Les protéines du groupe Polycomb (PcG) ont initialement été identifiées chez la drosophile comme répresseurs transcriptionnels des gènes homéotiques. Aujourd’hui, nous savons que ces protéines jouent un rôle bien plus large puisqu’elles régulent des gènes dont les produits sont impliqués dans de nombreux processus biologiques (régulation des gènes HOX, maintien de la plasticité des cellules souches, la différenciation cellulaire, l’inactivation du chromosome X, la régulation des gènes soumis à empreintes). Leur dérégulation est source de nombreux cancers chez l’homme. Hautement conservées, elles forment deux principaux complexes : PRC 1 et 2 (Polycomb repressive complex 1 and 2), dont l’activité est respectivement reflétée par la mono-ubiquitinylation de la lysine 118 l’histone H2A (H2AK118Ub) et la tri-méthylation de la lysine 27 de l’histone H3 (H3K27me3). Chez la Drosophile, les sites de fixation de ces complexes sont appelés PRE (Polycomb Responsive Elements) où ils sont recrutés via des facteurs de transcription (FT).La complexité du recrutement des complexes du PcG, chez la Drosophile comme chez les mammifères, est visible à différents niveaux : au niveau de la séquence même de leurs sites de fixations, au niveau des facteurs de transcription qui les recrutent, au niveau de l’interface entre les deux complexes PRC1 et PRC2 et enfin au niveau global, part le présence de ces complexes au niveau de sites transcriptionnellement actifs. L’ensemble de ces résultats démontre clairement la nature hétérogène des PRE. Ces derniers diffèrent non seulement par leur séquence, mais également par les FT qui les recrutent et enfin par la manière dont les complexes PcG sont recrutés (PRC2 recrute PRC1 ou le contraire). Mon projet de thèse s’est donc dessiné autour d’une hypothèse : il existe différentes classes de PRE chez la Drosophile. Mon travail a donc consisté à définir ces différentes classes et à les caractériser pour en déduire des rôles spécifiques à l’échelle génomique. En effet, l’implication des complexes du PcG dans l’apparition de cancer chez l’Homme requière que l’on comprenne comment ces protéines sont recrutées à la chromatine.Mes travaux de thèse ont permis d’identifier six classes différentes de sites de fixation aux protéines du PcG. Nous avons retrouvé une classe correspondant aux sites de fixations canoniques fixés par les protéines du PcG et présents au sein de larges domaines répressifs marqués par H3K27me3. Une seconde classe correspond à des éléments de régulation marqués par un état de pause transcriptionnelle. De façon surprenante, nous avons démontré qu’une grande partie des sites de fixation des complexes du PcG était localisée au niveau de régions transcriptionnellement actives. Ces classes de PRE diffèrent en particulier en éléments génomiques qui les composent. Deux classes correspondent à des enhancers développementaux. Une classe correspond à des promoteurs actifs pouvant réguler des gènes de ménage. Enfin, une dernière classe correspond à des bordures de TAD. Les sites actifs et réprimés fixés par le PcG fixent également des combinaisons différentes de FT. Des analyses in vivo associées à un transcriptome réalisé à partir de cellules mutantes pour une protéine du PcG révèlent que les complexes du PcG jouent également un rôle de répresseur transcriptionnel au niveau des sites actifs. L’ensemble de ces résultats suggère une hétérogénéité inattendue des sites de fixation des complexes du PcG et permettra de mieux comprendre les caractéristiques liées à ces protéines dont la dérégulation mène à l’apparition de cancers chez l’Homme marqués par leur agressivité. / Polycomb group (PcG) complexes were initially discovered in Drosophila as transcriptionnal repressors of homeotic genes. To date, we know that they are involves in a large pleithora of biological processes including the maintenance of stem cells plasticity, differentiation, X chromosome inactivation and imprinting. PcG complexes are highly conserved from Drosophila to Humans and can be divided into two main complexes: PRC1 and PRC2 (Polycomb repressive complex 1 and 2). Both complexes have a histone modifying activity: PRC1 catalyses the mono-ubiquitination of the lysine 118 on histone H2A (H2AK118Ub) and PRC2 catalyses the tri-methylation of the lysine 27 on histone H3 (H3K27me3).In Drosophila, these