Homeostasis within the neuro-glial unit is essential to the longevity of neurons. Conversely, loss of homeostasis, particularly of Ca2+ levels, of redox balance and of ATP, contribute to neuronal loss and dysfunction in many neurodegenerative and neurological disorders. This thesis is centred on better understanding the vulnerability of neurons to stress, as well as adaptive responses to these stresses. Since neurodegenerative conditions associated with Ca2+, redox and bioenergetic dyshomeostasis are often characterised by early dendritic pathology, I first studied dendritic vs. somatic responses of primary cortical neurons to these types of challenges in real-time. Using a wide range of genetically-encoded probes to measure Ca2+, ATP, NADH, glutathione and glutamate, I show that dendrites are selectively vulnerable to oxidative stress, excitotoxicity as well as to metabolic demand induced by action potential (AP) burst activity. However, I provide evidence that neurons undergoing energetically demanding AP burst activity can adjust their metabolic output by increasing mitochondrial NADH production in a manner dependent on the mitochondrial calcium uniporter (MCU), as well as increase their capacity to buffer their intracellular redox balance. Finally, I have studied transcriptional programs in astrocytes triggered by neurons and neuronal activity to better understand adaptive signaling between different cell types in the neuro-glial unit. I developed a novel system combining neurons and astrocytes from closely-related species, followed by RNA-seq and in silico read sorting. I uncovered a program of neuron-induced astrocytic gene expression which drives and maintains astrocytic maturity and neurotransmitter uptake function. In addition I identified a novel form of synapse-to-nucleus signaling, mediated by glutamatergic activity and acutely regulating diverse astrocytic genes involved in astrocyte-neuron metabolic coupling. Of note, neuronal activity co-ordinately induced astrocytic genes involved in astrocyte-to-neuron thyroid hormone signaling, extracellular antioxidant defences, and the astrocyte-neuron lactate shuttle, suggesting that this non cell-autonomous signaling may form part of the homeostatic machinery within the neuro-glial unit.

Les récepteurs nucléaires régulent l’expression des gènes importants pour le développement et l’homéostasie des cellules eucaryotes. Si certains d’entre eux sont activés par la liaison d’un ligand naturel, d’autres, comme le récepteur ERR, sont orphelins et ne possèdent pas de ligand naturel connu à ce jour. ERR se lie via son domaine DBD sur un élément de réponse ERRE au niveau des régions promotrices des gènes et son activité peut être modulée par la liaison de ligands synthétiques antagonistes. Par une approche de biologie structurale intégrative combinant biochimie, biophysique et cristallographie aux rayons-X, mon travail de thèse consistait en l’étude de l’organisation structurale d’ERR pour comprendre les mécanismes d’antagonisme induits par le ligand agoniste inverse XCT-790, et la topologie qu’il adopte en interagissant avec l’ADN. Mon projet était axé sur l’étude des domaines modulaires DBD et LBD en complexe avec ADN et ligand transposée à l’étude fonctionnelle du récepteur entier. Par un important travail biochimique, j’ai mis en place les protocoles de purification de ces protéines et caractérisé biophysiquement la stabilité et la stœchiométrie des complexes avec ADN et ligand. J’ai mis en évidence que le ligand XCT-790 se lie sur chacun des monomères de LBD avec une forte affinité et que le DBD interagit avec un ERRE sous la forme d’un monomère. De multiples essais de cristallisation ont abouti à l’obtention de cristaux pour les complexes avec le LBD et le DBD ayant diffracté respectivement jusqu’à 3,5 Å et 7 Å de résolution sans néanmoins pouvoir en déterminer les structures. L’ensemble de ces résultats permettent de conclure que le DBD est monomérique sur les ERRE et IR3 et semble être dimérique sur