in vivo.

Dystonia/dyskinesia are disabling movement disorders with limited therapeutic options. Alterations of intracellular signalling pathways modulating cAMP turnover in striatal projection neurons is a key process in their pathogenesis. Pathogenic variants in the ADCY5 gene, encoding the main striatal adenylyl cyclase (AC5), cause mixed hyperkinetic movement disorders in humans. Modelling and investigating ADCY5-dystonia/dyskinesia thus represent a promising framework to find novel molecular targets for various hyperkinetic movement disorders. The PhD student will characterize the motor phenotypes associated with ADCY5 variants in the nematode C. elegans and test their rescue by pharmacological intervention. This model could represent a versatile and robust tool to assess the functional consequences of ADCY5 variants and perform drug screening targeting the modulation of the striatal cAMP turnover.

The pharmacological treatment of dystonia/dyskinesia is disappointing for most patients. Effective yet invasive surgical treatments, especially deep brain stimulation, are suitable only for a small subset of selected patients. Therefore, most patients with dystonia/dyskinesia have disability with a significant alteration of quality of life (doi: 10.1002/mdc3.13796). This emphasizes the need to investigate novel molecular targets. Movement execution is under the control of the striatum, the key input structure of the basal ganglia. In the striatum, 95% of the neurons are GABAergic striatal projection neurons (SPNs). The striatum controls movements via a subtle balance between the activity of two types of SPNs, namely the striato-nigral SPNs of the direct pathway and the striato-pallidal SPNs of the indirect pathway. The activation of the striato-nigral neurons and the striato-pallidal neurons facilitates and inhibits movement execution, respectively. Accumulating evidence of a functional interaction and biological convergence of multiple dystonia/dyskinesia genes expressed in SPNs suggests that synaptic activity and intracellular signalling of these neurons are critical for dystonia/dyskinesia pathogenesis. Indeed, several genetic defects causing hyperkinetic movements were linked to alterations of intracellular signalling pathways modulating cAMP turnover in the SPNs (doi: 10.1080/14737175.2021.1840978). Pathogenic variants in the ADCY5 gene cause mixed hyperkinetic movement disorders in humans, especially dystonia/dyskinesia. Interestingly, the encoded protein, adenylyl cyclase 5 (AC5), is highly expressed in SPNs and is the final enzyme in the striatal cAMP synthesis pathway. AC5 plays a central role in regulating the intracellular cAMP turnover and thereby in controlling SPNs activity level. Preliminary evidence shows that modulating intrastriatal cAMP metabolism can have a strong effect on motor phenotype in some patients with dystonia/dyskinesia (doi: 10.1002/mds.29006; 10.1002/mds.30170). Modelling and investigating specifically ADCY5-dystonia/dyskinesia thus represent a promising framework to investigate how impaired cAMP turnover in the striatum contributes to dystonia/dyskinesia more broadly, and how targeting this pathway can restore both biochemical and motor phenotypes. Moreover, new generation sequencing reveals increasing numbers of ADCY5 variants of uncertain significance in patients with movement disorders, leading to uncertain diagnosis. Current functional assessment of ADCY5 variants is restricted to cAMP production assay in heterologous cells. Thus, an alternative integrated model would improve diagnosis in ADCY5-dystonia/dyskinesia.
Our hypothesis is that the nematode Caenorhabditis elegans (C. elegans) could represent a versatile and robust experimental model to assess the functional consequences of ADCY5 variants associated with dystonia/dyskinesia in humans and investigate drugs or drug combinations targeting the modulation of cAMP turnover that could be used to treat dystonia/dyskinesia. We will characterize the phenotypes associated with ADCY5 variants in C. elegans and test pharmacological intervention to restore these phenotypes.
In C. elegans, all major neurotransmitter classes are present, and abnormal neuronal function can be correlated with simple behavioural outputs, such as spontaneous motor activity and its alterations due to various pharmacological exposures (doi: 10.1098/rstb.1986.0056; 10.1242/dmm.046110). C. elegans provides a powerful in vivo model to investigate neuronal cAMP signalling with direct translational relevance to human neurobiology. The relatively simple, fully mapped nervous system and strong conservation of key molecular pathways of this organism enable precise functional dissection of adenylyl cyclase activity at the molecular, cellular, and behavioural levels. The C. elegans genome encodes 4 adenylyl cyclases (acy-1, acy-2, acy-3, and acy-4) each with distinct expression patterns and biological functions (doi: 10.1895/wormbook.1.75.1; 10.1093/emboj/17.17.5059). The cAMP pathway modulates acetylcholine release from ventral cord motor neurons, providing a direct readout of neuronal signalling. We focus on acy-2 because it is predominantly expressed in neurons and plays a central role in regulating neuronal cAMP signalling and locomotor behaviour. Importantly, acy-2 shares close functional and structural similarity with mammalian AC5, enabling a translational approach. The availability of acy-2 mutants, deficient in a neuronal adenylyl cyclase orthologous to human AC5, offers the opportunity to model disease-associated variants in a living organism. Taking advantage of the worm’s genetic tractability, short generation time, and quantifiable motor phenotypes, we will define the functional consequences of human ADCY5 mutations within an intact neural network. We will then express in these lines the wild-type human AC5 - or different disease-causing variants of human AC5 - under a pan-neuronal promoter to investigate their functional effects in neurons in vivo.
ADCY5-dystonia/dyskinesia has a broad clinical spectrum and the response to treatment varies widely from one patient to another (doi: 10.1002/mds.29006). It is not known whether this heterogeneity is linked to the genotype, especially to the gain or loss of function characteristic of the variant. Cell studies have indicated that several missense variants implicated in dominant forms of ADCY5-related disorders result in gain of function (doi: 10.1016/j.bcp.2019.02.005). In contrast, the functional consequences of variants associated with recessive forms have been poorly studied, and a large repertoire of missense or in phase-deletion variants still need to be studied. In the C. elegans strains expressing the various ADCY5 pathogenic variants we will perform a various locomotor tests including crawling assay in solid media to analyze spontaneous locomotion (trajectories and movement speed) and a swimming assay in liquid media to evaluate the sensitivity to the paralysis induced by levamisole (nicotinic acetylcholine agonist acting on post-synaptic muscle receptors), aldicarb (inhibitor of acetylcholine esterase), or dopamine (stimulation of dopamine release-induced paralysis) (doi: 10.1093/hmg/ddab296; 10.3791/59243). The potential morphological abnormalities of motor GABAergic and cholinergic neurons will also be analyzed in transgenic lines where the network is labeled with GFP reporter. Analysis of these variants will provide a better understanding of the functional consequences of ADCY5 pathogenic variants on cAMP turnover. This will also offer an opportunity to evaluate a possible link between functional consequences (gain or loss of function) and the phenotype or the therapeutic response of the patients. Recent investigations indicate that modulating intrastriatal cAMP metabolism using A2A antagonists (caffeine, istradefylline, theophylline) can have a strong effect on the motor phenotype in some patients with ADCY5-dystonia/dyskinesia (doi: 10.1002/mdc3.13067; 10.1002/mds.29006; 10.1002/mds.30170). We will then investigate whether caffeine and other modulators of this pathway would restore the motor phenotype in the C. elegans strains expressing the various ADCY5 pathogenic variants.
This project will set up a new animal model to study dystonia/dyskinesia with quantitative approaches of behaviors. Adding this C. elegans model to the ongoing human, mouse and cellular studies, the PhD student will participate in deciphering the genotype/phenotype correlations in ADCY5-dystonia/dyskinesia. Our findings will i) provide a functional test that could help the diagnosis in patients with variant classified “of uncertain significance” ii) refine targeted interventions for various cAMP signalling defects iii) deliver strong proof-of-concept for striatal cAMP turnover as a therapeutic entry point in various hyperkinetic movement disorders.

Alzheimer’s disease (AD) is the most common form of dementia and inevitably leads to death. Current treatments are mainly based on enhancing acetylcholine (ACh) levels in the brain, but do not provide long-term amelioration. Human and animal model data indicate that interactions between the cholinergic neuromodulatory system and the amyloid beta peptide play a key role in the pathogenesis of AD. However, how this interaction contributes to AD pathology remains largely unknown. In a recent study, we showed that nicotinic acetylcholine receptors (nAChRs) expressed by distinct GABAergic interneurons are differentially inactivated by the amyloid-beta during AD progression and provided a possible mechanistic explanation for why clinical trials targeting specific subunits of nAChR genes failed. Here, we propose to dissect how cholinergic neurotransmission affects cortical networks during cognitive behaviors in early stages of AD, with the aim of identifying novel therapeutic strategies.

The neocortex is critical for cognitive processing with a central role in executive functions, sensory perception, motor ability and goal-directed behavior. Cortical circuits are composed of two major neuron types: excitatory pyramidal cells (PCs) that project to long-range target areas and local inhibitory interneurons (INs). Inhibitory GABAergic INs orchestrate cortical networks and operate a finely tuned division of labor within the circuit that is characterized by their complex connectivity patterns [2]. Due to their large diversity, INs contribute greatly to the brain’s computational power along with their role of preventing runaway excitation.
Layer 1 INs are found within the only cortical layer that lacks excitatory cells, although this layer contains the apical tufts of dendrites from supragranular and infragranular PCs [3]. This layer is often referred to as “elusive” because its composition and functional role have been difficult to identify. Recent evidence suggests that disinhibitory mechanisms involving layer 1 GABAergic cortical INs represent key cellular components for behavior [3].
Activity of neocortical INs is tightly controlled by endogenous acetylcholine, an essential neuromodulator, which is released from basal forebrain cholinergic projections that are diffusely localized throughout the cortical mantle [4]. Acetylcholine neuromodulation supports cognitive functions such as learning, memory and attention. Anomalies in cholinergic modulation of GABAergic INs can have wide-ranging effects such as disruption of synaptic plasticity, alterations of the excitation/inhibition balance and eventually the emergence of psychiatric and neurodegenerative diseases, such as Alzheimer’s disease (AD). However, little is known about the cholinergic recruitment of layer 1 INs during cognitive behaviors and their contribution to the network hyperexcitability that is observed in Alzheimer’s disease.

This project aims to address the following key questions:
1. What is the functional role of layer I INs in PFC during cognitive behaviors?
2. How are layer I INs recruited by cholinergic signaling during behavior?
3. How is cholinergic modulation of layer I INs affected in Alzheimer’s disease?
Understanding the role of distinct inhibitory neuron subtypes in PFC function will provide fundamental insights into the cellular mechanisms of brain disease.

Methods
Our team uses multiscale experimental approaches, combining genetics, two-photon calcium imaging in awake mice, behavior, slice electrophysiology and optogenetics.


1 Koukouli, F. et al. (2025) The alpha7 nicotinic acetylcholine receptor mediates network dysfunction in a mouse model of local amyloid pathology. Molecular Psychiatry. 2025 Sep 23. doi: 10.1038/s41380-025-03241-4
2 Lourenço, J. et al. (2020) Synaptic inhibition in the neocortex: Orchestration and computation through canonical circuits and variations on the theme. Cortex. 132, 258–280
3 Schuman, B. et al. (2019) Four Unique Interneuron Populations Reside in Neocortical Layer 1. J. Neurosci. 39, 125–139
4 Hartung, J. and Letzkus, J.J. (2021) Inhibitory plasticity in layer 1 – dynamic gatekeeper of neocortical associations. Curr. Opin. Neurobiol. 67, 26–33
5 Ballinger, E.C. et al. (2016) Basal Forebrain Cholinergic Circuits and Signaling in Cognition and Cognitive Decline. Neuron 91, 1199–1218

Neurodevelopmental disorders, including autism spectrum disorder (ASD), affect nearly 15% of children in industrialized nations. Besides synaptic dysfunction, an enhanced inflammatory state, both centrally and peripherally, has been established as mechanistic hallmarks of the condition.
We study the newly described autism-like disorder, Okur-Chung Neurodevelopmental Syndrome, OCNDS, that linked to de novo variants in the CSNK2A1 gene coding for the catalytic subunit of the kinase CK2. We hypothesize that neuroinflammation is not only present in OCNDS patients but that it is in part responsible for disease symptoms which range from developmental delay, intellectual deficiency, stereotypies, to epilepsies, sleep disorders and more.
We have developed a mouse model of the condition, harboring a patient hotspot mutation. We will use primary glial and neuronal cultures derived from these mice, perform confocal imaging, functional biochemical assays and cell-type specific proteomics to cha

CSNK2A1, is among 25 kinase and phosphatase genes, that were identified through genome sequencing and GWAS as highly penetrant monogenic ASD genes. CSNK2a1 encodes the catalytic ? subunit of casein kinase 2 (CK2), a highly conserved tetrameric serine/threonine protein kinase, composed of two ? and two ? regulatory subunits, that is abundantly expressed in the brain. It plays a critical role in brain development and physiology and has been implicated in neurodegenerative diseases and glioblastoma. Knockout mice lacking CK2? or CK2? subunits exhibit embryonic lethality, emphasizing the essential role of this kinase. Our lab has demonstrated that 16 different OCNDS-linked CSNK2A1 missense mutations significantly reduce CK2 activity in vitro. We have also characterized the first mouse model of OCNDS (CK2?K198R, based on a hotspot mutation occurring in one third of OCNDS patients), which phenocopies patient symptoms.
Neuroinflammation, characterized by the activation of microglia and astrocytes, has emerged as a key contributor to the pathophysiology of autism spectrum disorder (ASD). Postmortem studies and analyses of cerebrospinal fluid and blood reveal elevated pro-inflammatory cytokines and reactive glial states in individuals with ASD. Animal models further show that glial dysfunction can disrupt synaptic development, neural connectivity, and behavior. These findings suggest that chronic neuroinflammatory signaling may underlie core cognitive and behavioral phenotypes of ASD. Understanding the specific contributions of different glial cell types is therefore critical for dissecting disease mechanisms and identifying potential therapeutic targets.
We stipulate that alterations in microglial and astrocytic function and activity are, besides neuronal alterations, likely to contribute to neurological manifestations and neuroinflammation in OCNDS. In our mouse model of OCNDS, we have found evidence of peripheral inflammation, including elevated inflammatory cytokines in plasma, morphological changes in the gut indicative of heightened inflammatory status, and alterations in gut microbiome composition. In the brain, preliminary data suggest that microglia in specific brain regions exhibit a more amoeboid morphology, suggesting heightened activation.
The proposed project aims to characterize neuroinflammatory processes in the OCNDS mouse model in greater detail and to determine whether the resulting effects on behavioral phenotypes are mediated in a cell-autonomous manner within the brain.

Aim 1: Phospho and total proteomics analysis of glia and astrocytes
We will study the in vivo p- and total-proteomes in a cell type specific manner by utilizing a biorthogonal proteome tagging method, based on Cre-induced expression of a mutant tRNA synthetase that utilizes a derivative of methionine for integration into nascent proteins. We will express Cre specifically in microglia, astrocytes and neurons, cross these mice with the OCNDs mice, then use a click chemistry protocol to enrich for labeled hippocampal and cortical proteins which will subsequently be analyzed by mass spectrometry. This technique is already established in the lab. This approach will allow a clear delineation of differences in protein expression and phosphorylation status across distinct cell types, thereby uncovering the molecular mechanisms underlying the disease. While the lab has prior mass spectrometry data derived from our OCNDS mice, the newly generated results will give the cell-type specific resolution that is required to establish how strongly neurons versus microglia versus astrocytes are altered in the OCNDS brain and to determine if one type of glial cells is more affected than the other.

Aim 2: Characterization of the degree of neuroinflammation in OCNDS model mice
a) In vivo: We will analyze the ratio of activated versus non-activated microglia and astrocytes in hippocampus and PFC of juvenile and adult K198R het and Wt mice by flow cytometry. Due to our findings of reduced blood brain barrier capacity in OCNDS mice, we will also determine if an invasion of peripheral CD4+/CD8+ T-cells has occurred into the brain.
a) We will further use confocal imaging to count microglia and astrocytes in the cortex and hippocampus and reconstruct their morphology using Imaris software, in order to determine and confirm an alteration of the activation state of these cells. For this aim, preliminary data already exists that needs to be confirmed in more animals.
b) In vitro: we will study primary astrocyte, microglial and hippocampal neuronal cultures from mutant and wt embryos/mice, to recapitulate in vivo morphology studies (Sholl analysis, spine quantification for neurons) but also functionally by assessing their cytokine expression and cytokine release profiles, basally and upon a pro-inflammatory challenge.

Aim 3: Assess effect of K198R expressing astrocytes and microglia on neuronal activity.
We will perform in utero viral injections leading to microglia- and astrocyte-specific expression of mutant eGFP-labeled CK2??K198R versus CK2??WT protein under CX3CR1 - and Aldh1L1-promoters into wt and mutant embryos. Adult mice will undergo testing for cognitive and sensorimotor processing deficits as well as for stereotypies to determine if the CK2?? mutation in either glial cell type is sufficient to drive ASD-relevant behaviors. Glia cells expressing GFP and thus the overexpressed mutant or WT protein, will be compared in their morphology by confocal microscopy followed by image analysis.
This part of the project investigates the causal roles of specific glial cell types in ASD-related phenotypes. It further examines how wild-type and mutant environments affect the survival, migration, and morphology of microglia and astrocytes. These studies will clarify whether observed effects arise cell-autonomously or through neuron–glia interactions.

Taken together, this project investigates the effects of a patient-linked CK2 mutation on glial cells and aims to determine through which mechanisms alterations in these brain-resident immune cells influence the cognitive and stereotypic phenotypes of OCNDS mice. Understanding these mechanisms will provide important insights into the contribution of neuroinflammation to ASD-like disorders.

Spinal cord injury (SCI) is a major public health concern for which there is still no cure. Interluminescence allows for non-invasive, activity-dependent and synapse-specific modulation of neural circuits. Our hypothesis is that interluminescence neuromodulation of genetically targeted interneurons can improve outcome after SCI. Our first aim is to identify cross-species genetic markers of V1 and V2a interneurons, which are crucial in reorganizing spinal circuits after injury. Our second aim is to achieve targeted expression of interluminescent tools in these genetically identified interneurons using a single ConVERGD vector. Our last aim is to test the potential of interluminescence-mediated reinforcement of proprioceptive inputs to V1- or V2a-microcircuits in promoting locomotor recovery after SCI in a mouse model. This project will provide a new translational framework for a more “ecological”, self-regulating, approach to drive reorganization of neuronal circuits after SCI.

Context, positioning and objective(s) of the project

Genetically targeted dissection of spinal circuits. In genetically tractable animal models such as mouse and zebrafish, classification of spinal neurons is increasingly based on molecular markers related to their developmental lineage. Ventral spinal interneurons involved in motor control have been divided in four cardinal classes (V0 to V3) defined by the selective expression of distinct transcription factors(Knafo & Wyart, 2018). Among these, V1 and V2a interneurons are of particular interest regarding spinal integration of proprioceptive inputs and their putative role in EES-mediated rehabilitation after SCI.

V1 and V2a-dependent sensorimotor circuits are key for recovery after SCI. Combining snRNA-seq and spatial transcriptomics in a mouse model of SCI, Kathe et al. showed that Vsx2-positive V2a neurons in the lumbar spinal cord are engaged by electrical epidural stimulation (EES) and that their targeted ablation prevents EES-mediated recovery following SCI (Kathe et al., 2022). The fact that V1 clusters are also transcriptionally activated by EES after SCI corroborates the implication of V1 and V2a subpopulations in reorganizing spinal circuits after injury (Skinnider et al., 2020).

Selective manipulation of circuits with bioluminescence-optogenetics. Contrary to EES, which is by design a non-selective electrical stimulation, optogenetics, i.e. genetically targeted expression of light-dependent actuators, enables dynamic and cell-specific modulation of neuronal circuits. However, various limitations relating to actuators expression and light delivery hinder successful translation of optogenetics in human. Bioluminescence-optogenetics has the potential to circumvent many of these limitations by using a luciferase as a biological light source (Berglund et al., 2013). Of particular interest is interluminescence, in which the expression of the luciferase is targeted to pre-synaptic and the opsin to post-synaptic partners (Prakash et al., 2022), allowing for noninvasive, activity-dependent and synapse-specific neuromodulation of targeted circuits.

Our research hypothesis is that interluminescence-mediated reinforcement of proprioceptive inputs to V1 and/or V2a neurons can enhance reorganization of spinal circuits, thereby improving functional outcome after SCI.

Aim 1. Identifying conserved V2a and V1 spinal interneurons in the mouse and human spinal cord

Task 1.1. Cross-species profiling of V1 and V2a interneurons. Transcriptomic profiling of spinal neurons in the adult human is hampered by the difficulty of obtaining fresh human samples. We have been granted authorization to harvest spinal cords from brain-dead organ donors and collected human spinal cord samples from 15 different patients so far. We used fluorescence activated nuclei sorting (FANS) to enrich the neuronal yield and collected 56,010 spinal neurons from 6 patients, a ten-fold increase over the largest dataset of human spinal neurons currently published (Yadav et al., 2023). Using transgenic mouse strains to restrict the analysis to En1-positive V1 interneurons and Vsx2-positive V2a interneurons, we trained a machine-learning algorithm (XGBoost) to predict V1 and V2a homologous clusters among human spinal neurons.

Task 1.2. Morphological validation of candidate V2a and V1 clusters. To validate the biological relevance of neuronal subpopulations identified from transcriptomic profiling, we combined in situ hybridization of identified markers with immunolabelling in sparsely labelled V2a and V1 neurons using the MORF3 reporter line allowing for Cre-dependent expression of a membrane-localized smFPV5-GFP (Veldman et al., 2020). We are currently using leveraging 3D-samples of mouse and human spinal cords to achieve a cross-species morphological analysis (localization, projections, known markers) that will allow us to determine which V2a and V1 clusters are most likely homologous to known human cell types.

Aim 2. Achieving targeted expression of interluminescent actuators in PA-V1/V2a synapses

Task 2.1. AAV-mediated delivery of actuators to PA-V1/V2a synapses. In collaboration with Pr Ute Hochgeswender’s Lab at Central Michigan University, we have designed a single ConVERGD vector strategy to direct expression of the luciferase to PV- and Advil-positive proprioceptive neurons and of the opsin to either V1 or V2a interneurons (Figure 3). At P30, triple heterozygous transgenic mice (Advil-FlpO::PV-Cre::En1-Cre for V1 synapses or Advil-Flp::PV-Cre::Vsx2-Cre for V2a synapses), will be stereotaxically injected bilaterally with AAV virus carrying presynaptic luciferase and postsynaptic opsin constructs. The expected expression of the opsin and the luciferase will be assessed by fluorescence microscopy (EYFP and dTomato, respectively). All mouse strains are already available in the lab, and the vector is currently undergoing final validation at CMU.