complexes are recruited to cis regulatory elements named Polycomb Responsive Elements (PREs) that drive the epigenetic inheritance of silent chromatin states throughout development. Importantly, PcG complexes do not contain DNA-binding activity but are recruited to PREs via their interaction with Transcription Factors (TF) recognizing DNA motifs clustered at PREs. However the mechanism how PREs target PcG complexes is still not well understood due to the complexity of PcG recruitment, which is reflected at different levels: The DNA signature between PREs can differ significantly and several TF are implicated in PcG recruitment, but none of them is sufficient to recruit PcG complexes to PREs. Moreover PcG complexes can cooperate in different ways to stabilize each other’s binding. Finally, another layer of complexity is found at a more global level since PcG complexes do not only bind repressed sites, but they are also found at active regions.Therefore, our working hypothesis is that different classes of PREs exist in Drosophila. My PhD work was thus to define these different classes of PREs on a genome-wide scale and to functionally characterize them in order to get a complete molecular description of PRE function. Understanding how PcG complexes are recruited is of high importance, since deregulation of both, PcG complexes and their recruiting factors can led to cancer and diseases. My work led to the identification of six different classes of PREs that are characterized by different chromatin and genomic features. Interestingly the majority of PREs are associated with active genes that can be divided into housekeeping regulatory regions and developmental enhancers. In addition another class comprises bona fide chromatin domain boundaries. On the other hand PREs associated with repressed chromatin states shows features of previously described PREs and associate with repressed genes and PcG-associated histone marks. Finally another class comprises PREs that are likely in a poised chromatin state. We further demonstrated that PREs located at repressed and active regions differ in their combination of TF. In vivo analyses along with a transcriptomic analysis performed in cell lines mutated for a member of PcG complexes revealed that PcG complexes play a repressive role at both, active and repressed PREs.Taken together, our result suggest an unexpected heterogeneity of PREs and contributes to the better understanding of their characteristics and function.

No description available.

Les protéines des groupes Polycomb et trithorax sont des régulateurs épigénétiques très conservés qui permettent le maintient de l'identité cellulaire en régulant le niveau d'expression des gènes. Ils agissent sur leurs gènes cibles à travers des éléments régulateurs en cis, appelés éléments de réponse aux Polycombs (PRE). Dans des tests transgéniques, il a été montré que deux copies du même PRE sont fréquemment regroupés dans la même région nucléaire. Dans le cas particulier du PRE Fab-7, ce regroupement corrèle avec sa fonction répressive. Durant ma thèse, j'ai tenté de cloner un outil bicolore qui permet la visualisation en 4D de deux PRE Fab-7 stablement intégrés dans le génome de Drosophila melanogaster. De plus, j'ai amélioré le protocole de DNA-FISH du labo. Ceci m'a permis d'identifier vestigial et apterous comme étant des loci qui forment des associations nucléaires, de façon dépendante de la transcription, dans Drosophila melanogaster. / Polycomb- and trithorax-Group proteins are highly conserved epigenetic regulators which maintain cell identities by maintaining states of gene expression. They act on their target genes through /cis/ regulatory elements, named Polycomb Response Elements (PREs). In transgene assays it has been shown that two copies of the same PRE are frequently found clustered in nuclear space and for one particular PRE named Fab-7 clustering is correlated with its repressive function. In the course of this thesis I tried to clone a two colour real-time tool which allows distinguishing in 4D two /Fab-7/s stably integrated into the genome of Drosophila melanogaster. Additionally, I improved the DNA-FISH protocol of the lab and identified vestigial and apterous as potential gene loci forming nuclear associations dependent on transcription in Drosophila melanogaster.