l’élément de réponse combiné ERRE/ERE. La topologie qu’adopte ERR sur l’ADN est dictée par la liaison de ligand et serait la conséquence d’une communication allostérique entre les domaines modulaires. Ces connaissances acquises faciliteront la détermination structurale d’ERR entier par cristallographie aux rayons-X ou par cryo-microscopie électronique. / Nuclear receptors regulate expression of important genes involved in development and homeostasis in eukaryotic cells. While some of them are regulated by the binding of a natural ligand, others, like ERR, are orphan and don’t have any natural ligand identified up to now. ERR binds to ERRE response elements through its DBD on promoter regions of genes and its activity can be modulated by the binding of synthetic antagonist ligands. By an integrative structural biology approach, my thesis work consisted in the study of the structural organization of ERR to understand the antagonism mechanisms induced by the inverse agonist XCT-790, and the adopted topology upon binding to DNA. My project focused on the study of the modular domains DBD and LBD complexes with DNA and ligand and extended to the functional study of the whole receptor. Thanks to significant biochemical work, I set up the purification protocols for these proteins and characterized by biophysics the stability and stoichiometry of the DNA and ligand complexes. This work revealed that XCT-790 binds on each LBD monomer with high affinity and that the DBD interacts to an ERRE as a monomer. Many crystallization trials led to crystals for the LBD and DBD complexes that diffracted respectively up to 3.5 Å and 7 Å of resolution but no structure could be determined. These results allow to conclude that the DBD is monomeric on the ERRE and IR3 and appears to be dimeric on the combined response element ERRE/ERE. The topology adopted by ERR on DNA is guided by the ligand binding and might be the consequence for an allosteric communication between the modular domains. These results provide the basis for a future structure determination of the full ERR by X-ray crystallography and cryo electron microscopy.

This work has focused on improving our knowledge of the thyroid specification process. Thyroid cells are derived from mouse embryonic stem cells (mESCs) by directed differentiation through multiple intermediate developmental stages, including anterior foregut endoderm (AFE), prior to NKX2-1+ thyroid progenitor specification. To investigate if transient Nkx2-1 expression can increase the efficiency of thyroid specification, we utilized a mESC line double knock-in GFP-T/hCD4-Foxa2 with a doxycycline inducible (Tet-On) Nkx2-1 transgene. Transient activation of the Nkx2-1 transgene at the AFE stage leads to stable induction of high levels of endogenous Nkx2-1 as well as early and mature thyroid-specific markers including Pax8, Foxe1, Tg, Nis, and Tshr. Lung and neuronal NKX2-1+ lineages were not derived in this system. The thyroid progeny mature and organize into follicle-like structures in 3D culture. These follicles express adherens and tight junction proteins indicative of an epithelial character and produce the thyroid hormone thyroxine (T4) in the presence of iodide. Critical determinants of thyroid lineage specification have been revealed by variations in developmental stage timing, signaling pathways, and sorting of AFE-stage subpopulations. To provide further insights into the mechanisms of thyroid specification, RNA-Seq data acquired from relevant stages has identified potential genes involved in early thyroid development. The results demonstrate that Nkx2-1 can act as a stage-specific inductive signal during directed differentiation of mESCs to thyroid follicular cells. We have also developed a mouse model to recapitulate these results in an in vivo context. This work has provided novel insights into thyroid specification and provides an efficient system for deriving and studying thyroid cells, which can be used for in vitro modeling of development and disease.