Task 2.2. Validation of interluminescence-mediated activity. The luciferin (coelenterazine, CTZ) will be injected intraperitoneally at 10 mg/kg, while administered vehicle (saline) will be used for control experiments. Peak concentration of CTZ will be reached about 25 minutes after injection and lasts for 45 – 60 minutes (Ikefuama et al., 2024). To confirm neural activity at the V2a and V1 synapses, we will compare cFOS protein levels of EYFP-labeled neurons in animals receiving either the luciferase substrate (CTZ) or a placebo (vehicle) and undergoing a simple locomotor task such as a rotarod assay.

Aim 3. Probing interluminescent modulation at the PA-V1/V2a synapses after spinal cord injury

Task 3.1. Locomotor recovery following modulation of the PA-V1 or PA-V2a synapses in a SCI mouse model. Twenty days after AAV transduction adult mice (P50) will undergo a severe crush injury to the spinal cord at the level of T10 after laminectomy of a single. After a one-week recovery period, animals will undergo daily locomotor training sessions for a period of 7 weeks. Before each session, intraperitoneal injection of either CTZ or saline (control group) will be administered, and animals will be tested 20 minutes afterward. Animals will undergo locomotor tasks emphasizing the recruitment of proprioceptive afferents (i.e. horizontal ladder stepping, stepping over treadmill at increasing speeds). Mice will be trained for two weeks prior to spinal cord injury to continuously walk on a flat, straight runway and once a week during the recovery period. Leg kinematics will be captured by tracking reflective markers and parameters describing gait timing, joint kinematics, limb endpoint trajectory, and trunk stability will be computed for each gait cycle and analyzed using principal component analysis (Takeoka et al., 2014; Takeoka & Arber, 2019).

Task 3.2. snRNA-seq analysis of the subpopulations engaged by interluminescence-mediated rehabilitation. Mice will be euthanized 8 weeks after injury and the lumbar spinal cord will be extracted and immediately frozen. Upon mechanical tissue dissociation of spinal cord samples, neuronal nuclei will be enriched using fluorescence activated nuclei sorting based on NeuN expression. Single nuclei will then be encapsulated and sequenced (Chromium 3’). Leveraging our cross-species profiling of spinal interneurons, we will provide a multimodal characterization of engaged V1 and/or V2a subpopulations and examine whether they might be conserved in the human spinal cord.

Impact and significance

This project will provide a new translational framework for genetically targeted, synapse-specific, activity-dependent, and noninvasive modulation of spinal circuits to promote neurological recovery. Adopting a more “ecological” and self-regulating approach to drive reorganization of spinal circuits, our interluminescent strategy has the potential to overcome the main hurdles to¬ward successful targeted neuromodulation after SCI.

Human iPSCs offer a controllable, human-specific system to study early mechanisms predisposing to Alzheimer’s disease (AD). We will dissect how the human-specific gene CHRFAM7A — an endogenous modulator of the alpha7 nicotinic acetylcholine receptor — shapes neurogenesis and interneuron (IN) lineage specification, maturation and vulnerability to amyloid-beta (Abeta). We will combine genomically defined iPSC panels (0/1/2 CHRFAM7A copies) with isogenic CHRFAM7A loss-of-function and add-back, harmonized 2D differentiations, molecular and imaging phenotyping, and standardized A? challenges to map causal relationships between CHRFAM7A dosage, alpha7 signaling and AD-relevant phenotypes. In a second phase, optimized protocols will be translated to 3D neuro-organoids and ventral–dorsal assembloids to model neuron migration and integration in a tissue-like context. The outcome will be a rigorous, human-relevant framework linking CHRFAM7A/alpha7 signaling to early AD trajectories.

Background and central hypothesis.
Developmental programs in human cortex — NPC fate, interneuron (IN) subtype specification, maturation and integration — are increasingly implicated in the network imbalance that precedes clinical Alzheimer’s disease (AD). CHRFAM7A, a human-specific partial duplication of CHRNA7, modulates alpha7-nAChR signaling and Ca²+ handling, but its causal role during human neurodevelopment and Abeta vulnerability is unresolved. We hypothesize that CHRFAM7A dosage calibrates alpha7 signaling to tune NPC proliferation/differentiation/migration and MGE/CGE IN trajectories, thereby shaping early Abeta responses.
Aim 1 — Genomic definition of CHRFAM7A in human iPSCs.
We will assemble a high-quality iPSC panel spanning 0/1/2 CHRFAM7A copies (copy number will be quantified by ddPCR/qPCR). In parallel, we will generate isogenic CHRFAM7A knockouts from a 2-copy line using CRISPR/Cas9 genome editing that preserves the intact CHRNA7 allele. Edited cells will undergo single-cell cloning, followed by PCR and amplicon sequencing to confirm on-target disruption, and allele-aware RT-qPCR to verify loss of CHRFAM7A transcripts with retention of CHRNA7 expression. For genetic rescue, we will create a KO+add-back line by inserting dupalpha7 cDNA under a controllable promoter at a safe-harbor locus. All lines (0/1/2-copy, KO, KO+add-back) will pass unified quality controls (pluripotency markers, normal karyotype, mycoplasma-free), be banked, and then compared side-by-side under identical NPC/MGE/CGE differentiation protocols.
Aim 2 — 2D differentiations of NPCs and INs.
Using standard operating procedure suitable for comparative biology, we will derive anterior NPCs by dual-SMAD inhibition, rosette selection and expansion in defined growth factors; and INs via dual-SMAD induction to anterior neuroectoderm, followed by SHH-pathway activation with WNT modulation and rostro-caudal cues to bias MGE (FGF8) versus CGE (retinoic acid ± Activin A/FGF19). IN progenitors will be banked at day ~21–28 and matured with neurotrophic support and enforced post-mitotic exit. Identity, purity and maturation will be verified with compact immunofluorescence/RT-qPCR panels and image-based morphology (neurite length; Sholl-derived dendritic complexity). Synaptic development will be tracked with a minimal set of presynaptic/postsynaptic (Synaptophysin, PSD95) markers to provide objective integration indices. Correct synaptic development and in vitro assessment of spontaneous functional neuronal activity will be validated following microelectrode array (MEA) analysis.
Aim 3 — ?7/CHRFAM7A axis and Abeta vulnerability in vitro.
We will quantify CHRNA7/CHRFAM7A expression, alpha-bungarotoxin binding, and simple alpha7 pharmacology (agonists/PAMs ± MLA) across the genotype series (0/1/2 copy, KO, add-back) processed under identical methods. Developmental readouts will emphasize NPC neurogenesis and proliferation (EdU/Ki67), early neuronal differentiation and morphological maturation (maturation markers analysis). A standardized, quality-controlled Abeta oligomer challenge will then be used to measure neuronal survival and neurite/synapse integrity. Rescue experiments with ?7 agonists/PAMs and dupalpha7 add-back will test causality. The deliverable is a coherent, human-specific map of how CHRFAM7A/alpha7 signaling affects lineage specification, maturation and Abeta sensitivity.
Aim 4 (Perspective) — Step-up to 3D neuro-organoids and assembloids.
Optimized 2D protocols will be translated to 3D systems aligned with the laboratory’s models: dorsal cortical organoids (NPC excitatory scaffolds) and ventral MGE/CGE organoids (IN progenitors). Each differentiation will be carried out independently using its specific protocol, followed by fusion of the resulting tissues. The assembly of ventral and dorsal brain organoids into ventral?dorsal assembloids will enable modeling of human interneuron migration and spatial positioning within a physiologically relevant, tissue-like environment. We will analyze readouts: cell identity and morphology, migration trajectories, and synaptic markers, with optional transcriptomics at a decisive timepoint to capture alpha7-dependent state shifts. To further validate physiological relevance, these assembloids will be integrated in vivo through transplantation into neonatal mouse cortex. This will enable assessment of interneuron/neuron survival, maturation, circuit integration, and long-range connectivity under a vascularized and dynamically active environment. It will also extend observation windows for long-term development and disease-relevant dynamics.
Methods integration and analytics : All experiments will be run in biological replicates across independent differentiations and, where relevant, multiple clones per genotype. We will also investigate genotype/sex interactions and sex-stratified secondary analyses (female/male iPSCs lines). Mixed-effects models will account for batch/clone effects; false discovery rates will be controlled for multi-marker panels. Standard procedures and analysis of notebooks will be versioned, with blinded image quantification and predefined acceptance criteria for differentiation batches and Abeta preparations.
Expected outcomes and impacts.
The project will produce: (i) a resolved genomic/expressional atlas of CHRFAM7A in the specific iPSCs cell line; (ii) causal evidence that CHRFAM7A dosage/perturbation impacting NPC development faith and MGE/CGE IN trajectories via alpha7; and (iii) a human-relevant link between CHRFAM7A/alpha7 signaling and early Abeta vulnerability. By anchoring organoid/assembloid readouts to rigorously benchmarked 2D phenotypes, the work will deliver transferable endpoints for future target validation in early AD.

Amyotrophic lateral sclerosis (ALS) affects 97% of patients through TDP-43 mislocalization and aggregation. Accumulating evidence demonstrates that TDP-43 localizes to mitochondria and binds key mRNAs, impairing respiratory complex function. However, the complete mitochondrial transcriptomic landscape disrupted by disease-associated TARDBP mutations is undefined. This project will identify a convergent “MitoTAR signature”, a core set of mitochondrial RNAs disrupted across different ALS mutations. Long-read RNA sequencing of mitochondria from patient iPSC-derived motor neurons via Oxford Nanopore will allow mapping of the full mitochondrial transcriptome. Functional validation via RNA knockdown and rescue will establish causality between specific RNA targets and metabolic dysfunction. In vivo validation in TDP-43 zebrafish models will test therapeutic targeting of this signature using mitochondria-directed small molecules. This work will define a novel therapeutic node for ALS treatment

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder that affects both upper and lower motor neurons, causing disruptions in signal transmission from the brain to the muscles. These symptoms are fatal, leading to a median survival time following diagnosis of three to five years1. A pathological hallmark of ALS is the mislocalization and aggregation of TDP-43, which occurs in approximately 97% of all ALS patients2. TDP-43 is an RNA-binding protein that regulates splicing of critical genes like STMN2 and UNC13A, whose dysregulation causes downstream synaptic function deficits3–6.
Mitochondrial perturbation is a recognized feature of ALS pathogenesis. Recent evidence suggests that TDP-43 localizes to mitochondria where it binds mRNAs encoding respiratory Complex I subunits (ND3, ND6), causing complex disassembly and bioenergetic failure7. While progress of understanding TDP-43 mitochondrial mislocalization continues, several gaps remain. First, the functional consequences of TDP-43 mitochondrial localization remain narrowly defined8. Second, while mutation-specific differences in mitochondrial transport and nuclear transcriptomes have been observed9,10, a systematic link between individual mutations, their specific mitochondrial RNA targets, and their corresponding metabolic phenotypes is missing. Lastly, the causality gap remains unexplored. While blocking TDP-43 mitochondrial import can rescue toxicity, it is unclear if this localization is a primary driver of disease or a secondary consequence of cellular stress2,11,12.
Preliminary data from the Kabashi lab provides evidence supporting mutation-specific heterogeneity in patient iPSC-derived motor neurons. While one patient line recapitulates a previously reported glycolytic shift, two other patient lines display paradoxical increases in mitochondrial membrane potential (??m). Importantly, metabolic phenotype severity correlates with the quantitative burden of TDP-43 localized to the mitochondria, suggesting convergent RNA-level disruptions may underlie this heterogeneity.
Therefore, we hypothesize that the mislocalization of TDP-43 to mitochondria is a critical event that disrupts the mitochondrial transcriptome. We predict this disruption converges on a core set of mitochondrial RNAs, which drives metabolic dysfunction in motor neurons. Defining this convergent “MitoTAR signature” will establish the mitochondrial transcriptome as a novel and therapeutically actionable node in ALS.
Aim 1: Defining the Core Mitochondrial Transcriptomic Signature
Aim 1 serves as the foundation of this PhD project by utilizing in silico sequencing approaches to define the core mitochondrial signature of TDP-43 proteinopathy. We will isolate mitochondria from iPSC-derived motor neurons representing isogenic controls, TARDBP mutation carriers, and C9orf72 hexanucleotide repeat expansion carriers. Both mutations converge on TDP-43 dysfunction, as C9orf72 patients show impaired mitochondrial respiration and TDP-43 pathology. To enhance neuronal-specific signatures, we will include cortical organoid-derived neurons to provide an autonomous neuronal model complementing the motor neuron focus13. Mitochondrial RNA will undergo long-read sequencing via Oxford Nanopore PromethION, enabling full-length, single-molecule reads that capture complete transcript isoforms, a critical aspect given the role of TDP-43 in splicing regulation. Bioinformatic analysis will employ Seurat for clustering and differential expression alongside hdWGCNA to identify co-expression modules. We will analyze RNA processing, transcript abundance, and RNA editing patterns. Isoform-switching analysis will identify mutation-specific versus convergent alterations. Functional enrichment via Enrichr will map affected metabolic pathways. Top candidates will be validated using RT-qPCR and customized Nanostring nCounter codesets to generate a high-confidence “MitoTAR signature” atlas.
Aim 2: Establishing a Causal Role in Metabolic Dysfunction
This aim will determine the functional necessity of the top candidate RNAs identified in Aim 1. We will select 5-6 high-priority RNA targets based on differential expression strength and metabolic pathway enrichment. We will employ complementary loss-of-function (LOF) and gain-of-function (GOF) approaches in our iPSC-derived motor neurons. In our LOF approach, control motor neurons will be transfected with antisense oligonucleotides targeting candidate mitochondrial RNAs. We predict this will recapitulate patient metabolic phenotypes. In our GOF approach, patient motor neurons will receive expression constructs or ASOs to restore RNA function. We predict this will rescue metabolic dysfunction.
Functional outcomes will be assessed using established phenotyping platforms. Seahorse analysis will be utilized to measure mitochondrial respiration and glycolytic function, and live-cell imaging with TMRM will be utilized via IncuCyte to quantify changes in mitochondrial membrane potential. To confirm that the ASO-mediated rescue occurs without altering the percentage of mitochondria containing TDP-43, the Opera Phenix high-content imaging pipeline will be employed. This experimental design will definitely establish that observed metabolic dysfunction is a direct consequence of the specific RNA interaction instead of a secondary effect from reduced TDP-43 localization in mitochondria.
Aim 3: In vivo Validation and Therapeutic Testing
This aim will validate the pathophysiological relevance of the MitoTAR signature in vivo and test a direct therapeutic strategy. We will utilize a TDP-43 transgenic zebrafish model crossed with an Hb9:GFP line to fluorescently label motor neurons14–16. FACS-sorted spinal motor neurons will undergo RT-qPCR/ Nanostring to confirm MitoTAR signature preservation. We will then test therapeutic approaches utilizing mitochondria-targeted small molecules (e.g., SBT-272) or other compounds that modulate mitochondrial dynamics and reduce oxidative stress. We will utilize quantitative video tracking to measure the rescue of motor functions, such as swimming velocity17. The outcome of this aim will be in vivo validation of the MitoTAR signature and pre-clinical evidence for the efficacy of mitochondrial-targeted therapies in TDP-43-mediated ALS.
This investigation is paramount for defining a core pathogenic mechanism in ALS. By identifying the convergent “MitoTAR signature”, this work will move beyond a narrow focus on rare mutations to define a core pathogenic mechanism for TDP-43-driven metabolic failure relevant to the vast majority of ALS patients. Beyond fundamental mechanistic understanding, this research has significant translational potential. The MitoTAR signature itself would represent a novel set of therapeutic targets. By validating that direct mitochondrial support is an effective therapeutic strategy, this work will provide a robust rationale for new treatment opportunities for ALS and offer a promising alternative to targeting elusive upstream regulators.
References 1.Morgan, S. Br Med Bull 119, 87–98 (2016).
2.Suk, T. Mol Neurodegeneration 15, 45 (2020).
3.Melamed, Z. et al. Nat Neurosci 22, 180–190 (2019).
4.Klim, J. R. et al. Nat Neurosci 22, 167–179 (2019).
5.Brown, A.-L. et al. Nature 603, 131–137 (2022).
6.Ma, X. R. et al. Nature 603, 124–130 (2022).
7.Wang, W. et al. Nat Med 22, 869–878 (2016).
8.Izumikawa, K. et al. Sci Rep 7, 7709 (2017).
9.Fazal, R. et al. EMBO J 40, e106177 (2021).
10.Schweingruber, C. et al. Nat Commun 16, 4633 (2025).
11.Kawamata, H. et al. Molecular Neurodegeneration 12, 37 (2017).
12.Zuo, X. et al. Nat Struct Mol Biol 28, 132–142 (2021).
13.Mearelli, M. et al. Glia doi:10.1002/glia.70080.
14.Li, P. et al. Transl Pediatr 13, 1201–1209 (2024).
15.Stil, A. Devel Neurobio 76, 642–660 (2016).
16.Flanagan-Steet, H Development 132, 4471–4481 (2005).
17.Kabashi E et al. Hum Mol Genet 19(4):671-83. (2010).

Adequate sensory processing from birth is critical for socio-emotional development. Children who later develop Autism Spectrum Disorder (ASD) already show early sensory atypicalities before reliable diagnosis. Yet, the impact of these atypicalities on sociocognitive development remains largely unexplored. This PhD project aims to study early sensory development, focusing on tactile perception and multisensory integration in typical and atypical conditions. By combining high-density EEG and behavioral assessments, we will: (1) Characterize early brain responses to sensory stimuli in 2-3-mo typical infants; (2) Link these early perceptions to later socio-cognitive development at 10-12m; (3) Assess whether early sensory profiles can predict later atypicalities in infants at risk for ASD. At the intersection of fundamental and clinical research, this innovative, interdisciplinary project seeks to advance understanding of how early multisensory perception shapes socio-affective development.

Neurodevelopmental disorders (NDD) including autism spectrum disorders (ASD) affect an increasing number of children who are likely to need ongoing care over their lifespan. Yet the early detection of neural and behavioral atypicalities in infants at risk of developing NDD is still inadequate while essential to predict outcome and implement targeted interventions to take advantage of the intense neuroplasticity in the first months after birth.

a. Early Development and Sensory Processing During Infancy
Sensorimotor functions and their associated brain networks develop early but follow distinct maturational trajectories. Touch, one of the first senses to emerge in utero, plays a crucial role in shaping body representations and sensory integration through the perception-action loop. While sensorimotor activity facilitates the integration of environmental information, few studies have explored how the immature brain processes tactile stimuli, particularly in response to uni- vs. bilateral tactile stimulation. This gap is notable given that bilateral integration could serve as a marker of sensory processing abilities.
Additionally, multisensory processing—such as the integration of auditory and tactile information—remains understudied in infants. While audiovisual integration has been examined, only a handful of studies have focused on audio-tactile processing, revealing differences between preterm and full-term infants. Understanding these processes is essential, as early sensory integration may influence later cognitive and social development, particularly in infants at risk for neurodevelopmental disorders (NDD) like Autism Spectrum Disorder (ASD).

b. Development of Affective Touch and Social Cognition in Infants
From birth, multisensory processing is involved in emotional and social development, enabling infants to imitate others and engage in appropriate social behaviors. Affective touch, such as gentle caresses, plays a vital role in bonding, well-being, and emotional communication. This type of touch activates specific neurobiological pathways linked to social bonding and affective processing.
Studies suggest that social touch modulates infants' responses to sensory stimuli, with differences observed based on the source of touch (e.g., parent vs. stranger) and the type of stimulation (affective vs. neutral). However, the relationship between early sensory profiles—particularly tactile perception—and the development of affective touch and social cognition during the first year of life remains poorly understood.

c. Atypical Sensory Profiles in ASD
Infants and children with ASD often exhibit sensory atypicalities, including hyper- or hypo-sensitivities and impaired multisensory processing. These differences may stem from early alterations in brain connectivity and excitation/inhibition balance. While sensory processing differences are frequently reported in older children with ASD, early infancy remains an understudied period.
Recent studies suggest that atypical tactile processing as early as 10 months may predict later ASD traits, and infants at risk for ASD (e.g., siblings of diagnosed children) show altered responses to caregiver touch and multisensory stimuli. Exploring these early sensory profiles could provide critical markers for early diagnosis and inform targeted interventions to support developmental trajectories in at-risk populations.

Main objectives:
The overarching goal of this PhD project is to investigate sensory development and in particular early touch perceptions and multisensory integration under typical and atypical conditions; with three main subobjectives:
WP1. Characterize early sensory profiles in 2-3 mo typical infants based on brain responses to sensory stimulations (tactile/auditory modalities, uni/bilateral, uni/bimodal) and behavioral questionnaires. We expect that some sensory integration is already present at this early age in the tactile modality (as it is for auditory and visual modalities), that integration strength differs between a single (tactile) modality and two sensory modalities (tactile - auditory) and between a uni- vs bilateral tactile stimulation, with modulations by behavioral profiles (higher integration for infants with higher sensory sensitivities).
WP2. Link these early sensory profiles to later socio-cognitive and affective development through a longitudinal evaluation of typical infants at 10-12 mo. We will evaluate how affective touch modulates sensory neural responses, and characterize the development of social cognition at the behavioral level. Our hypothesis is that these sensory and socio-cognitive markers will be associated with relatively stable inter-individual variability between ages, making it possible to “predict” measurements at 10-12 mo from those at 2-3 mo.
WP3. Assess whether early sensory atypicality could serve as a marker for predicting future impairments in social cognition and affective perception. We will perform a proof-of-concept study on infant siblings of ASD children, followed in a clinical context and evaluated at 2-3mo and 10-12mo similarly to typical infants. We assume that ASD siblings, known to have higher risk to develop NDD, will show a higher inter-individual variability in neural and behavioral sensory profiles compared with typical infants, and that infants that will develop ASD or other NDD will show early atypical patterns in brain sensory responses in relation with later socio-cognitive and affective development, providing discriminant information for identifying infants at greater risk for social impairments.
Altogether, these 3 sub-aims will provide key evidence about the possibility to predict socio-affective development from early sensory perception.