L’étude du contrôle de l’expression des gènes a été, au cours de ces dernières années, révolutionnée par la découverte des régulations épigénétiques. Parmi les différents acteurs participant à ces régulations se trouvent les protéines du groupe Polycomb (PcG). Ces protéines, initialement découvertes chez la drosophile, sont responsables de la mise en place et du maintien de "marques épigénétiques" au niveau de gènes cibles, qui sont alors réprimés. Les protéines PcG agissent sous forment de trois complexes dinstincts chez les animaux nommés PRC1 (Polycomb Repressive Complex 1), PRC2 (Polycomb Repressive Complex 2) et PhoRC (Pleiohomeotic Repressive Complex); le PRC2 possédant une activité histone méthyltransférase de type H3 K9/27. Chez les plantes, seules trois classes de protéines PcG sont retrouvées: la classe des Enhancer of zeste E(z), des Extra Sex Combs (ESC) et des Supressor of zeste 12 (Su(z)12), formant le complexe PRC2. Leur rôle dans le développement des plantes a été mis en évidence chez Arabidopsis, au niveau du gamétophyte femelle et de la graine, du maintien de l’état végétatif, de l’identité florale et de la vernalisation. Cependant leur implication dans le développement du fruit reste inconnue. Mon travail a permis d'identifier et de caractériser deux gènes PcG de la classe des E(z) de tomate exprimés dans le fruit, nommés SlEZ1 et SlEZ2. Les proteines SlEZ1 et SlEZ2 présentent l’ensemble des domaines caractéristiques des protéines de cette classe et sont localisées dans les noyaux. Les expériences de double hybride révèlent que les protéines SlEZ1 et SlEZ2 sont capables de former des complexes de type PRC2 avec certaines autres protéines PcG de tomate (de type ESC et Su(z)12). Ceci suggère que SlEZ1 et SlEZ2 sont effectivement des protéines fonctionnelles. L’analyse de des profils d’expression des gènes SlEZ1 et SlEZ2 révèle une expression ubiquitaire dans la plante au niveau de l’appareil végétatif, de la fleur et dans le fruit. Cependant, dans la fleur, seul SlEZ1 présente une expression dans les étamines tandis que les ARNm de SlEZ2 sont présent de façon spécifique dans le tissu de transmission du style. Dans le fruit, SlEZ1 est exprimé de façon constante, tandis que SlEZ2 semble faiblement exprimé dans les fruits en cours de mûrissement. Afin d’identifier la fonction de SlEZ1 dans le développement du fruit, des plantes transgéniques sous-exprimant SlEZ1 de façon constitutive ont été générées. Elles présentent une morphologie altérée de la fleur: les étamines sont torsadées et ne forment pas de cône staminal fermé. De plus, une augmentation du nombre moyen de carpelles par fruit est observée. / The control of gene expression has been challenged by the discovery of epigenetic regulation. Among the different factors involved in epigenetic regulations, the Polycomb (PcG) proteins are known to repress gene expression by setting epigenetic marks. The PcG protein, initially discovered in drosophila, act together in three distinct complexes named PRC1 (Polycomb Repressive Complex 1), PRC2 (Polycomb Repressive Complex 2) and PhoRC (Pleiohomeotic Repressive Complex). PRC2 complexes methylate histone H3 on lysines K9/27. In plants, only three classes of PcG protein has been found: the Enhancer of zeste (E(z)) class, the Extra Sex Combs (ESC) class and the Supressor of zeste 12 (Su(z)12) class, which belong to the PRC2. Their function in plant development has been brought to light in Arabidopsis thaliana. They control female gametophyte and seed development, maintain the vegetative development, and are involved in floral identity and vernalization. However, their function in fruit development is still unknown. My work was aimed to identify and characterize two PcG genes, named SlEZ1 and SlEZ2, encoding tomato E(z) class proteins. SlEZ1 and SlEZ2 proteins contain all the five E(z) characteristic domains and are both localized in the nucleus. Furthermore, as double-hybrid experiments reveal that both SlEZ1 and SlEZ2 proteins are able to form PRC2 complexes and interact with PcG proteins of other classes (ESC and Su(z)12 classes), it seems that these proteins are functional. Their expression profiles reveal ubiquitous expression during vegetative development (leaves, buds, stems) and reproductive development (flowers and fruits). However SlEZ1 is specifically expressed in the stamens whereas SlEZ2 shows specific expression in the transmitting tissue of the style. Moreover, their expression during fruit development shows some differences: if SlEZ1 expression is almost constant, SlEZ2 expression decreases during fruit development. In order to indentify SlEZ1 functions in fruit development, transgenic plants underexpressing constitutively SlEZ1 have been generated. These plants present altered flower morphology with twisted stamens and increased carpel number fruits.