Chez les eucaryotes, le génome est organisé en chromatine, une structure nucléoprotéique essentielle pour la régulation de l’expression génique ainsi que pour le maintien de la stabilité du génome. Les ciliés sont d’excellents organismes modèles pour étudier les mécanismes généraux qui maintiennent l’intégrité du génomes eucaryote. Chez Paramecium tetraurelia, la différentiation du génome somatique à partir du génome germinal est caractérisée par des événements massifs et reproductibles d’élimination d’ADN. D’une part, des éléments répétés (transposons,régions minisatellites), de plusieurs kilobases de long, sont imprécisément éliminés.D’autre part, 45000 séquences courtes et uniques, appelées IES, sont précisément éliminées au nucléotide près. Une classe de petits ARN, appelé scnRNAs, est impliquée dans la régulation epigénétique de l’élimination d’ADN, mais comment les scnRNA contrôlent l’élimination d’ADN reste mystérieux. Nous avons testé l’hypothèse selon laquelle une organisation particulière de la chromatine, en particulier des modifications post-traductionelles des histones associées à des formes répressives de la chromatine, est impliquée dans le processus d’élimination d’ADN. Nous avons montré que la triméthylation de l’histone H3 sur la lysine 9 et la lysine 27 (H3K9me3 et H3K27me3)apparaît transitoirement dans le noyau somatique en développement au moment où se produisent les événements d’élimination d’ADN. Nous avons identifié la protéine de type Polycomb, Ezl1, et montré qu’elle est une histone methyltransferase qui présente une dualité de substrat et catalyse à la fois la mise en place de K9me3 et K27me3 sur l’histone H3. Nous avons montré que la déposition de H3K9me3 et H3K27me3 dans le noyau en développement requiert les scnRNAs. Des analyses de séquençage haut débit ont montré que Ezl1 est requise pour l’élimination des longues séquences répétées germinales, suggérant que les scnRNA guident la déposition des marques d’histones au niveau de ces séquences. Au contraire des régions répétées du génome, les IES montrent une sensibilité différente aux scnRNAs et à Ezl1, suggérant que plusieurs voies partiellement chevauchantes sont impliquées dans leur élimination. Notre étude montre que des caractéristiques intrinsèques des séquences d’ADN, telles que leur taille, peut contribuer à la définition des séquences germinales à éliminer. De manière intéressante, nous avons aussi montré que Ezl1 est requise pour la répression transcriptionnelle des éléments transposables. Nous suggérons que les voies H3K9me3et H3K27me3 coopèrent et contribuent à préserver le génome somatique de Paramecium des parasites génomiques. / Eukaryotic genomes are organized into chromatin, a complex nucleoprotein structureessential for the regulation of gene expression and for maintaining genome stability.Ciliates provide excellent model organisms with which to gain better understandinginto the regulation of genome stability in eukaryotes. In the ciliate Parameciumtetraurelia, differentiation of the somatic genome from the germline genome ischaracterized by massive and reproducible programmed DNA elimination events. Longregions of several kilobases in length, containing repeated sequences and transposableelements are imprecisely eliminated, whereas 45,000 short, dispersed, single-copyInternal Eliminated Sequences (IESs) are precisely excised at the nucleotide level. Aspecific class of small RNAs, called scnRNAs, is involved in the epigenetic regulation ofDNA deletion. How scnRNAs may guide DNA elimination in Paramecium remains tobe discovered. Here, we investigated whether chromatin structure, in particular histonepost-translational modifications known to be associated with repressive chromatin,might control DNA elimination. We showed that trimethylated lysine 9 and 27 onhistone H3 (H3K9me3 and H3K27me3) appear in the developing somaticmacronucleus when DNA elimination occurs. We identified the Polycomb-groupprotein, Ezl1, and showed that it is a dual histone methyltransferase that catalyzes bothH3K9me3 and H3K27me3 in vitro and in vivo. Genome-wide analyses show thatscnRNA-mediated H3K9me3 and H3K27me3 deposition is necessary for theelimination of long, repeated germline DNA. Conversely, single copy IESs displaydifferential sensitivity to depletion of scnRNAs and Ezl1, unveiling the existence ofpartially overlapping pathways in programmed DNA elimination. Our study revealsthat cis-acting determinants, such as DNA length, also contribute to the definition ofgermline sequences to delete. We further showed that Ezl1 is required fortranscriptional repression of transposable elements. We suggest that H3K9me3 andH3K27me3 pathways cooperate and contribute to safeguard the Paramecium somaticgenome against intragenomic parasites.