We will recruit full-term infants and evaluate them longitudinally at around 2-3 mo (i.e. when the baby is expected to give smiling responses) and 10-12mo (i.e. after the key age of 9 mo for the development of social cognition, and affective touch perception), considering typical infants (WP1-2) and infants from family at high risk of ASD (WP3).
Participants: Inclusion criteria for the group of typical infants will be 2 to 3-month-old at the first visit (T1), to have no direct family antecedent of NDD, be born after 37 SA, and without known neurological history. Inclusion criteria for the group of infants at high risk of ASD will be 2 to 3-month-old at the first visit (T1), be born after 37 SA, and to have an older sibling with diagnosed ASD. Sample size: we expect to recruit 60 typical infants and 50 infants at risks of developing ASD, as we expect to have around 10-15% of drop out between T1 and T2 of testing for typical infants, while no drop out or below 5% is expected in the group at risks, given that they are followed at the hospital and that the follow-ups will be done in the same time as their medical appointment. In typical EEG longitudinal experiments, sample size is of 20-25 infants (e.g. Adibpour et al., 2020), while here some of our hypotheses are based on regression analyses, so we aim to double the sample.

At the interface between integrative neuroscience, developmental neuroimaging and pediatrics, this PhD project should have a strong impact for both fundamental questions about multisensory integration in infants and clinical impact for the “prediction” of cognitive, affective and social development over the first year.

This project investigates how striatal neural imbalance drives the emergence and persistence of compulsive behaviors and explores novel treatment strategies to restore circuit function. Using a genetically modified mouse model that enables testing our hypothesis of altered neurotransmission, we will examine links between circuit dysfunction and compulsive behavior and evaluate interventions to restore neural balance and improve behavioral control.
This project will have important theoretical and therapeutic implications given the central role of the dorsal striatum and dopaminergic transmission in the regulation of a wide range of behaviors. It aims to feed clinical research with new and innovative hypotheses validated in rodents in order to find new therapeutic strategies to help alleviate psychiatric illnesses with a compulsive component.

Context and Rationale
In everyday life, behaviors are governed either by conscious goal-directed actions or by automatic habitual behaviors. Goal-directed behaviors (GDB) are flexible and intentional, aimed at obtaining specific rewards or avoiding negative situations. In contrast, habitual behaviors (HB) are automatic responses to stimuli, independent of outcomes, and can persist despite punishment. Persistent HB are common in neuropsychiatric disorders like Tourette syndrome, obsessive-compulsive disorder, substance use disorders and eating disorders (EDs). Excessive development of habits has been proposed to underlie compulsive, maladaptive behaviors. It is therefore essential to understand the neurochemical mechanisms involved if we want to understand and treat major psychiatric disorders. mong compulsive diseases, anorexia nervosa (AN) is a deadly and life disrupting psychiatric condition with no current cure.
The dorsal striatum plays a key role in the transition from GDB to HB and compulsive behaviors. Striatal cholinergic interneurons (ChIs), though express the vesicular acetylcholine transporter (VAChT) and the type 3 vesicular glutamate transporter (VGLUT3) and therefore release both acetylcholine (ACh) and glutamate (glut). In addition, VGLUT3 enhances vesicular ACh accumulation and release through vesicular synergy. We previously showed that in mice, striatal ACh/glut cotransmission influences habit formation and maladaptive behaviors (Favier et al, 2024). Interestingly, ACh and glut have opposing effects on dopamine (DA) release in the nucleus accumbens (NAc) and the dorsomedial striatum (DMS), but not in the dorsolateral striatum (DLS).
We identified a missense polymorphism, VGLUT3-p.T8I, in patients with severe substance use disorders and EDs (Sakae et al, 2015). The VGLUT3-pT8I mutation in mice reduces vesicular synergy without affecting glut transports. This results in a decreased ACh release in the striatum and an imbalance of DA levels between the DMS and DLS. Consequently, VGLUT3T8I/T8I mice display excessive habit formation (Favier et al, 2024). This behavioral phenotype suggests a predisposition to compulsive-like behaviors, highlighting the VGLUT3T8I/T8I mouse as a powerful model for investigating the mechanisms driving compulsive disorders.
Endophenotypes of AN such as excessive physical activity and self-starvation can be partially modelled in mice with the activity-based anorexia (ABA) test. In this well-established rodent model, chronic undernutrition is associated with running wheel activity. We previously demonstrated the efficacy of donepezil in reducing self-starvation behavior in two hypocholinergic mouse lines: VAChTcKO mice and VGLUT3T8I/T8I mice (ip injections). In addition, we also reported the efficacy of donepezil for the treatment of AN patients. However, in human anorexic patients, donepezil is associated with several peripheral side effects, including cramps, nausea, and diarrhea. These side effects arise from ACh action at the neuromuscular junction. As a result, treating patients with anorexia nervosa (AN) requires low doses of donepezil (oral admin., 0.5–2.5 mg/day), which markedly slows the therapeutic process. One potential solution to mitigate these muscular side effects is to use brain-specific analogs of donepezil. Our collaborator N. Pietrancosta (chemist, CNRS) recently developed and protected several of these compounds such as Donquine. We will remain vigilant for possible side effects of the drugs used, especially any alterations in food intake.
We will first explore the neurochemical mechanisms of compulsion, using this pre-clinical model, and then the impact of our lead compound, Donquine, on ACh and DA release in striatal compartments of WT mice and VGLUT3T8I/T8I mice.

We hypothesize that reduced striatal cholinergic tone disrupts DA signaling in the DMS and nucleus accumbens (NAc), biasing behavior toward habits and compulsion. Restoring ACh tone should rebalance DA signaling and reverse compulsive phenotypes.

The PhD project has three main objectives and an additional perspective:
- Characterize striatal ACh and DA dynamics during the development of compulsive behaviors.
- Test the causal role of striatal DA imbalance in compulsive phenotypes.
Evaluate the therapeutic potential of ACh enhancement using donepezil and the novel brain-specific analog Donquine.

In parallel, we wish to refine this preclinical model of AN using a more ecological environment acquired through the BeCog project funded by DIM C-BRAINs.

Experimental Strategy:

1. Striatal DA and ACh dynamics during compulsion development
We will examine ACh and DA efflux in the DMS and DLS of WT and VGLUT3T8I/T8I mice using fiber photometry and genetically encoded biosensors (AAV9-hSyn-ACh3.0 and AAV9-hSyn-rDA3m). DA dynamics will be recorded for 30-min during the first and last day of habituation. Then, during the restriction phase, two daily 30-min recording sessions will be done, one overlapping the introduction of food and the second the restriction. We predict that an imbalance of DA tone in the DLS vs. DMS of VGLUT3T8I/T8I mice will accelerate the transition from normal to compulsive behavior in the ABA models.

2. Probing causal DA mechanisms
To test causality, we will manipulate DA tone in specific striatal compartments using chemogenetics.
In a first set of experiments, we will use DAT-CRE mice injected with retrograde AAV expressing inhibitory DREADD to reduce DA release specifically into the DMS. We predict that upon C21 local administration anorexic-like phenotypes will be precipitated in the ABA model. In a “reverse” experiment, we will use an excitatory DREADD to activate DA release in the DMS of VGLUT3T8I/T8I::DAT-CRE mice. We predict that C21 administration will alleviate self-starvation phenotypes in the ABA.
These two complementary experiments will validate the importance of the DA tone in the DMS for the development of compulsive behavior.

3. Pharmacological restoration of cholinergic balance
Donepezil, an acetylcholinesterase inhibitor (AChE-i), restores ACh tone and reverses compulsive grooming and self-starvation in VGLUT3T8I/T8I mice. Clinically, donepezil has shown efficacy in severely ill ED patients, though peripheral side effects limit its use. Our collaborator N. Pietrancosta (CNRS) has developed Donquine, a brain-selective AChE-i designed to minimize peripheral effects.
We will compare donepezil and Donquine in the ABA paradigm while simultaneously recording striatal ACh and DA dynamics. These experiments will reveal whether selective enhancement of central ACh transmission can normalize DA signaling and behavior.

We expect to demonstrate that:
- Reduced striatal ACh release drives DA imbalance between the DMS and DLS.
- This imbalance underlies the emergence of self-starvation and excessive activity in the ABA test.
- Restoring ACh tone rebalances DA signaling and rescues these behaviors.

Additionally, our lab will implement a home cage monitoring system allowing us to follow 24h a day small cohorts (4–6 mice) in a large, enriched home-arena for several weeks. Each mouse will carry a RFID tag. All key resources (food, water, wheels, nest zone…) are accessible, but food is dispensed at RFID-gated stations so we can quantify each individual’s intake, meal structure, and effort cost—without single-housing. This allows us to probe restriction, hyperactivity, cost-insensitive weight loss, anxiety-linked feeding suppression, social modulation of feeding, and cognitive inflexibility—constructs linked to anorexia.

In conclusion, this PhD project combines advanced neurochemical recording, behavioral modeling, and pharmacological innovation to dissect the striatal ACh–DA interplay driving compulsive behaviors. Its translational potential—bridging molecular mechanisms and clinical intervention—could pave the way for new treatments for new treatments for eating and compulsive disorders.

The PhD project is focused on theoretical development of the allostatic reinforcement learning (ARL). This framework proposes internal needs as coded through a space of drives, predictively govern motivation, thereby linking interoception and learning of allostatically oriented motivated behaviors. The candidate will develop neural theory for drive represention in the brain and how body states affect its structure. The project will analyse how the multiple interoceptive dimensions, reflecting bodily needs and sympathetic-parasympathetic activation, must be mapped to the lower-dimensional drive manifolds. The homeostatically regulated reinforcement learning theory from our team implies that the drive-space and the internal space representations should be separable. The project will develop an allostatic theory for introceptive encoding of motivations and rewards. Models will be used for approach-avoidance experiments from collaborators.

All living beings act on the environment to maintain the conditions of their existence, chiefly by regulating energy and material flows with the outside world. Beyond this “here and now” regulation, animals with a brain can forecast how external events (dangers and opportunities) will impact their needs in the future, at multiple temporal and spatial scales and use these predictions to guide adaptive behaviours. This predictive homeostatic regulation, or allostasis, requires the organism to blend complex information about its internal, physiological state and the state of the external world, combining the brain-body loop with the perception-action loop. Yet, while the neuroscience of interoception has made enormous progress in recent years, how the integration of body (internal) and world (external) signals drives behaviour and cognition is not fully understood. The project seeks to resolve this fundamental question by developing a novel neuro-theoretical account of allostasis in mammals.
The PhD project is focused on theoretical development of the allostatic reinforcement learning (ARL). This framework proposes how our internal needs, coded through a space of drives, predictively govern motivation, thereby linking interoception and learning of allostatically oriented motivated behaviors. In this project, the candidate will develop neural hypotheses for how drives are represented in the brain and how body states affect the structure of the drive-space. Notably project will analyse how the multiple interoceptive dimensions, reflecting bodily needs and sympathetic-parasympathetic activation, must be mapped to the lower-dimensional drive manifolds. The homeostatically regulated reinforcement learning theory, that has been developed in the Gutkin team, implies that the drive-space and the internal space representations should be separable. The project will develop an allostatic theory for introceptive encoding of the drive space and rewards. We will develop ARL-based models to determine the necessary drive manifold structures that explain learned behaviours and related neural activities, under competitive versus collaborative cross-interactions of drives. We will apply these models to approach-avoidance experimental data from our experimental collaborators.

Task1: Theory of drives and rewards.
We will examine two theoretical alternatives: either the multiple dimensions of the internal state space are convergently mapped as distances from the homeostatic set points to a unitary drive-manifold or mapping onto multiple independent manifolds as per ref.10. The two architectures yield distinct predictions: convergent unitary mapping predicts cross-modal interactions where manipulation of one interoceptive modality alters drive along others (e.g., hunger suppressing reward sensitivity or pain potentiating threat responses). Multiple independent manifolds predict autonomy of drives, with competition emerging only at the response generation stage. We will compare predicted behaviour from models with cross-modality interactions at the drive level to models where same interactions occur solely during response selection. By systematically varying which subsets of body variables converge onto which drive manifolds, we will characterize how interoceptive dimensionality reduction impacts motivated behaviour. This analysis will establish conditions under which different architectures yield equivalent versus distinguishable behavioural and neural signatures, to inform our experimental designs.

Task2: Theory of interoceptive interactions.
We will develop a novel theory for how autonomic state—particularly sympathetic/parasympathetic balance—modulates the drive-manifold architecture. We will formalize the induced allostasis hypothesis, that is, autonomic shifts alter the effective drive relative to anticipated state changes. One mechanism for this is that sympathetic activation remaps the interoceptive inputs and increases deviation from homeostatic optima while parasympathetic tone decreases such deviations—effectively shifting setpoints to adapt to different action contexts. Alternatively, autonomic balance may modulate sensitivity to homeostatic rewards and punishments, altering the steepness of drive gradients (i.e., the "looseness" versus "tightness" of homeostatic regulation). We will examine how these two options can be implemented through modulation of drive-manifold geometry—flattening versus steepening the manifold curvature.

Task 3: NeuroBehavioral Modelling.
We will develop HRRL models for the approach-avoidance tasks and the competitive interactions between the aversive thermal stimuli and the reward. The models will explain how body state modulation along the sympathetic/parasympathetic axis integrates across drives to modulate value structures and decisions, when multiple “orthogonal” dimensions are manipulated. The models will then be matched to behavior in approach-avoidance tasks where the homeostatic reward (e.g. food) is paired against a homeostatic punishment (e.g. unpleasant heat form a thermode cuff in primates) both in open field (rodents) and in operant conditioning set ups (primates). The models will be optimised for behavior and examined against recordings from multiple, interoception-related areas of the brain such as the amygdala, medial PFC, and insula.

The trainee will be part of a collaborative project that regroups the advisors team as well as labs of F. Battaglia (Radboud University, Netherlands) and K. Gothard (U Arizona) who will provide data from rodents and primates. Trainee will benefit from interactions with S. Palminteri and his team on multi-agent modelling. Trainee will be situated within the Group for Neural Theory that provides a rich intellectual environment for theoretical neuroscience.

In the primary visual cortex, distal dendrites of layer 2/3 and layer 5 pyramidal neurons act as key sites where sensory inputs and contextual signals converge and integrate nonlinearly. This integration is shaped by two major classes of inhibitory interneurons: somatostatin-expressing and NDNF-expressing cells. This thesis project will investigate how these interneurons coordinate to regulate dendritic computations and visual perception. Combining dual-color two-photon calcium imaging with high-density electrophysiology in awake mice, we will map their activation by sensory and contextual inputs, characterize their reciprocal interactions, and determine their impact on pyramidal neuron excitability and network dynamics. By linking single-cell mechanisms to cortical processing, this work seeks to reveal fundamental principles of inhibitory control in visual computation.

Background: Visual perception in the mammalian brain relies on the integration of feedforward sensory inputs and feedback contextual signals within the primary visual cortex (V1). Pyramidal neurons (PNs) in layers 2/3 and 5 are central to this process, receiving excitatory inputs along their dendrites, particularly in layer 1 (L1), where integration is modulated by dendrite-targeting inhibitory interneurons (INs). Two key IN subtypes—somatostatin-expressing (SST) and neuron-derived neurotrophic factor-expressing (NDNF) cells—exert distinct inhibitory influences on PN dendrites. SST INs provide spatially precise inhibition driven by sensory input, while NDNF INs deliver slower, diffuse inhibition in response to contextual signals. Preliminary data suggest reciprocal interactions between these INs dynamically shape dendritic integration windows, enabling flexible routing of information.
This PhD project aims to elucidate how SST and NDNF INs regulate dendritic excitability and PN ensemble activity during visual processing. By combining in vivo imaging, optogenetic perturbation, and electrophysiological recordings, we will uncover the spatiotemporal principles governing inhibitory control in V1.

Specific Aim 1: Characterize the in vivo activity and functional connectivity of SST and NDNF interneurons during feedforward and feedback visual processing.
We will use dual-recombinase SST-Flp::NDNF-Cre mice to express calcium indicators in both IN populations. Dual-color two-photon multiplane imaging in awake, behaving mice will reveal activity profiles of SST and NDNF INs in response to sensory (e.g., drifting gratings, contrast modulation) and contextual (e.g., size tuning, feedback receptive fields, Asahi illusion) stimuli. We will assess contrast sensitivity and stimulus-dependent recruitment dynamics.
To dissect input-specific connectivity, we will use PdCO, a bistable optogenetic silencer, to reversibly inhibit SST or NDNF INs during visual stimulation. These experiments will test whether feedforward inputs preferentially activate SST INs and suppress NDNF INs, facilitating sensory-contextual convergence on PN dendrites. We will also investigate whether SST and NDNF INs are synaptically interconnected and whether reciprocal inhibition modulates dendritic integration windows.

Specific Aim 2: Determine how SST and NDNF interneurons regulate dendritic activity and PN ensemble dynamics in response to visual stimuli.
We will perform multiplane two-photon imaging of sparsely labeled L2/3 and L5 PNs to monitor dendritic calcium signals during visual stimulation. By selectively silencing SST and/or NDNF INs using PdCO, we will assess how each IN subtype shapes dendritic nonlinearities and cortical excitability.
Complementary high-density Neuropixels recordings will capture PN population activity across cortical layers. We will evaluate how optogenetic silencing of SST and NDNF INs alters ensemble dynamics and information flow during feedforward and feedback visual processing. These experiments will reveal how dendritic inhibition sculpts PN responses and contributes to flexible sensory integration.

Significance: This project will define how SST and NDNF INs coordinate to gate dendritic integration in V1, providing insight into the cellular mechanisms underlying context-dependent visual processing. Understanding these dynamics is essential for decoding cortical computation and may inform therapeutic strategies for disorders involving dendritic over-inhibition, such as Down syndrome.

As we navigate our world we constantly move our eyes and head, yet our visual experience remains stable. This requires the brain to separate visual inputs caused by self-movement from environmental motion—but how cortico-thalamic circuits manage and transform these signals remains unclear. Self- versus externally-generated head motions engage distinct pathways, but there is evidences they converge onto the same structure: externally-imposed rotations signal through vestibular inputs to pulvinar thalamus, while self-generated rotations recruit motor areas, which also project to pulvinar. Through projection-specific optogenetics and high-density electrophysiological recordings in head-fixed and freely moving mice, we will determine how the pulvinar integrates these signals, whether the brain maintains self-centered or world-centered representations, and how visual cortex processes these sources of inputs. This work will establish the mechanisms enabling stable vision during self-motion.

Introduction: The construction of a stable spatial representation of the external world is fundamental to an organism's survival. Yet sensory inputs are acquired through organs in near-continuous motion, creating a fundamental challenge: the visual system must contend with constant eye and head movements that shift images on the retina, while interpreting motion of objects in the external world. To detect head motion, the brain relies on the motor system and on the vestibular system—sensory organs in the inner ear that signal head motion. Therefore, to accurately interpret the visual world, the brain must integrate retinal signals with self-motion information. Understanding these mechanisms is essential not only for basic neuroscience but also for insights into disorders where sensory-motor integration is disrupted, such as schizophrenia and autism spectrum disorders, where patients exhibit atypical responses in distinguishing the sensory consequences of their own actions from externally-caused sensory events.
The first cortical stage of visual perception, primary visual cortex (V1) represents an ideal entry point for investigating how head motion impacts visual processing. V1 neurons in mice respond to both eye movements and externally- or self-generated head movements, even in the dark. However, the neuronal circuits and mechanisms enabling V1 to integrate these signals remain largely unknown. Recent evidence suggests the pulvinar thalamus relays eye movement and vestibular signals to V1, while higher visual areas (HVAs) and retrosplenial cortex (RSC) may provide indirect routes or additional contextual inputs, forming a complex cortico-thalamic loop. Critically, the circuits mediating head motion differ fundamentally between self- versus externally generated head turns: while secondary motor cortex (M2) is dispensable for V1 modulation during passive (externally-imposed) rotations, M2 appears essential during active (self-initiated) turns, likely through projections to pulvinar. A fundamental question emerges from this picture: how does the brain manage and transform vestibular and motor signals to enable V1 to distinguish visual motion caused by self-motion from those caused by moving objects? This requires understanding: (1) Circuit architecture – Does the pulvinar act directly on V1 or through cortical intermediates, and how do pathways differ for active versus passive head turns? (2) Computational principles – How do vestibular and motor signals modulate visual processing? (3) Local circuit implementation – What is the role of V1 inhibitory and excitatory neurons in integrating self-motion signals with visual inputs?

Experimental Approach: Our strategy leverages the circuit difference between active and passive head turns to dissect distinct pathways. Using chronically implanted high-density silicon probes with an inertial measurement unit (IMU) for head tracking and a head-mounted camera for eye movements we will record neural activity under two conditions. For passive rotations, we will use head-restrained mice on a motorized turntable while presenting calibrated visual stimuli. This allows precise optogenetic targeting via a scanning galvo mirror systems and control of whether visual and vestibular signals are congruent or incongruent—enabling us to map Pulvinar-HVA-RSC pathways independently of M2. For active turns, we will record from freely moving mice. This allows us to test whether M2-dependent pathways converge with or operate parallel to passive pathways. Optic fibers will be chronically implanted above regions identified during passive mapping, enabling targeted manipulations during freely moving behavior.


Aim 1: Mapping Direct and Indirect Pathways from Pulvinar to V1: Pulvinar projects directly to V1 but also to HVAs and RSC, suggesting multiple potential pathways for vestibular signals. To isolate the cortical source of head and eye movement signals, we will first optogenetically silence individual HVAs (identified via intrinsic imaging) and RSC using galvo mirrors while recording V1 responses to passive head rotation and spontaneous eye movements. Second, to demonstrate the specific circuit connecting those areas and V1, we will perform projection-specific silencing by injecting rAAV-Cre in V1 and Cre-dependent eOPN3 in identified HVAs, silencing only HVA axons terminating in V1. In parallel, we will interleave silencing of pulvinar terminals in V1 with silencing of HVAs and RSC. If the pulvinar acts directly, silencing the pulvinar terminals in V1 should abolish head/eye movement responses; if indirect, silencing pulvinar projections to HVAs or RSC should reduce V1 modulation. Then, we will evaluate the role of those identified projections during active head turns in freely moving mice by placing an optic fiber above V1. If the same pathways mediate both conditions, effects should be identical. However, if M2 provides direct input to V1, bypassing the pulvinar-HVA/RSC pathway, we expect reduced efficacy of HVA/RSC silencing during freely moving mice. These approaches will establish whether eye and head movements (active and passive) engage segregated or shared circuit architectures.