Les protéines du groupe Polybomb (PcG) sont conservées de la drosophile jusqu’à l’homme et assurent la « mémoire cellulaire » d’un état transcriptionnel réprimé au cours du développement. Un modèle a été proposé pour expliquer leur fonctionnement, qui propose que les deux principaux complexes, PRC1 et PRC2 (Polycomb Repressive Complexes 1 & 2), sont recrutés ensemble au niveau de séquences spécifiques appelés PREs (Polycomb Responsive Elements) où ils collaborent pour maintenir la chromatine dans un été réprimé.Au cours de ma thèse, j’ai voulu tester ce modèle en utilisant les disques imaginaux d’œil-antenne de drosophile, qui sont des structures larvaires préfigurant l’œil adulte. Étonnamment, alors que les mutants PRC1 et PRC2 présentent des phénotypes similaires dans l’embryon, seuls les clones mutants PRC1 présentent une transformation néoplasique et une surcroissance dans l’œil. Pour comprendre les mécanismes moléculaires qui sous-tendent ce découplage fonctionnel, nous avons réalisé des ChIP-Seq contre plusieurs marques d’histones actives et répressives, ainsi que contre des protéines du PcG. La comparaison de ces ChIP-Seq avec les profils embryonnaires a d’abord révélé un redéploiement majeur du PRC1 au stade larvaire, sur environ 1000 promoteurs actifs. Cette nouvelle classe de cibles, que nous avons appelée « Neo-PRC1 », se trouve au niveau de gènes actifs où la marque H3K27me3 normalement déposée par le PRC2 est remplacée par la marque active H3K27Ac. Ces gènes sont impliqués dans la régulation de la polarité, la prolifération ou encore la signalisation cellulaires, et un nombre substantiel d’entre eux est surexprimé dans les mutants PRC1, mais pas PRC2. Ces résultats suggèrent que l’activité suppresseur de tumeurs du PRC1 au stade larvaire découle de la régulation précise de gènes classiquement dérégulés dans les cancers, et ce en l’absence du PRC2En plus des sites situés sur des promoteurs actifs, nous avons détecté des sites PRC1 sans PRC2 au niveau de régions enrichies pour des marques de séquences amplificatrices (« Enhancers » en anglais) actives. Ces sites correspondent à des séquences amplificatrices spécifiquement actives au stade larvaire, et sont localisés à proximité de gènes codant pour des facteurs de transcription cruciaux pour le développement de l’œil, tels que les gènes du réseau de détermination de la rétine (RDGN). Pour mieux comprendre l’action du PRC1 sur ces cibles, j’ai réalisé des expériences de Hi-C (High-throughput Chromosome Conformation Capture) dans l’œil et l’embryon, révélant ainsi que ces séquences amplificatrices contactent les promoteurs proches spécifiquement au stade larvaire. De plus, la fréquence des contacts est positivement corrélée au niveau de PRC1 fixé. Étonnamment, ces gènes cibles sont sous-exprimés dans les mutants PRC1 mais pas dans les mutants PRC2, ce qui suggère que les contacts PRC1-dépendants entre ces séquences amplificatrices et leurs promoteurs cibles promeuvent la transcription. Pour vérifier cette hypothèse, j’ai étudié l’impact de la délétion via CRISPR de deux sites PRC1 impliqués dans une boucle régulatrice ; l’un situé au promoteur d’un gène du RDGN appelé dac et l’autre sur une séquence amplificatrice putative située en aval du gène. Des expériences de 3D-FISH révèlent que leur délétion entraîne la diminution des contacts entre la séquence amplificatrice et le promoteur, avec pour effet la sous-expression de dac. Ces résultats suggèrent que le PRC1 est impliqué dans la formation de boucles entre les séquences amplificatrices et leurs promoteurs cibles, et que cette topologie est nécessaire pour l’activation de ces gènes au cours du développement.Ma thèse a donc contribué à la découverte de nouvelles fonctions pour le PRC1, qui acquiert de nouvelles cibles au cours du développement et régule la transcription de gènes impliqués dans le cancer ou le développement indépendamment du PRC2, via des mécanismes dédiés. / Polycomb Group (PcG) are a set of highly conserved proteins implicated in cellular memory of transcriptional gene silencing throughout development. A classical model of PcG mode of action proposes that the two main Polycomb Repressive Complexes (PRC), PRC1 and PRC2, are co-recruited at specific DNA sequences called PREs (Polycomb Responsive Elements) where they collaborate to stably maintain a repressed chromatin state.My PhD work has challenged this collaborative model, by using as an experimental system the Drosophila larval Eye-Antennal imaginal Disc (EAD) that prefigures the adult eye. Surprisingly, while PRC1 and PRC2 mutants exhibit similar phenotypes in embryos, only PRC1 mutant clones show neoplastic transformation and massive overgrowth