Transcriptional enhancers represent the primary basis for differential gene expression. These elements regulate cell type specificity, development, and evolution, with many human diseases resulting from altered enhancer activity. To date, a key gap in our knowledge is how enhancers select specific promoters for activation. To fill this gap, in this thesis, I first developed an Integrated Method for Predicting Enhancer Targets (IM-PET). Leveraging abundant “omics” data, I devised and characterized multiple genomic features for distinguishing true enhancer-promoter (EP) pairs from non-interacting pairs. I integrated these features into a probabilistic predictor for EP interactions. Multiple validation experiments demonstrated a significant improvement over extent state-of-the-art approaches. Systematic analyses of EP interactions across twelve human cell types reveals global features of EP interactions. Second, we used a well-established viral infection model to map the dynamic changes of enhancers and super-enhancers during the CD8+ T cell responses. Our analysis illustrated the complexity and dynamics of the underlying EP interactome during cell differentiation. Taking advantage of the predicted EP interactions, we constructed stage-specific transcriptional regulatory networks, which is critical for understanding the regulatory mechanism during CD8+ T cell differentiation. Third, recent progress in mapping technologies for chromatin interactions has led to a rapid increase in this type of interaction data. However, there is a lack of a comprehensive depository for chromatin interactions identified by all major technologies. To address this problem, we have developed the 4DGenome database through comprehensive literature curation of experimentally derived interactions. We envision a wide range of investigations will benefit from this carefully curated database.

Hematopoietic stem cell (HSC) development is accompanied by dynamic changes in the transcriptional program. How the corresponding transcriptional programs are related to the epigenetic mechanism is poorly understood. To fill this gap, we first profiled the transcriptomes and epigenomes using RNA-Seq and ChIP-Seq for five key developmental stages of HSC emergence in the mouse embryo. Using epigenetic markers, we identified novel 12,000~17,000 enhancers for each developmental stage. We applied a computational tool to link those enhancers to their target genes. Systematical analysis of enhancer-promoter (EP) pairs using network-based strategy reveals multiple novel key transcription factors for early specification of HSC in the mouse embryo. Second, we compared the 3D genome organization, epigenomes, and transcriptome of fetal and adult HSCs in the mouse. We found that higher-order genome structures are largely conserved between fetal and adult HSCs, including chromosomal compartments and topologically associating domains (TADs). However, chromatin interactions within TADs exhibit substantial differences. We found that promoters within 23% (242/1039) of TADs undergo interaction changes. Transcription factor motif analysis of HSC-specific enhancer-promoter loops suggests a role of KLF1 in mediating condition-specific enhancer looping and regulation of genes involved in cell cycle. Our result provides a comprehensive view of the differences in 3D genome organization, epigenome, and transcriptome between fetal and adult HSCs.

Les macrophages sont des cellules immunitaires innées présentes dans chaque organe. Ils sont des cibles thérapeutiques dans de nombreuses maladies, dont le cancer. En dépit de travaux récents sur l'origine des macrophages, les mécanismes régulant leur différenciation sont mal définis. L'expression de Jun, membre de la famille AP-1, augmente pendant la différenciation des macrophages, mais son rôle dans ce processus n'est pas connu.Au cours de mon doctorat, nous avons caractérisé le rôle de Jun dans le développement et l'homéostasie des macrophages, dans un modèle de souris avec délétion conditionnelle de Jun dans la lignée myéloïde (JunΔCsf1r). Nous montrons que Jun contrôle la différenciation, induite par CSF1, des monocytes en macrophages. In vivo, Jun régule l'accumulation de macrophages dérivés de monocytes dans les poumons et intestins. Les macrophages associés aux tumeurs (TAMs) jouent un rôle crucial dans la progression des cancers. L’absence de Jun freine la croissance d’un mélanome et la différenciation, induite par CSF1, des TAMs dérivés de monocytes qui participent à l’angiogénèse tumorale. Cependant, lors d'une inflammation aiguë, Jun n’affecte pas le recrutement de macrophages inflammatoires.En conclusion, nos résultats identifient Jun comme un régulateur central de la différenciation des macrophages. Dans un modèle de mélanome, les macrophages Jun-dépendants exercent des fonctions pro-tumorales. Le fait que Jun soit un