Aim 2: Understanding Computational Mechanisms for Dissociating Head-Motion-Induced from Object-Induced Visual Motion: We aim to understand how visual circuits distinguish retinal motion caused by head/eye movements from motion caused by moving objects. We hypothesize V1 represents head movements, eye movements, and visual flow in an egocentric reference frame. Alternatively, V1 may integrate signals into an allocentric reference frame where visual responses become invariant to head motion. During passive rotations, we will present full-field moving gratings while varying rotation parameters, creating conditions where visual and vestibular signals are congruent or incongruent. For example, we will utilize an innate eye reflex (VOR) and a unique protocol (cancellation) where head rotation is paired with matched visual motion, eliminating eye movement and retinal slip (VOR cancellation). If V1 neurons respond during VOR cancellation, this reveals V1 has an internal model that reconstructs expected visual motion from head movement—a hallmark of allocentric encoding. Using longitudinal recordings tracking the same neurons during passive rotations and active head turns, we will determine whether computational principles differ when motor signals are present. Through dimensionality reduction and decoding of V1, pulvinar, and identified higher order cortical areas, we will determine whether vestibular, oculomotor, and visual signals are represented at the population level and how they change through this pathway.

Aim 3: The Role of Inhibitory and Excitatory Neurons in V1: Finally, we will determine how these computations are implemented within local circuits. Somatostatin-expressing interneurons (SOM), which receive strong input from pulvinar and HVAs and which respond to large-scale visual motion, are positioned to integrate vestibular and visual flow signals. By selectively suppressing SOM or VIP interneurons while recording from excitatory neurons during both passive and active movement conditions, we will evaluate the contribution of inhibitory cell types to distinguishing self-generated from externally generated motion cues.


Significance: Stable vision requires the brain to correct for the visual consequences of self-motion so that the world appears stable even when we move. The circuits and computations that perform this correction remain unclear. By identifying how vestibular and motor signals are combined with visual input in visual cortico-thalamic circuits, this project will reveal how the brain removes self-induced visual motion to isolate motion in the environment. These mechanisms are fundamental for perception and are disrupted in disorders where distinguishing self-generated from external sensory events is impaired.

Zebrafish have a special type of radial glia called radial astrocytes (RAs). In contrast to those of mammals, RAs do not degenerate and display analogous functions to those of astrocytes. In a recent study, we found that in the optic tectum of zebrafish, RAs synchronize their Ca2+ transients immediately after the end of a spontaneous escape behavior. The RA synchronizations modulated the direction selectivity and the functional correlations among tectal neurons. This mechanism supports freezing behavior following a switch to an alerted state. Here, we propose to investigate the neuronal pathways and mechanisms underlying the behavioural-state transition. This study will shed light on the neuronal and glia pathways and mechanisms across the
entire brain, underlying the behavioral-state transitions.

Zebrafish have a special type of radial glia called radial astrocytes (RAs). In contrast to those of mammals, RAs do not degenerate and display analogous functions to those of astrocytes. Previous studies showed that RAs in the hindbrain enable larvae to switch to a quiescent behavioral state. In the forebrain, perturbations of the glia-neuron interactions lead to the emergence of epileptic seizures?. In a recent study, we found that in the optic tectum of zebrafish, RAs synchronize their Ca2+ transients immediately after the end of a spontaneous escape behavior. Using two-photon Ca2+ imaging, optogenetics and ablations we showed that RA synchronous Ca2+ events are mediated by the locus coeruleus. The RA synchronizations modulated the direction selectivity and the functional correlations among tectal neurons. This mechanism supports freezing behavior following a switch to an alerted state. Here, we propose to investigate the neuronal pathways and mechanisms underlying the behavioural-state transition. For this purpose, we will use transgenic zebrafish expressing the genetically encoded Ca2+ indicator GCaMP7f and the opsin ReaChR (optogenetic stimulation of
neurons), in combination with a 3D two-photon system and a digital micromirror device for pattern illumination for opsin activation.
Supported by preliminary results, we will investigate the hypothesis that the medula oblangata gates the behavioral-state transitions, by integrating information from the internal state of the animal (e.g. sensory inputs form the gut and/or oxygen levels). In addition, we will investigate the downstream neuronal and glial pathways following the spontaneous activation of the medula oblangata.
Overall, this study will shed light on the neuronal and glia pathways and mechanisms across the entire brain, underlying the behavioral-state transitions.

Large-scale recordings have led to important insights into the organization of neural activity in different sensory and behavioral contexts. While task-related activity is typically found to be low-dimensional, spontaneous activity has been shown to be much higher-dimensional, with surprising properties such as a scale-invariant covariance spectrum. Most studies that try to make sense of these data follow a ‘top-down’ approach and infer underlying functional networks or motifs. Here, we propose to investigate ‘bottom-up’ how spatial connectivity patterns can support different regimes of activity, a clear understanding of which is still lacking. Deciphering how spatially modulated connectivities shape the spatio-temporal dynamics both at the population level and at the level of individual neurons, for neural activity of variable dimension, would be an important step towards an integrated understanding of the substrate of neural computations in the brain.

The advent of new electrophysiological and imaging techniques made it possible to simultaneously record the activity of large numbers of neurons, and up to the whole brain for organisms such as zebrafish larvae [1,2]. These data led to important insights into the spatio-temporal dynamics of neural activity in different sensory and behavioral contexts. Task-related activity has often been found to be well described by low-dimensional manifolds embedded in the high-dimensional neuronal state space [3,4], lending weight to the idea that a limited number of latent variables or latent dimensions robustly encode neural computations [5]. Spontaneous activity in the awake state has been shown to be higher-dimensional [6-8], with specific properties like a scale-invariant covariance spectrum [9,10]. Whereas dynamics of neural manifold dynamics can in principle be linked to low-rank structure in the synaptic connectivity between neurons [11], the scale-invariance of covariance spectra has been hypothesized to arise from a specific form of functional connectivity between neurons [9]. Most studies that try to make sense of the large-scale data follow a ‘top-down’ approach and try to infer underlying functional networks or motifs from the observed neural activity. Here, we propose to follow a complementary approach and investigate ‘bottom-up’ how spatial connectivity patterns, including long-range connections between distinct brain areas, can support different types of activity, a clear understanding of which is still lacking. Specific spatially modulated synaptic connectivities have been hypothesized to shape the spatio-temporal dynamics both at the population level, observed e.g. in the form traveling waves across the cortical surface [12], and at neuronal resolution e.g. by influencing the correlation structure of neural variability [13]. Deciphering their relative contributions for neural dynamics of variable dimension would be an important step forward to an integrated understanding of the substrate of neural computations in the brain. Specifically, in this project we aim to leverage recent advances in connectomics allowing us to build models of neural networks with data-constrained intra- and inter-area connectivity patterns. Such data exist for mice [14] and zebrafish larvae [15]. We will use biophysically realistic population models that have been successfully employed to capture traveling waves in cortex [16] to investigate the effect of specific inter-area connectivity patterns on brain-wide population dynamics. We will furthermore employ rate-based recurrent neural network models incorporating spatially structured connectivity to study the interplay of structural and functional connectivity, and their role in the observed crossover between high- and low-dimensional neural activity. The project will benefit from a collaboration with the team of Quan Wen (USTC, Hefei, China) that studies zebrafish larvae and records brain-wide neural activity in vivo during behavior (see also ref. [9]).

1. Ahrens, M. B., Orger, M. B., Robson, D. N., Li, J. M., & Keller, P. J. (2013). Whole-brain functional imaging at cellular resolution using light-sheet microscopy. Nature methods, 10(5), 413-420.
2. Jun, J. J., Steinmetz, N. A., Siegle, J. H., Denman, D. J., Bauza, M., Barbarits, B., ... & Harris, T. D. (2017). Fully integrated silicon probes for high-density recording of neural activity. Nature, 551(7679), 232-236.
3. Ahrens, M. B., Li, J. M., Orger, M. B., Robson, D. N., Schier, A. F., Engert, F., & Portugues, R. (2012). Brain-wide neuronal dynamics during motor adaptation in zebrafish. Nature, 485(7399), 471-477.
4. MacDowell, C. J., & Buschman, T. J. (2020). Low-dimensional spatiotemporal dynamics underlie cortex-wide neural activity. Current Biology, 30(14), 2665-2680.
5. Ostojic, S., & Fusi, S. (2024). Computational role of structure in neural activity and connectivity. Trends in Cognitive Sciences, 28(7), 677-690.
6. van der Plas, T. L., Tubiana, J., Le Goc, G., Migault, G., Kunst, M., Baier, H., ... & Debrégeas, G. (2023). Neural assemblies uncovered by generative modeling explain whole-brain activity statistics and reflect structural connectivity. eLife, 12, e83139.
7. Benisty, H., Barson, D., Moberly, A. H., Lohani, S., Tang, L., Coifman, R. R., ... & Higley, M. J. (2024). Rapid fluctuations in functional connectivity of cortical networks encode spontaneous behavior. Nature Neuroscience, 27(1), 148-158.
8. Manley, J., Lu, S., Barber, K., Demas, J., Kim, H., Meyer, D., ... & Vaziri, A. (2024). Simultaneous, cortex-wide dynamics of up to 1 million neurons reveal unbounded scaling of dimensionality with neuron number. Neuron, 112(10), 1694-1709.
9. Wang, Z., Mai, W., Chai, Y., Qi, K., Ren, H., Shen, C., ... & Wen, Q. (2025). The geometry and dimensionality of brain-wide activity. eLife, 14, RP100666.
10. Moosavi, S. A., Hindupur, S. S. R., & Shimazaki, H. (2024). Population coding under the scale-invariance of high-dimensional noise. BioRxiv, 2024-08.
11. Mastrogiuseppe, F., & Ostojic, S. (2018). Linking connectivity, dynamics, and computations in low-rank recurrent neural networks. Neuron, 99(3), 609-623.
12. Rubino, D., Robbins, K. A., & Hatsopoulos, N. G. (2006). Propagating waves mediate information transfer in the motor cortex. Nature neuroscience, 9(12), 1549-1557.
13. Rosenbaum, R., Smith, M. A., Kohn, A., Rubin, J. E., & Doiron, B. (2017). The spatial structure of correlated neuronal variability. Nature neuroscience, 20(1), 107-114.
14. Allen Brain Map, https://portal.brain-map.org/connectivity
15. Max Planck Zebrafish Brain Atlas, https://mapzebrain.org/home
16. Kang, L., Ranft, J., & Hakim, V. (2023). Beta oscillations and waves in motor cortex can be accounted for by the interplay of spatially structured connectivity and fluctuating inputs. eLife, 12, e81446.

This thesis studies the role of post-stroke cerebral reorganization (multimodal MRI, 100 patients in the subacute phase recruited in the MOTIF-STROKE multicenter clinical trial at NeuroSpin). Its objectives are: 1) to construct an atlas of disconnections (diffusion MRI) and to study the reorganization of intrinsic functional networks (default mode, executive) using resting-state functional MRI. To achieve this, whole-brain tractography algorithms will be employed to investigate the underlying axonal reorganization of the white matter. By combining these analyses with a morphological atlas constructed from the same patients, it will be possible to generate a structural connectivity matrix for each individual and to associate it with the functional connectivity derived from resting-state fMRI. 2) To predict the potential for motor recovery by machine learning using this deep phenotyping and clinical biomarkers of gait and grasping.

CONTEXT
Ischemic stroke is a leading cause of long-term motor disability. Recovery of function relies on a complex reorganization of brain networks, a process that is not yet fully understood [Favre, 2014; Hommel, 2016]. The BrainSync project (November 2024–September 2029), funded within the framework of the CEA’s at-risk Audace! Research program, aims to develop innovative interventional strategies for motor rehabilitation of the upper limb in subacute stroke patients.
Specifically, the MOTIF-STROKE clinical trial proposes an innovative approach by combining ultra-high field (7T) and high-field (3T) magnetic resonance imaging (MRI), artificial intelligence (AI), and digital consultations to map motor intentions, identify brain lesions, and assess motor disabilities in hundreds of subacute stroke patients with upper limb impairments. Within this project, the current PhD thesis co-supervised by Dr. Philippe Ciuciu, Director of the Inria-CEA MIND unit, Dr. Ivy Uszynski (PhD), head of the Ginkgo team at CEA/NeuroSpin/BAOBAB, and Prof. Myriam Edjlali-Goujon (MD, PhD), neuroradiologist at Paris University Hospitals, brings together a consortium of experts in fMRI and dMRI data analysis on one hand and in neuroradiology and stroke on the other hand.

SCIENTIFIC OBJECTIVES

This study aims to elucidate the relationship between cerebral (cortical and subcortical) lesions, disconnections, and motor disability in 100 stroke patients, focusing on their structural impairments and functional reorganization.
The primary objective of this project is to construct a high-resolution morphological, functional, and structural atlas of post-stroke disconnection and reorganization, addressing a major gap in current stroke research. Building on the disconnectome framework [Thiebaut de Schotten et al., 2020; Forkel et al., 2022], this atlas will couple lesion-derived structural disconnection maps with multimodal MRI data to decode the relationships between anatomical disconnections, functional connectivity alterations, and behavioral outcomes. Moving beyond conventional MRI lesion maps, it will provide probabilistic representations that link white-matter disconnection to anomalies in resting-state functional connectivity and task-related motor activations, thereby yielding an unprecedented, multimodal, high-resolution understanding of post-stroke network reorganization.

This PhD project will focus specifically on the structural dimension of the atlas and its relationship—at both the individual and cohort levels—with functional connectivity measured using resting-state fMRI at 3 T. By combining advanced diffusion-based tractography, morphological contrasts, and functional imaging, we will delineate the core white-matter pathways whose disconnection critically impacts motor recovery, in analogy with how Forkel et al. (2022) decoded the reading circuitry from functional and disconnectome data.

By integrating multimodal neuroimaging with detailed clinical phenotyping, this atlas will enable accurate prediction of post-stroke deficits and recovery trajectories, support personalized rehabilitation strategies, and guide the targeted use of neurostimulation interventions. To achieve this, a large cohort of subacute ischemic stroke patients with motor impairments will be systematically recruited across diverse vascular territories, ensuring comprehensive coverage of disconnection patterns relevant to motor recovery. This approach will overcome the key limitations of prior studies (e.g., [Jaillard 2005; Hannanu 2017; Hannanu 2020])—typically restricted to small cohorts (20–25 patients), low-resolution T MRI, and a lack of integration between structural (diffusion) and functional (resting-state fMRI) connectivity.
The atlas will support complementary connectivity analyses across this large cohort, correlating them with motor-task activation maps (and specifically hand and foot movements recorded using dedicated 7 T task-based fMRI paradigms) with quantitative biomarkers of motor performance. Motor impairment will be quantified using both the Fugl–Meyer clinical scale and an innovative digital consultation tool that extracts objective indices such as gait asymmetry and lateralized grasping performance, thereby enriching standard measures of motor disability [Jaillard 2021] and recovery [Hommel 2016]. High-resolution morphological contrasts (T1-weighted and FLAIR images) will also be integrated to facilitate individual projection within the MNI (Montreal Neurological Institute) space [Brett 2002].
The ultimate goal of this atlas is to assist clinicians and neurosurgeons in optimizing WIMAGINE neuroprosthesis implantation [Lorach 2023] in stroke patients participating in BCI4STROKE, a forthcoming clinical trial within the BrainSync program. BCI4STROKE will develop innovative, intensive neurorehabilitation strategies grounded in brain–computer interfaces (BCIs), real-time decoding of cortical activity recorded through the WIMAGINE implant, and specialized effectors such as functional electrostimulation systems and the Glorea glove [Maleševi? 2012; Popovic-Manesci 2013; Bressi 2022; Vanoglio 2024].

REFERENCES
[Abraham et al, 2014] Abraham et al. Machine learning for neuroimaging with scikit-learn. Frontiers in neuroinformatics. 2014 Feb 21;8:14.
[Braaß et al, 2025] Braaß et al Altered functional connectivity between cortical premotor areas and the spinal cord in chronic stroke. Stroke. 2025 May;56(5):1159-68.
[Brett2002] Brett et al. The problem of functional localization in the human brain. Nature reviews neuroscience. 2002 Mar 1;3(3):243-9.
[Bressi2022] Bressi et al. Robotic-Assisted Hand Therapy with Gloreha Sinfonia for the Improvement of Hand Function after Pediatric Stroke: A Case Report. Appl. Sci. 2022, 12, 4206.
[Chauvel et al, 2024] Chauvel et al. (2024). Comparative analysis of the chimpanzee and human brain superficial structural connectivities. Brain Struct Funct. 229(8):1943-1977.
[Comby2023] Comby et al. Denoising of fMRI volumes using local low-rank methods. In IEEE 20th International Symposium on Biomedical Imaging (ISBI) 2023 Apr 18 (pp. 1-5).
[de Schötten et al, 2020] Thiebaut de Schotten et al. Brain disconnections link structural connectivity with function and behaviour. Nature communications. 2020 Oct 9;11(1):5094.
[Forkel et al, 2022] Forkel et al. Stroke disconnectome decodes reading networks. Brain Structure and Function. 2022 Dec;227(9):2897-908.
[Herlin et al, 2023]Herlin et al (2023). Cross-subject variability of the optic radiation anatomy in a cohort of 1065 healthy subjects. Surg Radiol Anat. 45(7):849-858.
[Hommel2016] Hommel et al. How to measure recovery? Revisiting concepts and methods for stroke studies. Translational stroke research. 2016 Oct;7(5):388-94.
[Jaillard2005] Jaillard et al. Vicarious function within the human primary motor cortex? A longitudinal fMRI stroke study. Brain. 2005 May 1;128(5):1122-38.
[Jaillard2021] Jaillard A. Structural or functional metrics to assess motor impairment in chronic stroke?. Stroke. 2021 Jan;52(1):250-2.
[Lorach2023] Lorach et al. Walking naturally after spinal cord injury using a brain–spine interface. Nature. 2023 Jun;618(7963):126-33.
[Maleševi?2012] Maleševi? et al. A multi-pad electrode based functional electrical stimulation system for restoration of grasp. J Neuroeng Rehabil. 2012
[Popovic-Manesci2013] Popovi?-Maneski et al. Multi-pad electrode for effective grasping: design. IEEE Trans Neural Syst Rehabil Eng. 2013 Jul;21(4):648-54.
[Vanoglio2024] Vanoglio et al. A Sensor-Based Upper Limb Treatment in Hemiplegic Patients: Results from a Randomized Pilot Study. Sensors (Basel). 2024 Apr 17;24(8):2574.
[Veraart et al, 2016] Veraart et al. Denoising of diffusion MRI using random matrix theory. Neuroimage. 2016 Nov 15;142:394-406.
[Jeurissen et al, 2014] Jeurissen et al. Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data. Neuroimage. 2014

The mechanisms underlying neuronal vulnerability to Tau neurofibrillary pathology in early Alzheimer’s disease (AD) remain poorly understood. Leveraging a groundbreaking scRT-QuIC (single-cell real-time quaking-induced conversion) technology, we detect Tau seeding activity in neurons isolated from AD patient brains, including those that likely represent the earliest tauopathy stage, characterized by the absence of detectable inclusions by immunolabeling yet the presence of pro-aggregative Tau.
By integrating scRT-QuIC with RNA sequencing on human neurons isolated from human brains, we will pinpoint neuronal susceptibility factors. Additionally, we will compare transcriptomic profiles of Tau RT-QuIC-positive neurons from AD and primary age-associated tauopathy (PART) patients to elucidate their relationship. From a translational perspective, this approach will also help identify early biomarkers, with a focus on phosphorylated Tau isoforms recently proposed as fluid biomarkers in AD.

Alzheimer’s disease (AD) is neuropathologically defined by the accumulation of extracellular amyloid-? (A?) plaques and intracellular neurofibrillary tangles composed of hyperphosphorylated Tau protein. While A? deposition can occur in cognitively normal individuals, the distribution and progression of Tau pathology are strongly correlated with cognitive decline and neurodegeneration. Despite this correlation, the mechanisms governing the initiation and propagation of Tau pathology remain poorly understood 1,2.
A critical gap in our understanding is the identification of cellular factors that render specific neurons susceptible to Tau aggregation, while others in the same brain region remain resistant. This differential vulnerability suggests that intrinsic molecular and genetic characteristics of neurons play a pivotal role in determining their susceptibility to Tau misfolding and propagation. To address this, we will investigate the earliest stages of Tau pathology, where neurons may harbor misfolded Tau with seeding potential but lack detectable Tau inclusions, such as those stained by the AT8 antibody.
Recently, our team successfully demonstrated the feasibility of single-cell RT-QuIC (scRT-QuIC) to detect misfolded Tau with pro-aggregative properties in neurons isolated from the cortex of AD patients. This breakthrough allows us to identify neurons that are negative for AT8 immunostaining but exhibit Tau seeding activity, representing the earliest stage of tauopathy. To our knowledge, this approach is unparalleled worldwide and provides a unique opportunity to study the initial events of Tau misfolding in AD.

PhD Project Objectives
The overarching goal of this doctoral project is to determine why some human cortical neurons are vulnerable to Tau aggregation while others remain resistant, even within the same brain region. We will focus on both AD and primary age-associated tauopathy (PART), a condition that may represent an early stage of AD 3. By establishing an ultra-sensitive single-cell pipeline (scRT-QuIC + scRNAseq), we aim to correlate Tau seeding activity with gene expression profiles across thousands of neurons.

Objective 1 (WP1): Cellular Factors Contributing to Neuronal Susceptibility to Tau Pathology in AD
We will develop and implement a method to perform RT-QuIC and scRNAseq on single human neurons isolated from the postmortem entorhinal cortex of AD patients at various Braak stages (1–4) and from age-matched controls (six AD patients at Braak stages 1–4 and three age-matched controls from the NeuroCEB brain bank).
Frozen entorhinal cortex samples will be dissociated, and neurons will be immunolabeled with anti-NeuN and AT8 antibodies 4. Cells will be encapsulated using the ModaFlow microfluidic platform from LiveDrop Bio, and reverse transcription will be performed overnight.
Amplification of cDNA from single neurons will be carried out using TS-PCR. Tau RT-QuIC will be performed on amplified neurons to detect seeding activity. scRNAseq and bioinformatic analyses will be conducted to identify differential gene expression and signaling pathways.

Objective 2 (WP2): Transcriptomic Comparison of AD and PART Neurons
Primary age-associated tauopathy (PART) has been proposed as an early stage of AD, although this remains debated. Immunodetection and cryo-electron microscopy analyses suggest that Tau fibers from PART and AD patients share structural similarities. By identifying a common transcriptomic signature in affected neurons—particularly AT8-negative/RT-QuIC-positive neurons—we aim to determine whether PART represents a pathophysiological continuum with AD.
This work package will involve single-cell analysis of neurons from the entorhinal cortex of PART patients using the same methodology as WP1. scRNAseq and bioinformatic analyses will be performed to compare the transcriptomic profiles of PART neurons with those from AD patients.