in EAD, while PRC2 mutant clones do not. To understand the molecular basis of this functional uncoupling, we generated ChIP-Seq directed against a large set of repressive and active Histone Marks (HTMs) as well as against core PcG proteins in EAD. A comparative analysis with Chip-Seq embryonic profiles firstly identified a massive de novo redeployment of PRC1 proteins at mostly 1000 active promoters that occurs only at larval stage. This new class of transcriptionally active PcG target genes, that we named “Neo-PRC1”, is devoid of the H3K27me3 epigenetic mark normally deposited by PRC2 and carry instead the active H3K27Ac mark. Moreover, this Neo-PRC1 category of PcG targets is enriched in ontologies linked to cell polarity, proliferation or signalling. A substantial subset of neo-PRC1 targets is up-regulated in PRC1 but not in PRC2 mutants, suggesting that the tumour-suppressor activity of PRC1 during Drosophila development might be exerted by fine-tuning the expression of cancer-related genes independently of PRC2.In addition to neo-PRC1 sites located at promoters, we next detected an enrichment of PRC1, but not PRC2, at regions enriched for active enhancer marks. These neo-sites which correspond to larval stage-specific enhancers are found in the vicinity of genes encoding for transcription factors playing a key role in EAD development, like genes implicated in the Retinal Determination Gene Network (RDGN). To understand the function of PRC1 at these enhancers, we performed comparative Hi-C (High-throughput Chromosome Conformation Capture) experiments between embryos and EADs, and discovered differential chromatin contacts occurring between the stage-specific neo-PRC1 enhancers and their closest promoters. The intensity of these 3D contacts is positively correlated with the PRC1-binding levels. Unexpectedly, in PRC1, but not in PRC2 mutants, these genes are down-regulated, suggesting that PRC1-dependent enhancer-promoter loops promote transcription. To study if larval 3D chromatin loops are PcG-dependent and functionally relevant, we analyzed the topological and transcriptional impact of two CRISPR-generated deletions affecting two PRC1 binding sites known to form a regulatory loop. These two PREs are respectively located close to the promoter and a putative 3’ enhancer of the dac locus encoding for a crucial member of the RDGN. 3D FISH experiments demonstrate that the removal of the dac endogenous PRC1 binding sites is sufficient to significantly decrease dac enhancer-promoter contacts as well as to trigger down-regulation of dac expression. Altogether, these results suggest that PRC1 might contribute to enhancer-promoter contacts at crucial developmental genes in EAD and that these PRC1-dependent long-range interactions could be necessary to allow a proper transcriptional induction during development.To summarize, my PhD project contributed in opening a new perspective, namely that in addition to conveying cellular memory, a main function of PcG correlates with a second wave of PRC1 recruitment during larval stage to subtly regulate and coordinate the expression of cancer-related and developmental genes through non-canonical molecular mechanisms.

In the Western world, breast cancer is the most frequent malignancy in women and still the leading cause of cancer related deaths, therefore, a better understanding of the disease is needed. Adequate therapeutic targets for all breast cancer types have not been identified yet, and patients with the same type of cancer have often different outcomes. Polycomb proteins are emerging as important factors involved in breast cancer formation. Polycomb proteins play a crucial role in embryogenesis, early development, stem cell renewal and establishing and maintaining cell identity. Their alteration leads to mis-regulation of several important cellular factors including tumour suppressors, DNA repair factors, cell cycle regulation factors and cell-cell interaction factors. In this thesis the importance of several polycomb proteins in breast cancer has been investigated. The effect of EZH2 knockdown has been tested in breast cancer cell lines expressing different level of the protein and with different features. The results obtained are in line with other studies and suggest that the effect of EZH2 down-regulation in breast cancer cells is dependent on cellular context. In vitro experiments, using both established breast cell lines and primary epithelial cells have been used for investigating the importance of CBX8 in breast cancer. The results obtained showed that the polycomb proteins CBX8 does not play a central role in malignant transformation of the mammary epithelial cells tested.