régulateur sélectif du développement des macrophages dépendants de CSF-1 permettra de définir de nouvelles approches ciblant sélectivement la différenciation des macrophages, sans altérer les réponses immunitaires dépendantes des monocytes. / Macrophages are immune cells present in every organ. Given their variety of functions, macrophages are therapeutic targets in many diseases including cancer. Despite the research efforts to characterise their origins, the molecular mechanisms regulating macrophage differentiation are still poorly defined. Expression of the AP-1 factor, Jun, increases during differentiation, but its role in macrophage development is not known.During my PhD, we characterised how Jun affects macrophage development and homeostasis. We developed a conditional mouse model in which Jun is deficient in the myeloid lineage (JunΔCsf1r). We showed that Jun controls CSF1-mediated monocyte to macrophage differentiation, proliferation and survival. In vivo, Jun loss limits macrophage accumulation in lungs and intestine. Tumour-associated macrophages (TAMs) play critical roles in cancer progression. We observed that Jun deficiency dampens melanoma growth and the differentiation of CSF1-dependent monocyte-derived TAMs. We further showed that Jun-dependent TAMs mediate vessel normalisation in melanoma. During inflammation, Jun was dispensable for the recruitment of monocyte-derived inflammatory macrophages.Altogether, our results identify Jun as a master regulator of macrophage differentiation, without altering monocyte effector functions. In a melanoma model, we showed that Jun-dependent TAMs play tumour-promoting roles. Therefore, Jun is a selective regulator of CSF-1-dependent macrophage development, which is redundant during inflammation; this observation should help to define novel approaches to selectively target macrophage differentiation, without altering monocyte-dependent immune responses.

ABSTRACT A multitude of complex cellular changes are required throughout development in order for a single cell to transform into a fully functioning organism. Cellular events including proliferation, migration, and differentiation have to be carefully controlled in order for development to proceed correctly. In order to study such dynamic processes, in vivo models are often utilized. Using the zebrafish (Danio rerio) as a model system, we have investigated the role of an axon guidance signaling pair, Semaphorin6A (Sema6A) and PlexinA2 (PlxnA2), in neurodevelopment. A previous investigation into the developmental expression patterns of sema6A and plxnA2 in zebrafish, revealed overlapping expression in the developing eye. At this early stage, the cells in the optic vesicles are undifferentiated retinal precursor cells (RPCs) and therefore do not require Sema/Plxn signaling for their canonical axon guidance role. To understand what the function of this early expression was, we knocked down both sema6a and plxna2 and observed 1) a loss of cohesion of RPCs within optic vesicles, and 2) a decrease in RPC proliferation (Ebert et al., 2014). Because these phenotypes were seen at an early stage and given that many developmental processes are dependent on genetic regulation, we hypothesized that Sema6A/PlxnA2 signaling could be regulating transcription of downstream target genes. To investigate this, we performed a microarray experiment and uncovered 58 differentially regulated genes (Emerson et al., 2017a). Prior to our study, it was not known that Sema/Plxn signaling led to changes in gene transcription. In an effort to understand the contribution of identified candidate genes to early sema6A/plxnA2 knockdown phenotypes, candidate genes with predicted functions in proliferation and migration were investigated. First, we show that rasl11b is important for regulation of RPC proliferation in the developing optic vesicles. Second, we show that shootin-1 is important in optic vesicle migration, retinal pigmented epithelium formation and optic tract patterning. Furthermore, PlxnA2 regulation of shootin-1 levels is important in sensory and motor axon patterning and branching in the peripheral nervous system. Belonging to a large family of proteins with the ability to cross talk, Semas and Plxns rely on spatially and temporally differential expression patterns to perform their tissue-specific roles. Here, we used in situ hybridization to comprehensively uncover the neuronal expression patterns of the PlxnA family in the early developing zebrafish (Emerson et al., 2017b). In addition, we present for the first time that zebrafish have two genes for PlxnA1, A1a and A1b, which show divergent expression patterns. Semas and Plxns are critical for many aspects of development and together, this body of work provides further insight into the downstream signaling mechanisms and roles of these essential developmental signaling proteins.