Objective 3 (WP3): Identification of Early Tau Biomarkers
We will evaluate recently identified early biomarkers of tauopathy, including phosphorylated Tau isoforms such as pTau217, brain-derived Tau (BD-Tau), and pTau212, which have been detected in the cerebrospinal fluid (CSF) and plasma of AD patients at early disease stages (three AD patients at Braak stages 1–4) 5-7. By comparing the transcriptomic profiles of neurons sorted using antibodies against these biomarkers, we aim to validate their potential for early diagnosis and patient stratification in clinical trials.

Expected Outcomes
This project will provide critical insights into: 1)Neuronal susceptibility factors that determine vulnerability or resistance to Tau aggregation in AD and PART; 2) The relationship between PART and AD, potentially establishing PART as an early stage of AD; 3) Early diagnostic biomarkers for Tau pathology, which could improve patient stratification and inclusion in clinical trials.
By deciphering the earliest molecular events in Tau pathology, this research will contribute to the development of targeted therapies and diagnostic tools for Alzheimer’s disease and related tauopathies.
Organization of the work: The selected PhD student recruited on the project will be in charge of running most of the experiments. She/he will be trained for postmortem brain dissociation and to the use of the microfluidic platform, RT-TSPCR and RT-QuIC. We will get strong support from the iGENSEQ platform of ICM to perform scRNAseq. The data analysis core of ICM will be involved in RNAseq analysis. The team has developed R-based bioinformatic tools.

1. Clavaguera F, Bolmont T, Crowther RA, et al. Transmission and spreading of tauopathy in transgenic mouse brain. Nat Cell Biol 2009; 11(7): 909-13.
2. Jucker M, Walker LC. Self-propagation of pathogenic protein aggregates in neurodegenerative diseases. Nature 2013; 501(7465): 45-51.
3. Duyckaerts C, Braak H, Brion JP, et al. PART is part of Alzheimer disease. Acta Neuropathol 2015; 129(5): 749-56.
4. Otero-Garcia M, Mahajani SU, Wakhloo D, et al. Molecular signatures underlying neurofibrillary tangle susceptibility in Alzheimer's disease. Neuron 2022; 110(18): 2929-48 e8.
5. Clemmensen FK, Gonzalez-Ortiz F, Gramkow MH, et al. The plasma p-tau217/BD-tau ratio improves biomarker short-term variability in memory clinic patients. Alzheimers Dement (N Y) 2025; 11(3): e70143.
6. Kac PR, Gonzalez-Ortiz F, Emersic A, et al. Plasma p-tau212 antemortem diagnostic performance and prediction of autopsy verification of Alzheimer's disease neuropathology. Nat Commun 2024; 15(1): 2615.
7. Montoliu-Gaya L, Salvado G, Therriault J, et al. Plasma tau biomarkers for biological staging of Alzheimer's disease. Nat Aging 2025.

Retinal organoids derived from stem cells have become a powerful tool to study human development and function. Current organoid models fail to fully recapitulate the regional complexity of the native retina, particularly with respect to the macula, a specific region of our eye that enables sharp vision. Without better insight into retinal development, organoids remain generic, limiting their relevance for disease modeling and therapeutic applications that target region-specific conditions like age-related retinal dystrophies. This PhD project will combine single-cell omics technologies and in vitro culture of human retinal tissues to elucidate the epigenetic mechanism underlying macular development. Building upon these findings, the project aims to manipulate intrinsic molecular cues to enhance the regional complexity of retinal organoids. The student will acquire expertise in stem-cell biology, retinal development, epigenetics and computational analysis.

Scientific context and objectives
The mechanisms underlying retinal regionalization are not yet fully understood, mainly due to the early timing of this process during development and the difficulties in accessing the relevant tissue samples. However, accumulating evidence indicates that epigenetic regulation, particularly through the establishment of repressive marks, plays a critical role in retinal development. Yet, the dynamic interplay between the epigenetic landscape and the process driving retinal regionalization remains largely unexplored. The outlined PhD project has two main objectives: first, to elucidate the epigenetic mechanisms that regulate the development of the human macula using in vitro model of human retina; and second, to apply this knowledge to develop physiologically relevant in vitro retinal organoid models that will advance our understanding and treatment of macular disorders.
This work integrates stem cell biology, epigenomics, and neurodevelopmental modeling, and directly aligns with DIM C-BRAINS Axis 1: genomics, human cells, reprogramming, and neuro-organoids.

Methodology
In vitro culture
To investigate early human retinal development, the candidate will use in vitro culture of retinal tissue. In addition, the candidate will culture induced pluripotent stem cells (iPSCs) and differentiate them into 3D retinal organoids following optimized protocols established by the host laboratory. Several iPSC engineered lines expressing fluorescent reporters previously generated by the team will be used to enable efficient identification of specific cell types.

Cellular and structural characterization
Retinal tissues generated from different retinal regions and organoids will be characterized across developmental stages. Cell proliferation will be measured by EdU incorporation, and cell fate will be assessed through immunostaining with established markers, combined with three-dimensional imaging using confocal and light-sheet microscopy.

Epigenetic profiling
To characterize the epigenetic and transcriptomic landscape of the developing human retina, the candidate will employ a single-cell multi-omics approach across different developmental stages and regions. Parallel experiments will then be conducted on developing organoids at matching stages for comparative analysis with native tissue. Integration of these datasets will allow the identification of candidate regulatory genes controlling cell fate decision and retinal regionalization over time. The student will work in close collaboration with computational scientists and the single-cell analysis platform present at the host institution to analyze and compare datasets.
Epigenetic perturbations
To better reproduce native tissue characteristics in organoids, the study will build on distinct epigenetic signatures identified between the human retina and organoids. A two-step strategy will be employed on organoids at several developmental stages, focusing on how epigenetic regulation shapes cell identity and tissue architecture. In the first step, pharmacological modifiers will be used to alter the activity of enzymes regulating repressive chromatin modifications. In the second step, CRISPR/dCas9-based tools will be applied to specifically target and modulate the expression of a selected set of genes, allowing detailed dissection of the mechanisms underlying retinal regionalization. The impact on progenitor proliferation and neuronal differentiation will be evaluated using single-cell transcriptomic analysis and immunostaining.

Expected results and impact
The outlined project will provide critical insights into the epigenetic mechanisms that ensure cell fate specification and proliferation during human retinal development. Overall, these findings will help refine organoid models to more accurately mimic the organization and function of the native retina. Furthermore, this enhanced organoid model will provide a unique platform for the development of innovative therapeutic strategies for macular disease. For the PhD candidate, this project offers a unique opportunity to gain cutting-edge skills at the interface of stem-cell culture, developmental neurobiology, genome editing, and computational biology.

Supervision and environment
The PhD candidate will work in Dr. Olivier Goureau’s laboratory at Institut de la Vision in Paris, a leading center for vision research and ophthalmology. The team provides access to facilities for stem cell culture, imaging, and single-cell omics technologies. In addition, the Institut de la Vision offers a dynamic and interdisciplinary research environment where the candidate will collaborate with computational scientists and neurodevelopmental teams. Participation in regular seminars and workshops hosted by the Institute will further strengthen the candidate’s knowledge and professional network.

For much of childhood, limited metabolic resources are unevenly distributed between brain and body, with neural maturation prioritized over physical growth. This energy trade-off critically shapes language acquisition, a cornerstone of human development. The PhD project investigates whether disruptions in this brain–body balance contribute to developmental language disorder (DLD). Combining behavioral, EEG, and MRI measures in 150 children aged 4–7, it will examine how energy allocation influences predictive speech coordination—the ability to synchronize one’s voice with rhythmic sounds—and its neural basis in beta–gamma oscillatory coupling. Body growth trajectories (BMI) and brain metabolic activity during speech perception will serve as complementary indices of the energy trade-off. The project aims to reveal how atypical energy allocation alters neural efficiency and compromises language development.

Language acquisition is one of the most demanding achievements of human development, relying on a long period of brain maturation during childhood. This stage involves a profound metabolic reorganization, as limited energy resources are redistributed between the growing body and the maturing brain1. In infancy, most energy is devoted to the body, supporting rapid somatic growth. Then, from the end of the first year to approximately six years of age, a growing share of metabolic energy is devoted to the brain2, fueling an intense period of neural plasticity that enables the development of higher cognitive functions, particularly the ones related to language3,4.
This shift—known as the energy trade-off between brain and body—creates a unique developmental context in which neural circuits can specialize for complex cognitive functions. We propose that, during this period, increasing metabolic investment in the brain promotes efficient communication between the auditory and frontal cortices, the core network for speech and language. Within this network, neural oscillations coordinate predictive and perceptual processes: beta oscillations (15–30 Hz) convey top-down predictions5,6, while gamma oscillations (30–100 Hz) encode the sensory details of incoming sounds7,8. Their progressive synchronization could reflect both the refinement of predictive mechanisms and the optimization of energetic efficiency in cortical communication.
Understanding language development therefore requires considering not only brain structure and function, but also the energetic constraints that shape maturation. When more energy is directed to the brain, cortical circuits can refine their communication patterns. Conversely, if energy allocation is disrupted by biological, environmental, or developmental factors, neural systems that support language may fail to organize efficiently. The AtyPi project builds on this framework, proposing that temporal coordination within the auditory–frontal network reflects how effectively the brain uses its energetic resources during development.
This PhD project aims to test whether an imbalance in the developmental energy trade-off between brain and body contributes to Developmental Language Disorder (DLD). Children with DLD show atypical neural oscillatory responses to speech, indicating altered temporal organization of brain activity9–11. We hypothesize that abnormal energy allocation during early development disrupts the coordination between auditory and frontal cortices, weakening predictive control of speech processing.
Objectives:
1. Determine whether the developmental trajectory of audio-vocal synchronization differs between typically developing children and children with DLD.
2. Characterize the oscillatory dynamics (beta and gamma) that support successful synchronization and test whether their interaction constitutes a key neural principle underlying speech acquisition.
3. Evaluate, via MRI and growth parameters, whether individual differences in brain–body energy allocation predict language outcomes and neural efficiency.
Experimental Plan
This project uses a within-participant design combining behavioral, EEG, and MRI data from the same cohort whenever possible. We will recruit 150 children in three age bands (4–5, 5–6, and 6–7 years), including both typically developing (?75) and DLD (?75) participants. This sample size accounts for potential dropout due to the two-session design. Each participant will complete behavioral assessments and EEG recordings in a single session, and an additional MRI sessions will be conducted when feasible.
Study 1: Behavioral and Neurophysiological Development of Audio-Vocal Synchronization. Our preliminary results show that the ability to synchronize vocal production with rhythmic auditory stimuli emerges between four and five years and refines until around seven (N = 400), paralleling heightened neural specialization and metabolic investment in the brain. Participants will perform a vocal-synchronization task in which they produce a simple syllable (e.g., /da/) in time with a rhythmic auditory stimulus (1 Hz). Behavioral measures will include the temporal delay between the stimulus and vocal onset, the variability of this delay, and the phase-locking value (PLV) quantifying synchronization strength.
Standardized language tests (vocabulary, phonology, morphosyntax) will provide complementary linguistic measures. Synchronization and language data will be modeled to test whether temporal coordination predicts language outcomes and whether children with DLD show reduced synchronization accuracy, marking their language deficit. The cross-sectional design will further enable modeling of developmental trajectories and between-group differences using structural-equation modeling approaches. We expect that atypical audio-vocal synchronization in children with DLD will be directly associated with poorer language performance, revealing a behavioral marker of their underlying language deficit.
Study 2: EEG Characterization of Oscillatory Mechanisms. EEG recordings acquired during the same synchronization task will characterize the oscillatory dynamics underlying predictive speech coordination. Analyses will focus on oscillatory power, coherence, and cross-frequency coupling between beta (15–30 Hz) and gamma (30–100 Hz) frequency bands in auditory and frontal regions. The study will test whether successful synchronization depends on efficient beta–gamma interaction and whether this coupling acts as a central neural mechanism supporting the developmental acquisition of predictive speech coordination. Group comparisons (DLD vs. controls) will reveal whether atypical temporal coordination arises from delayed or inefficient oscillatory organization.
Study 3: MRI Assessment of Neural and Metabolic Correlates of the Energy Trade-Off. MRI will provide complementary measures of brain maturation and metabolic activity underlying the energy trade-off between body growth and brain development. Body growth will be approximated using Body Mass Index (BMI) trajectories, collected from birth through the date of inclusion for each participant, based on medical records (carnet de santé) and verified at inclusion. These longitudinal anthropometric data will serve as proxies for somatic energy expenditure and will be integrated with neural measures to estimate the brain–body energy balance.
Structural MRI will quantify cortical thickness and myelination indices, while diffusion-weighted imaging will assess white-matter connectivity within the auditory–frontal network. In addition, functional MRI (fMRI) will approximate brain energy consumption during speech perception. Participants will listen to both natural and temporally compressed speech, the latter condition strongly engaging predictive mechanisms. This design will allow us to test whether activity within the temporo–frontal network—particularly the auditory cortex and inferior frontal gyrus—varies as a function of age and group (typically developing vs. DLD). We expect that younger children and DLD participants will show greater metabolic engagement of this network, reflecting higher energetic costs of predictive processing.

By integrating behavioral, neural, and biological data, this project will model how variations in the brain–body energy balance relate to language acquisition and its disruption in DLD. Based on previous findings1, we expect that children showing slower body-growth trajectories—reflecting greater energetic investment in the brain—should display higher language abilities and more efficient neural synchronization. Conversely, greater energy devoted to somatic growth may limit neural maturation and atypical language acquisition.

Primary Progressive Aphasia (PPA) comprises a group of rare neurodegenerative syndromes selectively impairing language. Their phenotypic diversity provides a unique model to investigate the neural architecture of language and its interactions with executive control, memory, and attention. Owing to their early onset and focal nature, PPA variants are ideal for multimodal neuroimaging and causal exploration. Yet, their cognitive and biological complexity, especially in relation to other dementias, remains challenging. This project addresses fundamental questions on the behavioral expression and neural organization of language by integrating large multimodal datasets (CSF, PET-MRI, fMRI-DTI, EEG, cognition) and applying advanced machine learning and AI approaches to extract biomarkers and model language phenotypes, ultimately informing diagnosis and rehabilitation.

Highly sophisticated language and communication skills represent one of the most distinctive and defining cognitive abilities of humans. They enable us to interact, learn, educate, exchange, and cooperate across large social groups, underpinning the exceptional degree of cultural and technological development our species has achieved. For more than a century, both normal and impaired language functions have been studied and characterized in relation to their cognitive, anatomical, and neurophysiological components. Although often examined in isolation, research in cerebrovascular and neurodegenerative patients strongly suggests that language networks interact extensively with other domain-general cognitive systems. Influential models have proposed that executive control, working memory, and attention are not merely auxiliary but integral to the language system, and that damage to these mechanisms compromises linguistic efficiency. These processes are supported by a distributed network of cortical and subcortical hubs, including the prefrontal cortex, inferior parietal lobule, temporo-parietal junction, striatum, and thalamus.

Primary Progressive Aphasia (PPA) encompasses a group of rare neurodegenerative conditions affecting language and communication, estimated to impact roughly 6,000 individuals in France. Three major variants have been described: the non-fluent/agrammatic form (nfv-PPA), characterized by deficits in syntax and phonology linked to left frontal atrophy; the logopenic form (lv-PPA), marked by lexical and verbal working memory impairments associated with atrophy of the left temporo-parietal junction; and the semantic variant (sv-PPA), featuring bilateral anterior temporal lobe degeneration and profound loss of semantic knowledge. Together with other early-onset neurodegenerative syndromes showing language dysfunction, such as Frontotemporal Dementia (FTD) and Progressive Supranuclear Palsy (PSP), PPA offers a unique model to explore the neural and cognitive architecture of language. These conditions provide a natural framework to investigate how executive, memory, and attentional deficits interact with language impairments, revealing both pathological mechanisms and principles of healthy brain organization.

Recent advances in computational neuroscience and neuroimaging open unprecedented opportunities to study PPA and language systems quantitatively. The integration of multimodal MRI and PET imaging, neurophysiological data (MEG, EEG), and causal brain perturbation methods (TMS, tCS) enables the precise mapping of the structural and functional anatomy of language. Clinically, the heterogeneity of PPA variants poses major diagnostic challenges that demand data-driven, multimodal approaches capable of generating integrative, quantitative classifications. The diversity of PPA phenotypes provides ideal lesion models for exploring the functional architecture of linguistic systems, while their focal and early-onset nature—particularly in sv-PPA (Semantic Dementia)—makes them promising candidates for personalized neurorehabilitation strategies using non-invasive brain stimulation.

This project aims to address both fundamental and translational questions concerning the organization, dysfunction, and rehabilitation of language systems in PPA. It leverages one of the largest French cohorts of early-stage PPA patients (PHRC CAPP; n=91) and matched healthy controls (n=24), including comprehensive linguistic, neuropsychological, structural MRI, diffusion, fMRI resting-state, and PET-FDG data, alongside cerebrospinal fluid biomarkers. To expand this foundation, two additional datasets will be integrated. The first, the ICM-sponsored STIMLANG protocol, includes well-characterized patients across the three PPA variants (sv-PPA n=18, lv-PPA n=6, nfv-PPA n=4), PSP (n=12), and bv-FTD (n=12) who underwent a crossover study involving transcranial Direct Current Stimulation (tDCS) with three modalities (left anodal, right cathodal, sham) targeting the anterior temporal lobe (ATL), temporo-parietal junction (TPJ), and inferior frontal gyrus (IFG). The second, the PHRC National STIMSD clinical trial, tests the effects of multi-day tDCS regimens in sv-PPA (n=39) versus healthy controls (n=20) using a randomized, double-blind, placebo-controlled design. All datasets include detailed assessments of language performance (semantic access, fluency, reading, speech and voice recordings) and broader cognitive domains (memory, executive control, attention, perception, humor) using validated neuropsychological tools.

Traditional inferential and hypothesis-driven analyses (correlation, regression models) will serve as a first analytical layer to identify associations among clinical, cognitive, neuroimaging, and electrophysiological measures. Priority will then shift toward advanced machine learning and AI-based approaches (univariate and multivariate classifiers, deep learning) to exploit these multimodal datasets. This framework will enable cross-validation across patient cohorts and control groups, the discovery of quantitative biomarkers of language phenotypes, and the development of predictive diagnostic and prognostic models.

AIM 1: To identify novel multimodal diagnostic biomarkers and establish new quantitative classification dimensions for language-impairing conditions through algorithms enabling automated PPA subtype classification based on neuroimaging, cognitive-linguistic, and biological data.
AIM 2: To characterize the structural and functional interactions within neural networks supporting phonological, syntactic, and semantic processing, defining variant-specific language networks and testing their causal relevance.
AIM 3: To integrate biological (CSF) data to stratify the PPA population by underlying pathology and identify vulnerable nodes and networks with high susceptibility to neurodegeneration, providing future prognostic markers of early disease progression.
AIM 4: To model current-field distributions from MRI-based simulations (SimNIBS) and determine optimal non-invasive brain stimulation strategies to enhance variant-specific language functions in PPA, paving the way for future personalized cognitive neurorehabilitation approaches.

The COGSTIMLANG project will be co-supervised by Dr. Antoni Valero-Cabré (MD PhD HDR, DR CNRS, ICM, Team FRONTLAB) and Dr. Marc Teichmann (MD PhD HDR, PH APHP Im2A, Pitié-Salpêtrière), both experts in the neural bases of language and its disorders, combining expertise in neuropsychology, multimodal neuroimaging (MRI, PET, fMRI, DTI), electrophysiology (EEG), and causal neuromodulation (TMS, tCS). Collaborations with Dr. Ninon Burgos (PhD HDR, DR CNRS, ICM, Team ARAMIS) in machine learning for neuroimaging, Prof. Aurélie Kas (MD, PhD HDR, PU-PH. APHP Salpêtrière, Nuclear Medicine), Prof. Nadya Pyatigorskaya (MD PhD HDR, PU-PH, APHP Salpêtrière, Neuroradiology, Team MOVIT), and Dr. Raffaella Migliaccio (MD PhD HDR, CR INSERM, Team FRONTLAB) for resting-state fMRI will ensure the project’s success and contribute to the quantitative, integrative understanding of language organization and its dysfunctions in neurodegenerative disease.

Myelination is an activity-dependent process termed adaptive myelination. Neuronal activity, experience and behavior can enhance oligodendrocyte precursor cell (OPC) proliferation, differentiation, and remyelination, offering a potential therapeutic avenue for demyelinating diseases. This project investigates how different motor training paradigms (simple vs complex running wheel tasks) modulate oligodendroglial activity and myelin regeneration after cuprizone-induced demyelination in mice. Using in vivo Ca2+ imaging, ex vivo patch-clamp recordings, and immunohistochemistry, Aim 1 will characterize activity-dependent Ca2+ signaling in OPCs and oligodendrocytes during training. Aim 2 will explore underlying molecular mechanisms, focusing on mitochondrial and proteomic changes driving adaptive myelination. This project will define how motor training regulates myelin repair and identify potential targets to promote remyelination in neurodegenerative disorders such as multiple sclerosis.

Emerging evidence in animals and humans has shown that myelination is a plastic process driven by neuronal activity and experience, the so-called “adaptive myelination”. Enhancing neuronal activity or behavioural training could therefore constitute an alternative to promote remyelination in myelin-related diseases. In line with this, findings from the team have shown that neuronal activity promotes the preferential myelination of active axons and enhances myelin regeneration in a preclinical model of remyelination. Moreover, enhanced experience, voluntary exercise and motor learning increases both the proliferation of oligodendrocyte precursor cells (OPCs) and the generation of new OLs in control and after demyelination. This remodelling of myelin appears essential for the consolidation of learning and memory.
The effects of enhanced neuronal activity and behavior on nodal remyelination could partly account for the impact of physical activity on improving many symptoms in disease or on reducing the risk of developing neurodegenerative disorders. In MS patients, both exercise and cognitive training can reduce cognitive deficits and high intensity aerobic exercise can improve clinical outcome. It has even been suggested that exercise could be prescribed as a treatment in early MS stages. Yet, it is unclear whether different motor training paradigms stimulate myelin formation and which motor-driven mechanisms underlie neuronal activity changes and adaptive myelination in health and disease. Since complex motor training involves learning and coordination skills, we could speculate that it has a greater impact on finely tuning myelin production than physical exercise, albeit intense exercise.
Re-establishing an adapted myelin pattern along partially myelinated neurons is a challenge, and adapted physical activity and motor training could modulate this process. Built on the aforementioned studies and considerations, our hypothesis is that simple and complex motor training could differentially modulate adaptive myelination and neuronal plasticity, and that specific motor intervention following demyelination enhances myelin regeneration. Understanding motor-driven mechanisms that stimulate myelination and remyelination may pave the way for the development of new therapies that protect axons in many neurodegenerative diseases.