Transcriptional repression is maintained by many molecular processes, including DNA methylation and polycomb repression. These two systems are both associated with chromatin modification at the promoters of silent genes, and are both essential for mammalian development. Previous work has shown that DNMT proteins are required for correct targeting of polycomb repressive complexes (PRCs). In this thesis, I investigate whether targeting of DNA modification has a reciprocal dependence on the polycomb machinery by mapping DNA modification in wild-type and PRC-mutant ES cells (Ring1B null, EED null, and Ring1B/EED duble null). I find that the loss of PRCs results in increased DNA modification at sites normally targeted by de novo DNA methyltransferase which lose H3K4 methylation upon PRC removal. This increased DNA modificaiton is associated with increased gene expression when found at CpG island shores of genes marked by the PRC-mediated histone modifications H3K27me3 and H2AK119ub, but not genes lacking these marks. Gene misregulation may be further linked to DNA modification changes by increased DNA modification at enhancers. While loss of either Ring1B or EED led primarily to increases in DNA modification at regions dependant on DNMT3A/DNMT3B, the combined loss of Ring1B and EED results in widespread loss of DNA modification at sites more dependent on DNMT1 activity. This thesis suggests an interplay between PRCs and DNA modification placement which is relevant to the cntrol of gene expression.

Polycomb proteins are epigenetic regulators that are critical in mediating gene repression at critical stages during development. Core and accessory proteins make up the Polycomb Repressive Complex 2 (PRC2), which is responsible for trimethylation of lysine 27 on histone 3 (H3K27me3), leading to maintenance of chromatin compaction and sustained gene repression. Classically, Polycomb accessory proteins are often thought of as having minor roles in fine-tuning the repressive action of PRC2. Their actions have often been attributed to chromatin recognition, targeting to specific loci and enhancing methyltransferase activity. In our previous work in mouse embryonic stem cells (ESCs), we showed that Polycomb-like 2 (Mtf2/Pcl2) is critical for PRC2-mediated regulation of stem cell self-renewal through feed-forward control of the pluripotency network. In moving beyond the ESC model system, we sought to interrogate the role of Mtf2 in vivo by creating a gene-targeted knockout mouse model. Surprisingly, we discovered a tissue-specific role for Mtf2 in controlling erythroid maturation and hematopoietic stem cell self-renewal. Via its regulation of other PRC2 members, Mtf2 is critical for global H3K27me3 methylation at promoter-proximal sites in developing erythroblasts. Thus, Mtf2 is required for proper maturation of erythroblasts. Loss of Mtf2 also reduces HSC self-renewal leading to stem cell pool exhaustion. Additionally, misregulation of Mtf2 in leukemia models contributes to massive leukemic blast expansion at the expense of leukemic stem cell self-renewal. In the developing hematopoietic system, Mtf2 functions as a core complex member, controlling epigenetic regulation of self-renewal and maturation of both stem and committed cells.

Chromatin structure and epigenetic mechanisms play an important role in initiating and maintaining the intricate patterns of gene expression required for embryonic development. One such mechanism, DNA methylation (5mC), involves the chemical modification of cytosine bases in DNA and is implicated in maintaining patterns of transcription. However, many fundamental aspects of DNA methylation are not fully understood, including the mechanisms by which it influences transcriptional states. Recent data suggest functional links between DNA methylation and a second epigenetic mechanism that has important roles in transcriptional repression, the polycomb group (PcG) repressor system. Here, I suggest that an intact DNA methylation system is required for the repression of many PcG target genes by influencing the genomic targeting of the polycomb repressor 2 complex (PRC2) and its signature histone modification, H3K27me3 (K27me3). I demonstrate differential genomic localisation of K27me3 at gene promoter regions in hypomethylated mouse embryonic fibroblast (MEF) cells deficient for the major maintenance DNA methyltransferase, Dnmt1. Globally, Dnmt1-/- MEFs have a higher level of the K27me3 mark than controls, as assessed by western blot and immunofluorescence. I observe increased K27me3 at a relatively small number of gene promoters in Dnmt1-/- MEFs that often are associated with high levels of DNA methylation in wildtype MEFs, consistent with the notion that DNA methylation is capable of antagonising PRC2 binding at certain loci. Conversely, I show that a large number of developmentally important genes that are normally repressed and highly bound by K27me3, including classic polycomb targets, the Hox genes, display dramatically reduced association with