Transcription of the Pseudomonas aeruginosa type III secretion system is controlled by ExsA, a member of the AraC/XylS family of regulators. ExsA is comprised of an amino terminal domain that is involved in self-association and regulatory functions, and a carboxy-terminal domain that contains two helix-turn helix (HTH) DNA-binding motifs which contact promoter DNA. Previous work from our lab determined the function of the two independent ExsA domains and found that each ExsA-dependent promoter contains two adjacent binding sites for monomeric ExsA. The promoter-proximal site (binding site 1) consists of highly conserved GnC and TGnnA sequences that are individually recognized by the two HTH DNA-binding motifs of an ExsA monomer. Nevertheless, the details of how ExsA recognizes and binds to ExsA-dependent promoters were still unknown. In chapter II I show that the two ExsA monomers bind to promoter regions in a head-to-tail orientation and identify residues in the first HTH of ExsA that contact the GnC sequence. Likewise, residues located in the second HTH motif, which contribute to the recognition of the TGnnA sequence, were also identified. While the GnC and TGnnA sequences are important for binding to site 1, the promoter-distal binding sites (site 2) lack obvious similarity among themselves or with binding site 1. Site 2 in the PexsC promoter region contains a GnC sequence that is functionally equivalent to the GnC in site 1 and recognized by the first HTH motif of an ExsA monomer and the second HTH interacts with an adenine residue in binding site 2. A comparison of hybrid promoters composed of binding site 2 from one promoter fused to binding site 1 derived from another promoter indicates that ExsA-binding affinity, promoter strength, and the degree of promoter bending are properties that are largely determined by binding site 2. Through the course of the ExsA studies I observed that the amino acids that comprise the HTH motifs of ExsA are nearly identical to those in LcrF/VirF, the activators of T3SS gene expression in the pathogenic yersiniae. In chapter III I tested the hypothesis that ExsA/LcrF/VirF recognize a common nucleotide sequence. Here I report that Yersinia pestis LcrF binds to and activates transcription of ExsA-dependent promoters in P. aeruginosa, and that plasmid expressed ExsA complements a Y. pestis lcrF mutant for T3SS gene expression. Mutations that disrupt the ExsA consensus-binding sites in both P. aeruginosa and Y. pestis T3SS promoters prevent activation by ExsA and LcrF. All of the data combined demonstrate that ExsA and LcrF recognize a common nucleotide sequence. Nevertheless, the DNA binding properties of ExsA and LcrF are distinct. Whereas two ExsA monomers are sequentially recruited to the promoter region, LcrF binds to promoter DNA as a preformed dimer and has a higher capacity to bend DNA. An LcrF mutant defective for dimerization bound promoter DNA with properties similar to ExsA. Finally, I demonstrate that the activators of T3SS gene expression from Photorhabdus luminescens, Aeromonas hydrophila, and Vibrio parahaemolyticus are also sensitive to mutations that disrupt the ExsA-consensus binding site. Taken together, this work shows that ExsA binding and activation at T3SS gene promoters serves as a model system by which the DNA binding properties of other AraC family transcriptional activators can be predicted.

Le silencing transcriptionnel limite la transcription des gènes et des éléments transposables dont l’expression pourrait être délétère à la cellule. Il dépend d’une diversité de modifications de la chromatine comme la méthylation ADN ou les marques répressives des histones. De façon à mieux comprendre les mécanismes moléculaires à l’origine du silencing transcriptionnel, nous avons mené une approche de génétique directe à l’aide d’un transgène soumis au silencing dans la plante modèle Arabidopsis thaliana. Cette stratégie nous a permis d'isoler à la fois des mutants déficients pour le maintien du silencing transcriptionnel et des mutations qui empêchent la réactivation transcriptionnelle des éléments transposables en réponse à un stress thermique. Nous avons caractérisé les défauts provoqués par ces mutations en combinant des approches de biologie moléculaire, de cytologie et de génomique.Nous montrons ainsi que MED14, la sous-unité centrale du complexe Mediator, et UVH6, composant du complexe TFIIH, sont requis pour la transcription de l'hétérochromatine en stress thermique. MED14 stimule aussi la transcription de l'hétérochromatine en l'absence de stress, mais ne semble fonctionner qu'en présence de la méthylation ADN. En plus de cette fonction originale, nous identifions un nouveau rôle de MED14 dans le maintien de