AIM 1: Motor-driven responses of oligodendroglia during myelin regeneration

Here, we will test how the activity of oligodendroglia in control and following demyelination is modulated in mice running in a regular wheel and in a complex wheel (wheel with rung gaps). In addition, we will use the cuprizone (CPZ)-induced model in which CPZ administration causes a global forebrain demyelination that is followed by a progressive spontaneous remyelination after CPZ withdrawal. The CPZ model has the advantage of reproducing many aspects of the MS pathology (increased astrogliosis and microglial activation, decreased OL density).
Intracellular Ca2+ signals of OPCs and OLs may play key roles during training because they affect OPC proliferation and differentiation and myelination. We hypothesize that simple and complex motor trainings differentially modulate oligodendroglia Ca2+ signals and OPC development in the motor cortex, contributing to myelin formation and regeneration. We expect that these signals will be particularly enhanced during complex motor learning, potentially improving remyelination to a larger extent. To visualize Ca2+ signals from OL lineage cells in vivo, we already set up the protocols to use the Miniscope-V4, a miniature microscope that can be mounted on the head of PDGFRaCreERT;GCaMP6f/f mice to image GCaMP6+ cells in behaving mice. We will train the mice to run on a simple or a complex wheel in control, during demyelination and remyelination while imaging oligodendroglia.
In addition to in vivo experiments, we will compare patch-clamp recordings of neurons in motor cortex slices from control mice and mice that have undergone motor training. These recordings will allow us to assess the functional changes induced by motor adaptations. Furthermore, we will perform immunostaining to evaluate the impact of motor training and the potential modulation of OPC proliferation, differentiation, and (re)myelination.

AIM 2: Mechanisms of motor-driven responses of oligodendroglia during myelin regeneration

In this aim, we will assess the molecular and cellular mechanisms underlying motor-driven remyelination during simple and complex motor training paradigms. Our unpublished data indicate that mitochondria Ca2+ flux plays an important role in OL regerenation in vivo (Maas et al., submitted). We hypothesis that the activation of neuronal networks during motor behavior promotes mitochondrial Ca2+ signals, enhancing the renewal of OL and myelin regeneration. First, we will analyze the number and distribution of mitochondria in OPCs and OLs during simple and complex motor training paradigms in control and demyelinated lesions using a mouse model expressing fluorescent mitochondria. Then, we will use proteomics to identify potential signaling pathways that mediate the effects of neuronal activity on oligodendroglial function and myelin repair. Protein expression profiles should help identifying potential mitochondria-related pathways contributing to activity-dependent OL lineage cell maturation and (re)myelination during motor training. This analysis will help uncovering molecular changes associated with motor training. We will validate protein candidates using Western blot and immunohistochemistry. This approach will provide a foundation for targeting specific pathways aiming at enhancing adaptive (re)myelination using pharmacology in this project and/or genetic manipulations in the future.
In conclusion, by establishing how mitochondrial Ca2+ signaling regulates oligodendrocyte behavior and promotes remyelination in response to motor-driven neuronal activity, we hope to identify targets for promoting OL regeneration and myelin repair after injury or in demyelinating diseases.

In vivo investigation using optogenetics and behaviour to compare the predictions of two competing theories of how iterative motor learning is implemented by the cerebellum.

The cerebellum is thought to contribute to the automation and optimisation of movements, particularly those that are rapid and complex. Such movements are often too difficult to control using real-time feedback. Indeed, sometimes the results of a movement can only be established after it is complete, when feedback would obviously be too delayed. In such cases, the only possible strategy is to learn through repetition: practice makes perfect.

There is a long-established theory linking cellular observations of plasticity of the hugely abundant granule cell--Purkinje cell synapses to the correction of movement errors signalled by climbing fibres, whose activity elicits "complex spikes" in Purkinje cells. However, our group analysed this classical theory and discovered that its algorithm is rather limited: it cannot account for the optimisation of arbitrary, complex movements. This issue is a manifestation of the "credit assignment problem", the difficulty of optimising a complex neural network on the basis of rudimentary and delayed sensory information.

We went on to propose an explicit implementation of a more general and capable algorithm, in simple terms a trial-and-error mechanism. The key feature of this implementation is that complex spikes play a dual role: in addition to conveying evaluation information (about movement errors), they would also constitute trial modifications of the motor command and thus of the movement and its error. The modifications are consolidated by plasticity if they improve the movement. Our theory predicted unexpected plasticity rules, which we were able to verify in vitro. This work is described in [1]. Since this publication, independent groups have tested two key elements of our theory: one verified that complex spikes can perturb movements |2] and the other replicated our unexpected plasticity rules [3].

Since our original publication, we have sought to test the competing theories in vivo. In particular we have performed chronic multielectrode recordings in the cerebellar cortex during eyeblink conditioning, a well characterised model behaviour of cerebellar motor learning. Preliminary analyses of these data support some aspects of our theory, but more direct, causal optogenetic interventions will be required to definitively distinguish the two theories. These experiments are the central objective of this PhD project. The key element that separates the two theories is the role of "perturbation" complex spikes occurring during the motor command. In the classical theory they play no role, whereas in our theory they are essential for the plasticity underlying the learning. The project is therefore to use optogenetic techniques to manipulate climbing fibre activity and to observe the effects of these interventions on learning-related changes of behaviour and cellular activity. The feasibility of a behavioural readout of optogenetic manipulations has already been established in precisely this experimental context [4]. The most direct experiment will involve inhibiting "perturbation" complex spikes without affecting "error" (or "evaluation") complex spikes, with our theory predicting the prevention of learning and the classical theory predicting no change.

These experiments will offer clear insight into the algorithms operating during iterative motor learning, linking in vitro, in vivo and computational approaches.

1. Bouvier et al. (2018) Cerebellar learning using perturbations. Elife 7:e31599. doi: https://doi.org/10.7554/eLife.31599
2. Muller et al. (2023) Complex spikes perturb movements and reveal the sensorimotor map of Purkinje cells. Curr Biol. 33(22):4869-4879. doi: https://doi.org/10.1016/j.cub.2023.09.062
3. Titley et al. (2019) Complex spike clusters and false-positive rejection in a cerebellar supervised learning rule. J Physiol. 97:4387-4406. doi: https://doi.org/10.1113/JP278502
4. Kim et al. (2020) A cerebello-olivary signal for negative prediction error is sufficient to cause extinction of associative motor learning. Nat Neurosci. 23:1550-1554. doi: https://doi.org/10.1038/s41593-020-00732-1

The vestibular system is essential to ensure proper balance, postural control and gaze stabilization. In vertebrates, brainstem vestibular efferent neurons project bilaterally back to the inner ear where they synapse on all hair cells. Until now however, the role of the vestibular efferent system (EVS) has remained undetermined. In this proposal, we investigate the following hypothesis: (i) EVS would differentially modulate vestibular activity in passive and active contexts. (ii) EVS could participate to rebalance the vestibular system in pathological situations.
The results of these studies have significant implications for rehabilitation strategies leveraging spinal circuit plasticity.

The vestibular system is a multimodal sensory motor system essential for postural stabilization, gaze control and spatial orientation. Within vertebrates, a brainstem efferent vestibular system (EVS) provides bilateral projections to the labyrinth, where efferent axons establish synaptic contacts with both type I and type II hair cells, as well as with vestibular afferent calyces and bouton terminals. While this circuitry is anatomically conserved across vertebrate taxa, its physiological role in mammals remains insufficiently resolved. Classical hypotheses propose that the EVS regulates receptor sensitivity, adjusts afferent firing dynamics, and participates in activity-dependent plasticity of peripheral vestibular encoding. However, direct causal evidence is scarce.
This PhD project targets a central working hypothesis: the EVS provides behavior-dependent modulation of peripheral vestibular output. During development, the EVS might participate to the calibration of vestibular sensori-motor reflex by adjusting the tuning of peripheral inputs. In addition, efferent feedback may dynamically adjust the gain of vestibular transduction during active behaviour, thereby improving sensorimotor performance when head movements increase. In pathological situations, efferent signaling may contribute to compensatory rebalancing in conditions of asymmetric input, such as unilateral vestibular lesions. Identifying such mechanisms would redefine the vestibular periphery as an active, state-dependent processing stage rather than a purely feed-forward sensory element.
Our recent work in rodents demonstrated that activation of spinal central pattern generators (CPGs) facilitates spino-ocular coupling during locomotion. These data parallel observations in Xenopus in which swim-driven CPG discharges modulate vestibular afferents via efferent pathways. Although the precise sources of inputs to mammalian EVS neurons are incompletely mapped, evidence suggests direct projections from spinal locomotor circuits to brainstem efferent nuclei. Our preliminary data in mice indicate a significant enhancement of vestibulo-ocular reflex (VOR) gain during running. The latency and magnitude of this modulation are compatible with rapid efferent-mediated alterations of hair cell or afferent terminal excitability, mediated by cholinergic receptors expressed peripherally.
The proposed research will combine anatomical, physiological and behavioral approaches to determine whether CPG-mediated modulation of EVS neurons causally drives locomotion-dependent changes in VOR gain. First, monosynaptic and polysynaptic projections to EVS neurons will be characterized using anatomical tracers and immunohistochemical markers defining neurotransmitter identity. This will allow localization of spinal or supraspinal neurons projecting to efferent nuclei and quantification of projection density along the rostrocaudal axis.
Next, efferent modulation at the labyrinth will be perturbed using transtympanic delivery of pharmacological agents targeting specific nicotinic receptors, muscarinic receptors, or associated modulatory pathways. Behavioral VOR responses will be quantified in behaving mice, while eye movements will be recorded at high temporal resolution. By systematically varying behavioural context and pharmacological conditions, we will determine whether VOR gain modulation scales with CPG output and whether it is modulated by EVS activity.
In parallel, in vivo calcium imaging of identified EVS neurons will allow real-time assessment of efferent recruitment during locomotion.
This integrative framework will provide the first causal demonstration, in mammals, that the EVS is a behavior-dependent sensory control system allowing rapid modulation of labyrinthine encoding. Such findings would significantly advance vestibular neuroscience by linking locomotor CPG output, efferent brainstem circuits, and peripheral sensory tuning. From a translational perspective, the results may open new avenues for rehabilitation after unilateral vestibular hypofunction, exploiting spinal circuit plasticity or pharmacological targeting of efferent pathways to restore vestibular function.

This project establishes the theoretical foundations of an integrative generative modeling framework connecting radiological, histological, and molecular representations of brain tumors. Building on advances in computational histopathology and spatial transcriptomics, it aims to formalize the geometric, molecular, and informational relationships linking tissue architecture to gene expression across scales and time. By developing trans-modal registration and longitudinal modeling strategies, the project will enable coherent integration of heterogeneous biomedical data. The ultimate goal is to create responsible generative AI models capable of simulating disease evolution, predicting treatment response, and revealing the multiscale mechanisms driving tumor progression, thereby bridging imaging, pathology, and genomics in a unified computational paradigm.

Extended Summary / Scientific Description
High-grade gliomas (HGG) and primary CNS lymphomas (PCNSL) exhibit pronounced spatial and temporal heterogeneity, driven by microenvironmental, metabolic, and immune processes. Clinically, MRI provides macroscopic structural and dynamical information; histopathology reveals cellular organisation and tissue architecture; spatial transcriptomics captures in situ molecular programmes at high resolution. Yet these modalities remain fundamentally disconnected—separated by differences in scale, resolution, spatial anchoring, and temporal sampling.
This PhD project aims to establish the theoretical and methodological foundations of an integrative generative modelling framework capable of connecting radiological, histological, and molecular representations of brain tumours within a unified computational paradigm. It builds upon advances in computational pathology, spatial transcriptomics, mesoscopic mechanistic modelling (high-dimensional PDEs), and next-generation physics-guided generative AI (GANs, diffusion models, hybrid physics-informed architectures).

Overall Objective
To design a responsible, physics-informed generative AI framework that can:
1. Synthesize missing modalities (virtual IHC, virtual spatial transcriptomics, virtual 7T MRI from 3T),
2. Register and integrate heterogeneous modalities into a coherent multiscale representation,
3. Model tumour evolution longitudinally,
4. Reveal the latent multiscale mechanisms linking imaging, tissue architecture, and gene expression.
In essence, the project seeks to move beyond mere correlations across modalities and identify the causal and mechanistic drivers underlying tumour architecture, molecular expression, and temporal dynamics.

Scientific Axes & Fundamental Questions
1. Identifiability theory for multiscale tumour modelling
The PhD will investigate the mathematical conditions under which latent mechanistic fields—diffusion, proliferation, infiltration, metabolic gradients, transcriptional programmes—are identifiable given incomplete or partially observed multimodal data.
This includes:
- structural invariants of mesoscopic PDEs,
- non-degeneracy conditions,
- sensitivity analysis,
- impact of noise, staining/sequence variability, and resolution gaps.
2. Physics-informed generative AI (PI-GANs and physics-guided diffusion)
The candidate will design hybrid generative models constrained not only by data but also by physical and biological laws, such as:
- diffusion-reaction equations,
- mass conservation,
- feasible kinetic regimes of gene expression and cell motility.
The models will generate:
- virtual IHC faithful to microstructural patterns,
- virtual spatial transcriptomics consistent with morpho-molecular gradients,
- 7T-like MRI from 3T, consistent with MR physics and suitable for registration.
3. Multimodal registration and cross-domain translation
The project involves developing a 3D WSI ? MRI registration framework enabling voxel-to-cell alignment, leveraging:
- virtual modalities to bridge appearance gaps,
- biophysical deformation priors,
- robust diffeomorphic registration tolerant to histology artefacts.
Cross-domain translation (MRI-like histology, histology-like MRI) will also be explored to stabilise similarity metrics and enhance cross-scale consistency.
4. Longitudinal modelling and digital twins
The candidate will implement patient-specific digital twins based on high-dimensional PDEs that model tumour–immune–stromal interactions, capable of predicting:
- tumour progression,
- infiltration patterns,
- therapy response,
- relapse vs. pseudoprogression.
Parameter estimation will rely on hybrid physics-guided learning and Bayesian filtering over longitudinal series.

Methodological Overview
- Mathematical formulation of structural constraints, invariants, identifiability conditions.
- Design of physics-informed GANs and diffusion models (MRI physics, tissue biophysics, transcriptional kinetics).
- 3D multimodal registration using landmarks, diffeomorphisms, and learned bridging modalities.
- Longitudinal inference with dynamic updating of latent fields and calibrated uncertainty.
- Evaluation via out-of-distribution stress tests, biological plausibility criteria, and clinically meaningful benchmarks.

Expected Scientific Contributions
- A new theoretical framework linking MRI, histopathology, and spatial omics through mechanistic identifiability.
- Generalisable methods for hybrid physics-informed generative modelling.
- Novel cross-scale registration strategies for neuro-oncology and beyond.
- Clinically relevant digital twins enabling interpretable predictions and counterfactual simulations.
- Transferability to other cancers (prostate, breast, lung) and to non-tumoural complex tissues.

Candidate Profile
The ideal candidate holds a Master’s degree in applied mathematics, machine learning, image analysis, modelling, or physics.
Desirable skills: deep learning (PyTorch), PDEs, optimisation, computational imaging, and strong motivation for interdisciplinary work at the interface between AI, imaging, biology, and clinical neuroscience.

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 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 determine whether this is unique to Human, we will use mice and gerbils 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 [1, 2]. 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 [2-4]. 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 rodent auditory cortex has never been systematically tested with controlled stimuli.

Experiment 1: To test whether there is a lateralization of spectrotemporal functions in mice and gerbils, we will use a framework developed for Human psychoacoustic studies to filter natural mouse vocalizations in the temporal and/or spectral domains [5]. 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 awake 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. Two rodent species will be used to 1) demonstrate a potential universality of lateralized functions, 2) benefit from various genetic models of deafness with the mouse model, and 3) study an organism (the gerbil) that extensively uses vocalizations for communication.

Experiment 2: However, several unanswered questions call for a more mechanistic description. Specifically, where does 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 asses the causal implication of the auditory cortex in processing spectral or temporal information, we will realize a behavior experiment. Animals will be trained to discriminate between a target vocalization and its filtered versions (same vocalization filtered in either the spectral or temporal domain). Once the animal has learned the task, we will optogenetically inactivate the left or the right auditory cortex and measure behavioral performance. Two optical fiber connected to the auditory cortices will alternatively or simultaneously inactivate these regions by activating the inhibitory neurons population while the animal is performing the behavioral task.

Experiment 3: To further analyze the underlying cortical network, 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 arises 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 use 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.

References:
1. Marlin, B.J., et al., Oxytocin enables maternal behaviour by balancing cortical inhibition. Nature, 2015. 520(7548): p. 499-504.
2. Levy, R.B., et al., Circuit asymmetries underlie functional lateralization in the mouse auditory cortex. Nat Commun, 2019. 10(1): p. 2783.
3. Oviedo, H.V., et al., The functional asymmetry of auditory cortex is reflected in the organization of local cortical circuits. Nat Neurosci, 2010. 13(11): p. 1413-20.
4. Neophytou, D., et al., Differences in temporal processing speeds between the right and left auditory cortex reflect the strength of recurrent synaptic connectivity. PLoS Biol, 2022. 20(10): p. e3001803.
5. Flinker, A., et al., Spectrotemporal modulation provides a unifying framework for auditory cortical asymmetries. Nat Hum Behav, 2019. 3(4): p. 393-405.

This PhD project investigates why music is so universally enjoyed by testing the intrinsic reward for learning hypothesis. It proposes that music engages ancient neural mechanisms linking prediction, surprise, learning, and reward, driving listeners to prefer sounds that slightly challenge their internal models. Project 1 will model how musical prediction and learning dynamics shape listening choices using behavioral data and AI-based prediction models. Project 2 will examine neural correlates of this process with intracranial EEG, focusing on interactions between auditory cortex, hippocampus, and amygdala within the VTA–hippocampal loop. Together, these studies aim to clarify how sound transforms into pleasure and learning, advancing both theoretical neuroscience and personalized music-based interventions.

Across every human culture and era, music has played a key role in society. However, its evolutionary purpose remains elusive. What neural mechanisms link the individual processing of patterned sounds to the rewarding and social experiences that music generates? The near-universality of a seemingly unnecessary auditory stimulus suggests that understanding how music influences the brain may reveal fundamental principles of neural function and how these mechanisms motivate cultural behaviors.

At the individual level, the link between sound and reward is often hypothesized to rely on prediction. These theories propose that the surprise arising from an unexpected note or chord generates musical tension, which is then translated into a rewarding experience once resolved. Yet it remains unclear why prediction should drive reward, since predictive processing is normally associated with efficiency—seemingly at odds with the pursuit of unnecessary surprise.

Interestingly, both prediction and reward have been separately linked to a third domain: learning and memory. For example, abstract pseudoword learning has been causally linked to dopaminergic systems: generating reward from a process with no extrinsic value. Similarly, structured surprise has been shown to enhance statistical learning, sequence-order memory, and the chunking of memory events. These findings suggest that learning may be a key mediator connecting prediction to reward.

I therefore hypothesize that the drive to experience music is motivated by an intrinsic reward for improving our internal models—that is, for enhancing our understanding of our own experiences. This mechanism likely extends beyond music, serving as an evolutionarily useful principle of human–environment interaction: predictions trigger surprise; surprise enables learning; learning leads to reward; and reward consolidates learning into memory, improving future predictions. Music provides a spark that activates these deeply ingrained evolutionary mechanisms, encouraging a better understanding of our environment. A key principle of this theory concerns how listeners select music, suggesting that individuals seek music whose complexity lies just beyond their internal models. However, music-listening behaviors and musical navigation have not been well studied in an openly accessible manner.

The proposed PhD thesis will investigate the intrinsic reward for learning hypothesis in two complementary projects focused on musical exploration.

In Project 1, we ask which features of musical prediction, reward, and learning guide listeners’ choices to continue listening to or discard a specific piece of music. Answering this question will provide deeper insight into how abstract sound sequences generate rewarding experiences and may lead to AI models capable of self-curating training datasets. To address this question, we will study the music-listening behaviors of a large cohort of participants by collecting an online behavioral dataset. Listeners will navigate a musical database through a series of possible actions: continue listening, restart the current piece, skip to the next piece, or switch to a different genre.

We will use recently developed AI models, fit to participants’ reported musical histories, to estimate how the dynamics of prediction, learning potential, and experienced reward drive action selection. Our ultimate aim is to develop a new computational model of music selection that integrates these factors. We expect this model to be guided by principles of exploitation versus exploration: an initial exploration phase allows listeners to find a piece or genre consistent with their internal models; subsequently, they will select music that challenges and refines those models, maximizing reward. Over time, preference for a given piece or genre may wane as predictive models adapt to its statistical features and learning potential diminishes.

In Project 2, we will investigate the neurobiological correlates of this process. Previous research examining the intrinsic reward for learning hypothesis in other domains has identified the VTA–hippocampal loop as a key mechanism for dopamine release that consolidates learning. We will test the hypothesis that prediction dynamics optimizing learning trigger dopamine release in the hippocampus, thereby enhancing sequence memory for specific levels of surprise.

We will collect intracranial EEG (iEEG) data from epileptic patients with clinically implanted electrodes as they perform a pared-down version of the behavioral experiment—making individual choices to re-listen to 30-second clips, continue to the next part of a piece, or begin a new one. Our analysis will focus on interactions between auditory cortical processing and activity in deep brain structures—particularly the hippocampus and amygdala—which are frequently targeted in clinical settings. Using both our established AI models and the new choice-selection framework, we will investigate how surprise encoding in the auditory cortex translates into learning and reward signals in the hippocampus and amygdala, and whether activity in these regions can predict participants’ decisions to continue or skip their current music. We highlight these areas for their accessibility via iEEG depth electrodes and their close functional and anatomical connections within the VTA–hippocampal loop.