K27me3 in Dnmt1-/- MEFs. Many of these genes, but not all, show reciprocal increases in promoter H3K4me3 modification and are transcriptionally de-repressed in Dnmt1-/- MEFs. I suggest that these genes are mostly associated with CpG-rich promoters with low levels of DNA methylation in wildtype cells, implying that their silencing is not dependent on the canonical role of DNA methylation. Consistent with the findings of recently published work, I suggest a working model where PRC2 binding in wildtype cells is restricted by CpG methylation. According to this model, the differential genomic location of K27me3 in hypomethylated Dnmt1-/- MEFs is explained by a redistribution of PRC2 to normally DNA methylated, unbound loci, resulting in a titration effect and coincident loss of K27me3 from normal targets. It was also apparent that certain PRC2-target genes, including the developmentally important Hox gene clusters, are strongly affected in Dnmt1-/- MEFs, displaying striking loss of K27me3. As intergenic transcription has been implicated in relief from polycomb silencing and abundant intergenic transcription has been reported within Hox clusters, I measured RNA expression at Hox clusters and a small number of other PcG target genes in Dnmt1-/- MEFs using highdensity tiling arrays. In Dnmt1-deficient MEFs, widespread increases in intergenic transcription were observed within Hox clusters. In addition, mapping of the elongatingpolymerase- associated H3K36me3 histone modification showed widespread increases in this mark at intergenic and promoter regions in Dnmt1-/- MEFs. Increased local intergenic RNA and H3K36me3 were found to correlate with K27me3 loss for this cohort of genes. I suggest a working model where increased intergenic transcription and H3K36me3 in Dnmt1-/- MEFs leads to accelerated loss of K27me3 at certain loci, including Hox clusters. Taken together with recently published data, this work suggests that a major role of DNA methylation is in shaping the PRC2/K27me3 landscape. The potential implications of this putative role for DNA methylation are widespread, including our knowledge of how DNA methylation influences transcriptional regulation, and the consequence of rearranged DNA methylation patterns that are observed in many diseases including cancers.

Polycomb Group (PcG) proteins are responsible for post-transcriptional modifications in histone tails leading to chromatin condensation and changes in gene expression. In Arabidopsis thaliana, curly leaf (CLF) is a member of the Polycomb Reporssive Complex 2 (PRC2), which cnfers a repressive epigenetic mark, namely trimethylation of histone H3 at lysine 27 (H3K27me3). In the clf mutant, the expression of the floral organ identity gene AGAMOUS (AG) is derepressed in vegetative stages and coincides with loss of H3K27me3 at the AG locus. Recent whole genome prfiling studies have suggested that PcG genes regulate mang more developmental regulators than AG (about 15% of Arabidopis genes). However, it remains unclear what the relevance of PcG regulation of these targets is for plant development; in addition, it is not known how changes in J3K27me3 casue gene repression in plants. To unravel the role of CLFcin A. thaliana, a T-DNA mutagenesis in the clf background was performed to identify mutations enhancing or suppressing the Clf- phenotype, as these may identify additional PcG genes and targets. Firstly, I screened an A. thaliana T-DNA mutagenized population and identified four mutations suppressing the Clf- phenotype: suppressor of polycomb 1 to 4 (sop1, sop2, sop3 and sop4). Secondly, I characterized these four mutants. The sop1 mutant had normal flowering time and the suppressed phenotype is due to a loss of function mutation in SEPALLATA3 (SEP3). I establied the SEP3 is an activator and a co-factor of AG. Also, I found that SEP3 is stronlgy mis-expressed in clf mutants and SEP3 chromatin is enriched the H3K27me3, which stronly suggests that SEP3 is a direct target of CLF. In addition, I showed that a mutation in Flowering Locus T (FT), which is a positive regulator of SEP3, suppreesed the Clf- phenotype suggesting the FT is also a target of CLF. Suppressors sop3, sop3 and sop4 are late flowering, unlike sop1, and show increased expression of Flowering Locus C (FLC), a MADS-box transcription factor gene that represses flowering. I found that the sop4 mutation in likely casued by disruption of FPA, a predicted RNA binding protein that promotes flowering time by repressing FLC. Consistent with this, sop4 mutants show hight levels of FLC. Unexpectedly, fpa clf (sop4) mutatns are much later flowering than clf FRI mutants, which have similarly high levels of FLC. This suggests that FPA may regulate other genes controlling flowering thant FLC. The genes involved in sop2 and sop3 mutants remain to be identified. In this thesis I brought genetic and molecular evidence showing that CLF, though the PRC2, control floral induction (FLC), floral integration (FT) and floral organ formation (SEP3 and AG) in A. thaliana.