la méthylation ADN, possiblement via la voie de méthylation ADN dirigée par les petits ARN.Par ailleurs, nos résultats nous ont permis d’identifier le rôle des protéines MAIN et MAIL1, qui définissent une voie de silencing transcriptionnelle indépendante des voies connues jusqu'alors. De façon intéressante, MAIN et MAIL1 possèdent un domaine protéique partagé avec les éléments transposables, qui aurait successivement été capturé par les éléments transposables et leur hôte au cours de l’histoire évolutive des plantes à fleurs.Enfin, en isolant une nouvelle mutation du gène POL2A, nous confirmons le rôle de l’ADN polymérase epsilon dans le silencing transcriptionnel et caractérisons les propriétés chromatiniennes qui dépendent de POL2A. Nous montrons que les défauts de silencing des mutants pol2a corrèlent avec une désorganisation importante de l’hétérochromatine sans diminution drastique des marques qui y sont associées. Au contraire, nous détectons une hyperméthylation ADN prononcée dans le mutant, et explorons différentes hypothèses pour expliquer ce phénotype particulier. Nos données suggèrent que plusieurs mécanismes moléculaires sont à l’origine des défauts des mutants pol2a. Elles confirment le rôle prépondérant de la chromométhylase CMT3 dans la régulation de la méthylation ADN, et suggèrent qu’un stress réplicatif pourrait causer une hyperméthylation de l’ADN.Dans l’ensemble, ces travaux de thèse proposent des pistes de travail dont l’exploration pourrait permettre d’expliquer les effets des déficiences réplicatives dans le maintien du silencing transcriptionnel et de l’homéostasie de la méthylation ADN. Ils suggèrent en outre que MED14 a une fonction dédiée à la transcription de l’hétérochromatine qui pourrait stimuler le maintien de la méthylation ADN. / Transcriptional gene silencing hinders deleterious transcription of some genes and transposable elements. Silencing is maintained by numerous chromatin modifications such as DNA methylation and repressive histone marks. To better understand the molecular mechanisms of silencing, we conducted a forward genetic screen using a transgene reporter system targeted by transcriptional gene silencing in the model plant Arabidopsis thaliana. We isolated a first type of mutants with diminished maintenance of silencing and a second category that displayed deficient release of transgene silencing upon heat stress. We then combined molecular, cytological and genomic methods to characterize the defects associated with these mutations.First, we show that the Mediator subunit MED14 and the TFIIH complex subunit UVH6 are required for heat-stress-induced release of silencing. We further show that MED14, but not UVH6, promotes transcriptional activation of transposable elements in mutant contexts where silencing is defective. Importantly, MED14 is only required when DNA methylation is not affected, suggesting that MED14 has a specialized function to promote transcription of heterochromatin. Furthermore, we show that MED14 promote DNA methylation at targets regulated by RNA-directed DNA methylation.Characterizing mutants from the first category, we unveil the contribution of the MAIN and MAIL1 proteins into transcriptional gene silencing, and show that they likely act through a pathway independent of known silencing factors. Interestingly, MAIN and MAIL1 bear a protein domain that is shared with transposable elements, and that has been captured by transposable elements and genes throughout the evolutionary history of flower plants.Additionally, we confirm the involvement of the DNA polymerase epsilon in transcriptional gene silencing by isolating a new mutation of the POL2A gene among mutants of the first category. We characterize the effects of the pol2a mutation on several heterochromatin properties, and show that the pol2a mutant retains high levels of heterochromatin marks despite having highly disorganized heterochromatin. We actually detect a strong elevation of DNA methylation in the pol2a mutant and explore different hypothesis to explain this unusual phenotype. We show that increased expression of the CMT3 chromomethylase is a likely cause, but that additional molecular mechanisms are probably involved. Further exploration suggests that constitutive replicative stress occurring in pol2a mutants could be an additional cause of DNA hypermethylation.To summarize, this work provide putative causes for DNA hypermethylation and silencing defects in a situation of replicative deficiency. Further investigation will be required to identify the molecular components involved in the mechanism. Our data further suggest that MED14 has a function dedicated to heterochromatin transcription that could promote DNA methylation maintenance.