The findings from this project will deepen our understanding of the interface between sensory and pleasurable experiences—a question fundamental to human experience—by emphasizing learning as a key motivator of behavior. Furthermore, they may allow us to predict musical preferences based on individual listening histories, enabling the development of more effective, personalized music-based clinical interventions. Future work will aim to identify how perceptual, predictive, reward, and learning disorders alter engagement with music, ultimately guiding the tailored deployment of musical interventions to improve clinical outcomes.

The project will be housed within the Human and Artificial Perception (HArP) team at the Institut de l’Audition, a center of the Institut Pasteur. The HArP team comprises passionate and supportive researchers at all career stages, dedicated to developing explainable computational models that advance our understanding of how natural sequences such as speech and music are processed. The selected PhD candidate will join this team as well as neighboring groups in human cognitive neuroscience, contributing to a vibrant community engaged in both fundamental and translational research in the speech and hearing domains.

Our lab recently demonstrated that damaging mutations in the SMARCC2 gene were associated with bipolar disorders. Interestingly, numerous de-novo mutations in SMARCC2 have already been reported in individuals with other neurodevelopmental disorders, but no study has shown the pathophysiological impact of such mutations. Our project aims to understand the pathophysiological mechanisms resulting from mutations in the SMARCC2 gene by studying their impact on the early steps of brain development, structure, and function. In this project, the candidate will generate and characterize induced pluripotent stem cell-derived neurons, brain organoids and assembloids carrying or not mutations identified in patients. In addition to the phenotypic characterization, the candidate will conduct ATAC-Seq and single-cell RNA-seq experiments to identify the genes targeted by the SWI/SNF complex and those deregulated when SMARCC2 is mutated in affected individuals.

Bipolar disorder (BD) is a major public health concern for which a high heritability has been widely demonstrated. Recent genome-wide association studies (GWAS) on tens of thousands of subjects have reported several loci associated with the disease with an additive effect of many genetic vulnerability variants, each having a low effect size (1). However, common polymorphisms explain only 7% of the genetic variance in BD suggesting the major heritability could reside in highly penetrant rare damaging variants. Using multiplex families with early onset forms of BD, we have recently identified several genes with rare damaging variants segregating with the disease (2,3). An enrichment pathway analysis highlighted that these genes encoded proteins involved either in neural gene expression regulation or in neural development and morphogenesis or both, suggesting for the first time that familial BD with an early age of onset could be a neurodevelopmental disorder (3). Among the genes identified in these multiplex families, we observed that SMARCC2, encoding the SWI/SNF related, Matrix-Associated, actin dependent Regulator of Chromatin subfamily C member 2, showed a higher frequency of damaging variants in individuals with BD than in the general population (3). Interestingly, damaging variants in this gene had also been reported in individuals with a rare disease named Coffin-Siris syndrome and autism spectrum disorders, but no functional studies have been performed yet to assess the impact of these variants and explain the variability in the phenotypic outcome.
The chromatin remodeling complex SWI (switch mutants)/SNF (sucrose non-fermenting) plays an essential role in the regulation of gene expression. This ATP-dependent multiprotein complex, also known as BRG1-associated factor (BAF), is composed of 15 proteins that interact to modulate the nucleosome and change chromatin conformation and accessibility. Among these proteins, BAF170, a common core protein of the BAF complex, which directly interacts with BRG1, is encoded by SMARCC2. SMARCC2 plays a crucial role in embryogenesis and corticogenesis (4), suggesting that amino acid changes in its protein could result in brain abnormalities and contribute to the BD vulnerability as well as to other neurodevelopmental disorders.
Modeling bipolar disorders in animals is a real challenge to understand the pathophysiology of the disease, particularly for studying early stages of brain development. The aim of our project is thus to use neuron culture and brain organoids and assembloids to study early brain development and assess the impact of disease associated SMARCC2 mutations identified in individuals with BD. First, the candidate will study two BD-associated mutations to compare the phenotypic consistency of SMARCC2 mutations identified in patients. These mutations will be introduced in two commercial lines of human induced pluripotent stem cells (hIPSCs) that are already available in the lab, using CRISPR/Cas9 approach to have isogenic lines varying only for the mutations of interest. These mutated lines will be compared to the wild-type lines. Two hIPSC lines will be studied to confirm that the mutation effect is independent of the genetic background. Then, mutated and non-mutated hIPSCs will be differentiated in 2D-cell culture into human neurons and into 3D-cortical organoids to assess the impact of mutations on neuron and dendritic growth as well as on more complex structure. The candidate will compare the gene expression and protein level of SMARCC2 and its SWI/SNF complex partners between wild-type and mutants cells/organoids. The neurite length of differentiating neurons will be measured with live cell imaging using an Incucyte® Live Cell analysis instrument. In addition, immunocytochemistry will be performed on neurons after several days in culture using neuronal markers (e.g. TUJ1, DCX, MAP2) to monitor the neuronal differentiation. In parallel, a higher degree of organization and cortical differentiation as well as synapse characterization and quantification will be analyzed by immunohistochemistry on 3D-brain organoids. Organoids will be fixed at multiple steps of differentiation and incubate with undifferentiation markers (e.g., PAX6 and SOX2), neuron specific markers (e.g. MAP2, TUJ1) and synaptic specific markers (e.g. SV1, VLGUT1, VGLUT2, VGAT, GPHN).
Phenotypic rescue of mutated neurons and organoids will be tested using common mood regulators, e.g. lithium and valproic acid, and other psychotropic drugs that will be added to the cell culture media.
In parallel, the candidate will dedifferentiate hair follicle cells collected from families in which SMARCC2 mutations have been identified and will differentiate them in neurons and brain organoids in order to assess the impact of the patients’ genetic background on the phenotypic expression of the disease-associated mutation. We have already obtained the ethical and legal authorization to collect hair follicles in these families. These mutations will then be corrected in vitro in patient’s cells using CRISPR-Cas9 approach to demonstrate that the observed phenotype results from the mutation.

As a second step, as the SWI/SNF complex regulates the expression of many genes by modulating the chromatin architecture, the candidate will perform single nuclei Assay for Transposase Accessible Chromatin combined with sequencing (ATAC-Seq) as well as single cell RNA sequencing of IPSCs-derived brain organoids to identify the molecular pathways affected in mutated organoids. This will result in the identification of chromosomal loci for which the chromatin accessibility changes with the mutations in SMARCC2, as well as the genes and molecular pathways for which the expression is affected by these changes. The candidate will thus compare in a cell-type specific manner, the loci with open chromatin regions (OCRs) in the genome between organoids with and without SMARCC2 mutations.

Finally, as bipolar disorder is known to show impairment in cortico-striatal connectivity, the candidate will also differentiate hIPSC into region-specific brain organoids that resemble the developing human striatum and cortex. He or she will then assemble these organoids in three-dimensional cultures to form cortico-striatal assembloids and characterize axon projections and interneuron migration using viral tracing and functional assays (5). Then, the candidate will assess whether the SMARCC2 mutations affect the connectivity between the two brain areas, as observed in subjects with BD.

This project is already founded by the FondaMental Foundation, which have identified a sponsor for the running costs of the project. Altogether, these results should allow the identification of molecular pathways that are affected in individuals with mutations in SMARCC2 and should shed new light on the pathophysiological mechanisms underlying BD.

1. Mullins N et al. Nat Genet. 2021; 53(6): 817-29. doi: 10.1038/s41588-021-00857-4.
2. Sitbon J et al. Mol Psychiatry. 2022; 27(2): 1145-57. doi: 10.1038/s41380-021-01151-9.
3. Courtois E et al. Transl Psychiatry. 2020; 10(1): 124. doi: 10.1038/s41398-020-0783-0.
4. Tuoc TC et al. Dev Cell. 2013; 25(3): 256-69. doi: 10.1016/j.devcel.2013.04.005.
5. Miura Y et al. Nat Biotechnol. 2020; 38(12):1421-1430. doi: 10.1038/s41587-020-00763-w.

This PhD project will investigate how the auditory thalamocortical circuit contributes accurate to learning and perception by examining its synaptic plasticity and dynamic regulation during behavior. Using two-photon calcium imaging in mice performing auditory discrimination tasks, we will track activity in thalamic inputs and cortical neurons across different learning stages to reveal how experience shapes circuit function. A schizophrenia pathology–related mouse model will be used to identify alterations in thalamocortical communication associated with impaired perception. In a second phase, targeted modulation of thalamic activity will test its causal influence on cortical plasticity and behavioral adaptation. Together, these studies will uncover fundamental principles governing how thalamic input drives cortical learning, providing mechanistic insight into the cellular basis of adaptive cognition.

Background and Rationale
Learning and memory rely on the brain’s capacity to modify synaptic connections and dynamically adjust information flow across cortical and subcortical networks. The thalamus plays a critical role in these processes by regulating how sensory signals are integrated, filtered, and relayed to cortical areas. Beyond its traditional role as a sensory gateway, the thalamus participates actively in higher-order functions such as associative learning, contextual modulation, and cognitive flexibility.
Within the auditory system, the secondary auditory thalamocortical (2ATC) pathway — linking the auditory thalamus (AT) with the secondary auditory cortex (2AC) — is a central hub for integrating sensory experience with learned associations. This circuit also communicates extensively with limbic and frontal regions, positioning it as a key structure for transforming perceptual input into memory and decision-relevant representations.
Alterations in thalamocortical communication and excitatory–inhibitory balance have been observed in several conditions associated with impaired sensory integration and cognitive deficits, including schizophrenia. Understanding the synaptic and circuit mechanisms that underlie these disturbances could inform strategies for circuit-level interventions, but fundamental questions remain about how thalamocortical plasticity supports normal learning and memory processes.
This project will therefore investigate how activity-dependent plasticity within the auditory thalamocortical circuit contributes to associative learning, and how its disruption in a schizophrenia pathology–related mouse model affects information processing and behavioral adaptation.

Hypotheses and Objectives
We hypothesize that the auditory thalamocortical circuit exhibits bidirectional plasticity that encodes learning-related changes in stimulus–outcome associations, and that disruption of this mechanism leads to altered sensory and cognitive performance. We further propose that targeted modulation of thalamic input can influence cortical plasticity and behavioral outcomes, revealing causal links between thalamic dynamics and learning efficacy.
To address these hypotheses, the project is organized into two experimental aims.

Aim 1 – Characterizing Thalamocortical Plasticity During Auditory Learning
The first aim seeks to define how synaptic and circuit-level plasticity within the 2ATC pathway evolves during associative learning. Mice will be trained in an auditory Go/No-Go discrimination task where specific tones predict rewarding or aversive outcomes. Using two-photon calcium imaging, we will record activity from presynaptic boutons of thalamic projections and postsynaptic dendrites of cortical neurons during task performance.
This approach will allow longitudinal tracking of structural and functional plasticity across different phases of learning — acquisition, consolidation, and reversal. By comparing activity patterns between control animals and a schizophrenia pathology–related mouse model, we will determine how changes in synaptic strength and coordination reflect learning performance and memory flexibility.
The analysis will focus on how thalamic inputs shape cortical representations over time, and how their modulation corresponds to behavioral indices such as discrimination accuracy, response latency, and learning speed. These results will establish a cellular and functional map of thalamocortical plasticity during auditory learning.

Aim 2 – Causal Modulation of Thalamic Activity and Its Impact on Learning and Plasticity
The second aim will test the causal contribution of thalamic activity to cortical plasticity and learning behavior. To this end, we will combine two-photon imaging with focal electrical stimulation of the auditory thalamus during behavior.
By systematically varying stimulation timing and frequency, we will assess how enhanced or suppressed thalamic drive influences cortical dynamics and behavioral performance. Longitudinal imaging will reveal whether specific patterns of thalamic activation can promote functional synaptic remodeling in the auditory cortex, thereby accelerating or stabilizing learning.
In parallel, we will compare these effects between control and schizophrenia pathology–related models to determine whether circuit-level interventions can compensate for impaired plasticity. This aim will thus provide mechanistic insight into how thalamic input regulates cortical learning rules and whether this regulation remains flexible under altered network conditions.

Expected Outcomes and Significance
This project will elucidate fundamental principles of thalamocortical plasticity in learning and memory. Specifically, it will:
1. Define how coordinated synaptic changes in the auditory thalamocortical circuit underlie associative learning.
2. Identify how circuit dysfunction affects sensory–cognitive integration in a validated model of schizophrenia-related pathology.
3. Establish causal evidence that modulation of thalamic activity can influence cortical plasticity and behavioral adaptation.
Beyond its relevance to auditory processing, the project will contribute broadly to our understanding of how thalamic circuits support flexible learning and memory formation. These findings may also provide a mechanistic foundation for future circuit-based interventions targeting cognitive and perceptual disturbances, though the primary focus remains on basic neurophysiological mechanisms.

TALES (Tracking And Leveraging Engaging Story-elements) aims to uncover the computational bases of narrative pleasure — a key form of intrinsic reward underpinning motivation and mental health. Combining experimental psychology, behavioral economics, and artificial intelligence, the project will quantify how suspense, surprise, and emotional valence shape the enjoyment of stories. Three work packages structure the thesis: (1) experimental measurement of the links between these features and pleasure; (2) use of large language models to predict story success from temporal profiles of suspense, surprise, and valence; and (3) investigation of how sensitivity to these signals varies across individuals and relates to depression and addiction. TALES fits within the DIM C-BRAINS axis on cognition, behavior, and computational modeling.

What makes stories — in books, films, or games — so compelling that people devote hours to them, often at the expense of other activities? Stories are intrinsically rewarding: they provide pleasure for their own sake, not because of external outcomes. Yet, despite centuries of research in narratology and psychology, the quantitative features that make stories rewarding remain unknown. The TALES project aims to identify and model the computational features that drive narrative pleasure, focusing on three key dimensions: emotional valence, suspense, and surprise.

*Scientific objectives

TALES will combine approaches from experimental psychology, behavioral economics, and artificial intelligence to investigate how these dimensions interact to shape intrinsic pleasure. The central hypothesis is that suspense and surprise interact with affective valence to determine enjoyment: highly suspenseful narratives resolving positively, as well as (positive) surprising outcomes should be the most pleasurable on average. The project has three main objectives:

1)Quantify the contribution of suspense, surprise, and valence to subjective pleasure in controlled experimental tasks and their textual equivalents.

2) Test the ecological validity of these features by ausing large language models to published stories, assessing how suspense, surprise, and valence predict real-world success (downloads, ratings).

3) Investigate interindividual variability in sensitivity to these intrinsic rewards and its relation to mental health dimensions, particularly depression and addiction.

*Background and positioning

While narratology and psychology have proposed qualitative notions such as emotional valence, suspense, and surprise, these concepts lack formal definitions that allow quantitative testing. Recent developments in computational modeling provide a way forward. In economics, suspense can be defined as the expected change in belief — the difference between what we currently believe and what we expect to believe in the near future. Surprise corresponds to the actual change in belief once new information is revealed. These definitions can be formalized mathematically and applied to sequences of events, such as story plots.

Importantly, positive surprises (prediction errors associated with good outcomes) are known to be intrinsically rewarding, while suspense can be conceptualized as the anticipation of such belief changes. TALES will integrate these notions into a unified quantitative model of narrative pleasure. Beyond its contribution to fundamental knowledge, the project seeks to bridge cognitive modeling, natural language processing, and computational psychiatry, in line with the interdisciplinary vision of the DIM C-BRAINS initiative.

*Methodology and work plan

**Work Package 1 – Experimental quantification of intrinsic reward features

In WP1, participants will complete behavioral tasks designed to generate well-controlled sequences of events that vary in suspense, surprise, and valence. Using a card-based paradigm inspired by the Black-Jack game, the task will allow independent manipulation of these dimensions. Participants will rate their feelings of suspense (before outcomes), surprise (after outcomes), and pleasure (liking). The same paradigms will be adapted into narrative (text-based) versions to test generalization to story contexts. The main prediction is that objective suspense and positive surprise (as defined by computational models) will best predict subjective enjoyment.

A second experiment will assess wanting rather than liking: participants will freely choose between lotteries differing in expected value, variance, and ambiguity. Preference for suspenseful or surprising options will reveal the motivational component of intrinsic reward.

**Work Package 2 – Computational modeling of story pleasure using neural networks

WP2 will extend these principles to real-world stories. Using pre-trained Bayesian recurrent neural networks, the project will compute paragraph-level time series of suspense, surprise, and valence in large text corpora (e.g., Project Gutenberg, IMDb subtitles). The model will estimate suspense as the expected divergence between predicted and future sentence distributions, and surprise as the realized deviation. Valence will be fine-tuned using affective annotation datasets.

These features will be used to predict the success of published books and movies (e.g., downloads, ratings). The hypothesis is that stories with higher proportions of suspense and positively valenced surprises will achieve greater popularity, validating the intrinsic reward framework on ecological data.

**Work Package 3 – Individual differences and links to psychopathology

In WP3, a large-scale behavioral study (N > 200) will examine how individual sensitivity to suspense and surprise relates to mental health. Participants will perform the tasks from WP1, alongside standardized questionnaires assessing depression, apathy, anhedonia, and addictive behaviors. Dimensionality reduction (e.g., PCA) will be applied to identify latent components. It is hypothesized that hyposensitivity to suspense and positive surprise will correlate with depressive symptoms, while hypersensitivity will correlate with addictive tendencies. This approach aligns with the Research Domain Criteria (RDoC) framework, emphasizing computational mechanisms underlying transdiagnostic dimensions of psychopathology.

*Expected outcomes and significance

TALES will provide the first quantitative framework linking narrative structure, intrinsic reward, and mental health. The project will bring:

**A computational model of narrative pleasure, combining psychological theory with mathematical definitions of suspense and surprise.

**Validated neural network tools for estimating narrative features from text, applicable to cultural analytics and affective computing.

**New insights into mental health, by identifying how sensitivity to intrinsic rewards varies across individuals and contributes to depression or addiction.

Beyond academic impact, TALES addresses societal questions about how media engage human cognition and emotion. Understanding the mechanisms that make stories pleasurable could inform education and digital media design.

*Integration within DIM C-BRAINS

This PhD project fits squarely within the DIM C-BRAINS themes on cognition, behavior, and computational modeling. It bridges psychology, computational psychiatry, and AI to explore how cognitive and affective mechanisms drive engagement and well-being. Hosted at the Centre d’Économie de la Sorbonne (Paris 1) and involving collaborations with experts in cognitive psychology, decision theory, and language modeling, TALES embodies the interdisciplinary spirit of the DIM network.

By quantifying what makes stories rewarding and how this relates to mental health, TALES will contribute to understanding the neural and computational bases of intrinsic motivation—a central challenge for cognitive science and psychiatry.

This PhD project investigates how bodily states shape emotional positive and negative valence by integrating systems neuroscience, physiology, and behavior. It examines how gut-brain pathways, metabolic cues, and interoceptive hubs (NTS, PBN) modulate neural circuits underlying reward and aversion, with a focus on mesolimbic dopamine dynamics. Using fiber photometry, chemogenetics, metabolic profiling, and behavioral assays, the project will determine how vagal inputs influence valence, how metabolic dysregulation (e.g., obesity) reshapes reward and aversion processing, and how brainstem-VTA projections causally integrate peripheral signals. The work aims to build a mechanistic framework for body-brain regulation of valence and provide insights relevant to psychiatric and metabolic disorders.

Emotional valence, the ability to interpret stimuli as rewarding or aversive, emerges from tightly coupled interactions between the body and the brain. Rather than being solely a cortical construct, valence arises from dynamic loops integrating peripheral physiological signals, interoceptive processing, and neuromodulatory circuit activity. Despite major advances in systems neuroscience and physiology, we still lack a unified mechanistic framework explaining how bodily inputs shape neural encoding of positive and negative valence, and how dysregulation of these body-brain interactions contributes to psychiatric and metabolic disorders.
This PhD project aims to fill this gap by dissecting the pathways through which peripheral physiological states, especially those arising from the gut-brain axis and autonomic system, modulate neural circuits involved in reward and aversion.

Background and significance
The brain continuously integrates signals from the body via visceral afferents, hormones, immune mediators, and metabolic cues. These inputs converge on key regions implicated in motivation, learning, and emotional processing, including the brainstem, hypothalamus, and midbrain dopaminergic system, which are all intricately interconnected.
Recent work demonstrates that gut microbial metabolites, vagal sensory neurons, and metabolic hormones can powerfully modulate dopamine release, aversive learning, stress responses, and reward processing. Conversely, physiological alterations such as inflammation, obesity, or microbiome dysbiosis bias animals toward anhedonia, anxiety, compulsive reward seeking, or heightened aversion. Emotional states also modulate bodily processes such as cardiac rhythms, gastric motility, and energy expenditure, making valence a fundamentally bidirectional phenomenon.
A body-brain perspective therefore offers a powerful framework for understanding a wide range of disorders (including depression, anxiety, addiction, binge eating, obesity, metabolic syndrome, and functional gastrointestinal disorders) all of which feature disruptions in interoception and valence encoding.

Aims and hypotheses
This PhD project will investigate how peripheral physiological states modulate neural circuits encoding positive and negative valence. The work is structured around three major aims:
1. Identify how gut–brain pathways influence reward and aversion circuits.
Hypothesis: Gut-brain vagal sensory neurons differentially modulate mesolimbic dopamine events and aversion-related circuits (e.g., central amygdala, hypothalamus, brainstem), biasing behavior toward positive or negative valence. We will compare how gut-driven signals regulate mesolimbic dopamine during rewarding stimuli (e.g., food-related cues) versus aversive internal states (e.g., visceral malaise).
2. Characterize how metabolic states reshape valence processing.
Hypothesis: Physiological and pathological metabolic states, including high-fat diet exposure and obesity, alter the balance between reward and aversion by modifying dopamine responsivity, brainstem integration, and interoceptive representations.
We will track how short- and long-term metabolic perturbations rewire gut-to-brain pathways involved in positive and negative valence.
3. Establish the causal role of interoceptive hubs in integrating peripheral signals.
Hypothesis: Interoceptive hubs such as the nucleus tractus solitarius (NTS) and parabrachial nucleus (PBN) integrate peripheral inputs and are required for coherent valence states. Using chemogenetic or optogenetic manipulation of NTS/PBN-to-VTA projections, we will test whether these circuits dynamically gate positive and valence depending on bodily states.

Methodological approaches
To address these aims, the project will combine cutting-edge systems neuroscience tools with detailed physiological and metabolic measurements.

Neural circuit monitoring: Multi-sites single and/or dual fiber photometry will be used to monitor calcium and dopamine dynamics in key brain regions involved in valence processing, including the mesolimbic system (VTA-to-NAc), and the brainstem regions, PBN and NTS, projecting to the VTA. These recordings will allow quantification of spontaneous and stimulus-evoked neural activity under varying bodily states.

Circuit perturbation: Chemogenetic tools (DREADDs) will enable causal manipulation of defined neural populations. For example, activating or inhibiting vagal sensory neurons, and/or PBN/NTS-to-VTA projections will reveal their role in shaping positive and negative valence responses. Optogenetics may complement these manipulations where temporal precision is required.

Peripheral physiological measurements: The project will integrate metabolic profiling (nutrient and hormone fluctuations, energy metabolism and indirect calorimetry), gastrointestinal motility and vagal activity monitoring.

Behavioral paradigms: A battery of assays will measure positive valence (reward learning, operant conditioning, approach behavior, sucrose preference) and negative valence (fear conditioning, avoidance tasks, stress responses). Additional paradigms will combine interoceptive manipulations with ambiguous or conflicting stimuli to probe integration mechanisms.

Planned experiments
1. Gut-brain modulation of reward and aversion.
Mice will undergo chemogenetic or metabolic manipulations of vagal sensory pathways while dopamine and brainstem activity is recorded during valence tasks. Expected outcomes include bidirectional modulation of dopamine dynamics and shifts in preference or avoidance behaviors.

2. Impact of metabolic dysregulation on valence circuits.
Using diet-induced obesity, we will test how altered peripheral states shape the neural encoding of positive (food reward) and negative (visceral malaise) valence. We predict altered dopamine responsivity during reward or aversive tasks and maladapted NTS/PBN-to-VTA circuit activity patterns, mirroring symptoms seen in metabolic-psychiatric comorbidity.

3. Brain integration of interoceptive signals.
Targeted DREADD manipulation of PBN- and NTS-to-VTA projections will assess their necessity for integrating bodily signals into valence states. Using simultaneous recordings from peripheral sensors and central circuits, we will model the directionality and temporal structure of body-brain interactions.

Expected results and impact
This project will produce a mechanistic map of the pathways through which bodily states influence emotional valence. It will clarify:
• How specific vagal and metabolic inputs bias neural computations in reward and aversion circuits;
• How interoceptive hubs integrate these signals;
• How pathological physiological states reshape emotional processing.
Beyond fundamental insights, the work has clear translational potential for psychiatry, neurology, and metabolic medicine. By identifying peripheral modulators of valence, it could inspire new therapeutic strategies based on vagal manipulation, microbiota-targeted interventions, metabolic control, or brain–body neuromodulation.

Training environment
The project offers multidisciplinary training at the intersection of neuroscience, physiology, and behavioral neurosciences. The student will gain expertise in neural circuit manipulation, in vivo calcium and dopamine imaging, metabolic profiling, and advanced behavioral paradigms. The diverse methodological toolbox ensures excellent career development prospects in both academic and biomedical research.

Conclusion
This PhD project will advance our understanding of how the body shapes emotional valence and how disruptions in body-brain communication contribute to neuropsychiatric and metabolic disorders. By integrating circuit neuroscience with physiology, it aims to build a comprehensive framework of affective regulation grounded in whole-organism physiology and physiopathology.

Early-life exposure to obesogenic diets profoundly impacts metabolic health, yet molecular mechanisms underlying developmental programming of metabolic circuits remain poorly understood. This project asks: How do obesogenic diets consumed early in life, i.e. during development alter transcriptional and translational responses across brain, gut, and adipose tissue?

Using Drosophila complemented by mouse studies, the student will track nascent transcription and translation in vivo across tissues. This will reveal tissue-specific temporal signatures of molecular plasticity, identify critical developmental windows, and uncover conserved mechanisms governing metabolic circuit development.

This project explores new territory through multinational collaborations on adipose tissue innervation and transcription-translation dynamics. Findings will provide insight into inter-organ coordination during metabolic programming with implications for disease prevention.

SCIENTIFIC BACKGROUND

The rising prevalence of metabolic disorders represents a major global health challenge, with early-life environmental exposures playing a critical role in disease susceptibility. The "developmental origins of health and disease" hypothesis posits that nutritional and metabolic challenges during critical developmental windows program lifelong physiological responses. However, the molecular mechanisms underlying this developmental programming remain poorly understood, particularly regarding how multiple metabolic organs coordinate their responses.

The brain-gut axis exemplifies bidirectional communication between the central nervous system and peripheral organs in metabolic control. Our team, has demonstrated that enteric neurons exhibit remarkable functional plasticity in response to environmental challenges, adapting their activity to meet changing energy demands. Peripheral organs, including adipose tissue, are extensively innervated by the nervous system, which plays crucial roles in metabolic regulation. Yet, we know surprisingly little about how these multi-organ neural circuits are established during development and how early metabolic challenges alter their formation and function.

A gap in our understanding lies within the temporal dynamics of molecular responses. Recent breakthrough studies have revealed that transcription and translation of the same gene can be temporally uncoupled, with transcription occurring days before translation in some developmental contexts. For example, in neocortex development, specific transcription factors can be transcribed days before their translation, restricting protein expression to particular layers. This temporal control has profound implications for cell fate specification and functional outcomes. However, such transcription-translation dynamics have never been examined in the context of metabolic organ development or in response to environmental metabolic stressors.

Furthermore, obesogenic diets during development trigger adaptations across multiple organs simultaneously—brain circuits regulating appetite and energy expenditure, gut nutrient sensing and hormone secretion, and adipose tissue storage and endocrine function. Understanding how these organs coordinate their molecular responses, and whether coordination is disrupted by metabolic stress, represents a major unresolved question with implications for understanding metabolic disease etiology.

RESEARCH OBJECTIVES AND HYPOTHESES

This PhD project will investigate how developmental exposure to obesogenic diets alters the temporal dynamics and inter-organ coordination of transcriptional and translational responses in metabolic tissues. We will interrogate if:

1. Obesogenic diets during development trigger tissue-specific changes in transcription-translation timing that persist beyond the exposure period
2. Different metabolic organs exhibit distinct temporal response profiles, with varying rates of molecular plasticity
3. Inter-organ coordination of molecular responses is disrupted by developmental dietary stress, contributing to long-term metabolic dysfunction
4. Core molecular mechanisms governing these responses are evolutionarily conserved across Drosophila and mammals

Specific objectives:

1. Map transcription and translation dynamics in brain, gut, and adipose tissue during normal development and following obesogenic diet exposure
2. Identify critical developmental windows when metabolic organs are most susceptible to dietary programming
3. Characterise tissue-specific molecular signatures and inter-organ coordination patterns
4. Determine whether transcription-translation uncoupling contributes to developmental metabolic programming
5. Validate key findings in mammalian systems through collaborative studies

METHODOLOGY and EXPERIMENTAL APPROACH

This project leverages Drosophila melanogaster as the primary experimental model, offering genetic tractability, well-defined development, rapid generation time, and remarkably conserved metabolic pathways. Key findings will be validated in mouse models through established collaborations, enabling cross-species comparison and assessment of evolutionary conservation.

Developmental obesogenic diet paradigm. We will expose developing larvae to high-sugar, high-fat diets at defined developmental stages, building on our established protocols for dietary manipulation in Drosophila. This will include acute exposures at specific developmental windows as well as chronic exposure paradigms to model different aspects of human dietary patterns. The student will employ cutting-edge molecular barcoding approaches developed by collaborators expert in transcription-translation timing to simultaneously label and track newly synthesized mRNA and protein molecules in vivo. This methodology enables hour-timescale resolution of transcription and translation activity in whole organisms throughout development. The student will apply and potentially adapt these techniques to examine brain, gut, and adipose tissue simultaneously.

Using genetic tools available in Drosophila, we will dissect tissue-specific responses through targeted molecular profiling of specific neuron populations including enteric neurons, fat-innervating neurons, and central appetite circuits. We will perform quantitative analysis of transcription-translation lag times for key metabolic regulators and identify genes showing altered temporal dynamics following dietary exposure. These molecular changes will be correlated with cellular and physiological phenotypes to establish functional relationships. We will develop novel analytical approaches to assess coordination between tissues, including temporal correlation analysis of molecular responses across organs and network analysis to identify inter-organ signaling pathways. Perturbation experiments will test tissue-autonomy versus inter-organ communication, and modeling approaches will help understand multi-organ metabolic circuit dynamics.

Through collaboration with experts in peripheral nervous system and adipose tissue innervation, we will perform targeted validation experiments in mice, focusing on adipose tissue innervation and comparing molecular plasticity between species. This will establish evolutionary conservation and translational relevance.

The project will connect molecular dynamics to physiological and behavioral consequences through metabolic phenotyping including energy storage, feeding behavior, and metabolic rate measurements. Neural circuit activity will be assessed and long-term metabolic health outcomes tracked. Reversibility experiments will identify potential intervention windows for therapeutic approaches.

OUTCOMES and IMPACT

This project will aim to shed light on how transcription and translation are temporally coordinated across multiple metabolic organs during development and how this coordination is altered by environmental metabolic stress. This could reveal fundamental principles of developmental gene expression control in physiological contexts, with specific insight into tissue-specific temporal signatures of molecular plasticity.
By mapping temporal response profiles, we will identify specific developmental periods when metabolic organs are most plastic and susceptible to dietary programming. This has direct implications for understanding sensitive periods in human development and may reveal critical windows for intervention. The project will reveal how brain, gut, and adipose tissue coordinate their molecular responses and whether developmental dietary stress disrupts this coordination, potentially contributing to metabolic syndrome.

Navigation in space is a fundamental cognitive capacity that relies on a brain wide circuit of Head Direction (HD) neurons. The vestibular based HD signal functions like a compass, providing spatial orientation information to hippocampal place cells and parahippocampal grid cells. However, the internally generated HD signal needs to be combined with visual landmark signals, to be useful for navigation. This project aims to dissect the pathways for visual updating of the HD signal.
Combining optogenetics and patch clamp recordings in mouse brain slices, the candidate will test the hypothesis that modulatory top-down cortico-thalamic projections, from the retrosplenial cortex and from the presubiculum, are integrated in anterior thalamic HD neurons. They will examine in vivo the specificities of presubicular output signals sent to the thalamus and to the mamillary bodies. Findings of this project will define the functional pathways that realign HD signals with visual landmarks.

Background and significance:
The head-direction (HD) system, the organism’s “neural compass”, has been studied extensively in both insects and rodents [1], and increasingly also in humans [2,3]. Single HD neurons increase their firing rate when the head is oriented in a specific direction, and they feed directional information to hippocampal place cells and entorhinal grid cells. A crucial hub in this system is the dorsal presubiculum, also termed postsubiculum, in which head-direction signals from the anterior thalamus, derived from vestibular signals, are integrated with visual landmark signals from the retrosplenial cortex [4-7]. The presubiculum can be considered as “head-direction cortex”, and the connection from the anterior dorsal thalamus to the presubiculum pathways exhibits key structural and functional similarities with canonical primary sensory pathways, including a driver thalamic input, specific laminar targeting, and receptive field transformations [8,9]. While intrinsic vestibular information such as angular velocity can be used to compute head direction for a limited time, precise determination of head direction requires “resetting”, that is, the realignment of the intrinsic signals with extrinsic visual cues [10]. As we and others reported, this visual information is relayed from the retrosplenial cortex (RSC) [6,11], via projections to the presubiculum. Yet, there are also modulatory top-down projections to the anterior thalamus, which have been insufficiently examined at the cellular and synaptic level, in the dorsal (AD) and in the ventral portion (AV) of the anterior thalamus. We hypothesize that anterior thalamus is not just a relay, but a key player for visuo-vestibular integration, contributing to landmark integration to the HD signal.

In this project we will study the anatomical and functional pathways for landmark updating of the HD signal. We ask, what is the role of the cortico-thalamic top-down projections to the anterior thalamus ? One prime candidate for carrying the visual landmark signal is the direct projection from the RSC to the anterior thalamus. Also the deep layers of the presubicular cortex send their axons to the anterior thalamus, while presubicular bursting neurons of layer 4 send their axons to the lateral mammillary nucleus [7,12]. Current research has not systematically decoupled differential contributions of projection-specific feedback connections to either the thalamus or to the mammillary bodies. These different output channels may play complementary roles depending on the visual context for resetting.

This project evolves around the idea that the visual anchoring mechanism in the HD system may flexibly involve the top-down cortico-thalamic and cortico-mamillary feedback. The disambiguation of their respective roles will be of particular functional interest when motion-related signals inform HD coding either alone, in the dark, or by integrating visual inputs from the environment.

Specific aims:
Aim 1: Functional mapping of cortico-thalamic pathways from the RSC and the presubiculum to the anterior thalamic nuclei.
Aim 2: Defining computations performed by thalamic neurons that receive HD signals from the mammillary bodies and landmark signals from RSC.
Aim 3: Testing visual updating of HD signals in vivo while manipulating cortico-thalamic or cortico-mammillary pathways using projection-specific optogenetic inhibition.

The present project proposal follows from Dr Fricker's previous work on the functional neuroanatomy and physiology of the parahippocampal region. To understand the strengths and efficiency of retrosplenial and presubicular feedback pathways to the anterior thalamus, the candidate will combine optogenetics and patch clamp recordings in brain slices. In awake mice, they will use large-scale neural recordings in the HD system, in controlled visuo-vestibular environments. Projection-specific silencing of the cortico-thalamic or cortico-mamillary pathways will test their relative contributions to the visual anchoring and resetting of HD signals. This research will provide a mechanistic understanding of visuo-vestibular integration in the HD system. It promises important insights, with broader impact for neuroscience and cognition in general.

Methodology.
Our lab routinely combines slice optogenetics with single or double whole cell patch clamp recordings [4-6]. The candidate will record from thalamic neurons in AD and AV, using a patch clamp setup equipped with a Multiclamp 700B amplifier and pCLAMP software. We have recently implemented dual wavelength optogenetic stimulation (Chronos and Chrimson opsins [6]) to examine synaptic integration from two input streams. Prior to the experiments, AAV5.Syn.Chronos-GFP.WPRE.bGH and AAV5-Syn-ChrimsonR-tdT will be injected into the granular portion of RSC and into the mammillary bodies, respectively, to be able to visualize and activate these two sets of axons innervating the thalamus. Converging inputs can be activated independently with dual wavelength LED stimulation. Ih channel antagonists will pinpoint the role of this ion channel for frequency dependent EPSP summation. Drugs to block or activate the cholinergic system will be applied to investigate potential gating effects of cholinergic neuromodulation.

Large-scale neural recordings will be carried out in head fixed mice that are exposed to vestibular stimulation in the presence of visual landmarks. The candidate will use Neuropixels probes to record HD activity from populations of thalamic neurons. My team has already validated the passive rotation setup, where mice are installed at the center of a motorized platform, such that rotating movements over 360° activate the vestibular system. A visual projection onto a surrounding dome creates controlled coherent or mismatched visuo-vestibular stimulation. Data are acquired using SpikeGLX and sorted with kilosort. Typically 40-60 well identified units can be recorded per session with directional tuning evident for >50% of units. Stereotaxic viral injections will be used to optogenetically inhibit distinct synaptic projections of retrosplenial cortical neurons, or presubicular neurons, targeting either thalamus or mammillary bodies. We will follow an intersectional approach, using retrgrade pAAV-Ef1a-mCherry-IRES-Cre and pAAV-CAG-DIO-PPO-Venus, such that blue light stimulation will rapidly and reversibly inhibit transmitter release at specific sets of axon terminals [14]. The candidate will analyze single neuron tuning curves, carry out population analysis of HD cells, and determine decoding lags of visual and vestibular variables across the different established conditions [15,16]. Manipulating specific pathways that contribute to landmark updating of HD signals will address their causal roles in different contextual settings and different moments in time.

References:
1. Hulse, B.K. and V. Jayaraman. Annu Rev Neurosci, 2020. 43: p. 31-54.
2. Kim, M. and E.A. Maguire, 2019. 29(7): p. 619-629.
3. Griffiths, B.J., et al. Nat Hum Behav, 2024. 8(7): p. 1334-1350.
4. Nassar, M., et al. J Neurosci, 2018. 38(28): p. 6411-6425.
5. Simonnet, J., et al. Nat Commun, 2017. 8: p. 16032.
6. Richevaux, L., et al. eLife, 2025, eLife12:RP92443
7. Yoder, R.M., et al. J Neurosci, 2015. 35(4): p. 1354-67.
8. Peyrache, A., et al. Prog Neurobiol, 2019. 183: p. 101693.
9. Duszkiewicz, A.J., et al. Trends Neurosci, 2025. 48(11):829-840.
10. Angelaki, D.E. and J. Laurens. Curr Opin Neurobiol, 2020. 60: p. 136-144.
11. Kononenko, N.L. and M.P. Witter. Hippocampus, 2012. 22(4): p. 881-95.
12. Huang et al. eNeuro, 2017. 4(2):ENEURO.0370-16.2017.
13. Page, Hector J. I., and Kate J. Jeffery. Front Cell Neuroscience, 2018. Jul 13:12:191.
14. Copits, Bryan A. et al. Neuron, 2021. 109(11): 1791 - 1809.e11
15. Chaudhuri R et al. Nat Neurosci. 2019. 22(9):1512-1520.
16. Siegenthaler, D., et al. Science, 2025. 389(6765): p. eadu9828.

The project aims at investigating the neural mechanisms that enable us to use all our senses to perceive others. While the discoveries of specific brain areas for face and voice processing have been important breakthroughs in science, little is known on how odors are processed by the primate brain, eventhough olfaction plays essential roles in social cognition, even already at birth. In this project the student will bridge novel approaches to examine neural mechanisms during social perception in different sensory modalities including the understudied sense of olfaction, by using a combination of neuroimaging in monkeys, electrophysiological recording of neurons, and large datasets analyses. He/she will investigate the neural mechanisms that enable transformation of social percepts into social knowledge and how olfaction is integrated with vision or hearing. This approach will be a first step to begin understanding the neural mechanisms underpinning the creation of social concepts.

Primates extract a wealth of information about their peers, including their gender, age, attractiveness, reproductive status, health status, dominance status or identity. These social concepts about individuals are sampled through several senses. Vision and audition are important sources of social information, through face, body or voice, which are processed by dedicated cortical areas, also called face areas, body areas and voice areas. But touch and olfaction are also major sources of information about others, though often in more unconscious or subliminal way. Olfactory signals found in sweat (but also in urine, saliva, and in some primate species, other than rhesus monkeys or humans, in specialized scent glands) advertise gender, social, reproductive and health status, and even kinship. Neural processing of odors occurs in piriform and insular cortex. But it is not known if, social odors are processed by separate neural areas compared to non-social odors, similarly to how faces and voices are processed. Further odor processing extends to the temporal and frontal lobes, where visual and auditory processing streams are located, suggesting that their processing streams may interact with the visual and auditory pathways to process social information. How can information from different sensory modalities give rise to same social concepts? One hypothesis would be that social concepts are encoded in amodal associative parts of the brain. The temporal and frontal cortex could be areas where social information from different modalities interacts and is integrated. An alternative hypothesis would be that each sensory system might have its own sensory-dependent representation of social concepts. What are the dynamics between secondary sensory cortices for social perception and amodal associative cortices? We will study at a fine temporal scale if social concepts about individuals are propagated back to secondary sensory cortices or first computed within these areas and next integrated in amodal associative cortices.

Aim1. Analyzing the stimuli collection.

We collected a large set of odor stimuli from 4 different species of primates (rhesus monkeys, tonkean monkeys, squirrel monkeys, humans) from sweat samples. Samples are indexed with Gender, Age and Dominance status, assessed through food priority tests in monkeys and 3 scales of dominance status in humans. Two samples per individual have been analyzed for chemical compounds, focusing on Volatile Organic Compounds (VOCs), which contribute to body odor characteristics and carry social information; while the others are stored for fMRI experiments.

Aim2. Providing a complete description of how individuals are perceived through different sensory channels, including undiscovered areas for social odor areas in the monkey brain.

fMRI of basic olfaction with contrast agent has been tested successfully in anesthetized macaques and we are adapting this procedure to awake fMRI and to our protocol. We will assemble samples of odors from monkeys with different social characteristics (Age, Gender, Dominance status) and present them as puffs, using a computer-controlled MRI-safe olfactometer. All odors will be delivered through tubes placed directly in animals’ nose. Sniffing will be recorded with a nasal cannula linked to a temperature sensor. Sniffing can potentially provide a ubiquitous signal for neuronal networks entrainment across the whole-brain, it will hence be regressed out of further analyses. To identify social odor areas in monkeys, we will parallel at best the design used to identify face areas, by contrasting neural activation to social odors to that to non-social odors of two types (1. natural odors of fruits, leaves, wood and 2. odor of synthetic materials such as plastics). To localize brain regions engaged by social percepts, we will scan four rhesus monkeys with fMRI in a classic human 3T MRI scanner. For comparison, we will localize face and voice areas using classical localizers, with sequences for T2*-weighted functional images adapted for perusing a brain of smaller size collected using a gradient-echo EPI sequence. Prior to scanning, monkeys will be trained in a mock scanner setup for habituation to scanner environment and for learning a fixation task. Whole-brain analyses will be carried using classic monkey or human fMRI pipeline, with tools from FS-FAST Freesurfer, AFNI, JIP, and FSL.
The result will be a description of how individuals are perceived through different sensory modality channels. Most recent studies show that the olfactory abilities of some primates are on a par with those of olfactory-dependent mammals such as rodents, in part due to the growing understanding of olfaction in primate social behavior. However most comparative analyses focused on anatomically defined areas of the primate brain. Carrying a functional comparative study in humans and rhesus monkeys, as opposed to an anatomical study, will allow to decipher if social odors and touch areas lay in similar brain regions, their relative size compared to total brain volume, and their number.

Aim3. Charting undiscovered territories of the brain for social concepts about individuals

Once areas for social perception are mapped, we will study how social concepts of Gender, Age and Dominance status are encoded within and outside regions found in Aim 2 using a cross-modal paradigm. To discover brain areas processing “social concepts about individuals”, we will identify those areas where measured contrast-enhanced signal is significantly higher for congruent association then for either stimuli presented alone. Incongruent associations might signal prediction error and we will track those areas where measured contrast-enhanced signal is significantly higher for incongruent association than for either stimuli presented alone. We will assess magnitude of signal change resulting from presentation of stimuli from two modalities: while congruent face-voice association can elicit signals of double the magnitude than each modality alone, or even superadditive effects, odors could rather act as modulators of brain activation, because of their more subliminal/unconscious processing compared to visual and auditory modalities.