Les mélanomes oculaires sont des tumeurs rares représentant environ 5% des mélanomes. Les mélanomes oculaires peuvent provenir de deux tissus : l’uvée (~500 cas/an en France) et la conjonctive (~30 cas/an). Les mélanomes uvéaux présentent un très faible taux de mutations somatiques. Ces tumeurs sont également porteuses d’altérations du nombre de copies caractéristiques (gains du 8q, 1q, 6p, pertes du 3, du 1p, du 6q ou du 8p). L’évolution du génome de ces tumeurs durant la progression métastatique est à ce jour mal décrit. Afin d’explorer l’évolution métastatique du mélanome uvéal, nous avons séquencé les exomes de 14 tumeurs primaires et 79 métastases provenant de 24 patients. Il existait une grande proximité génétique entre tumeurs primaires et métastases avec une médiane de 11,5 mutations dans les tumeurs primaires, et 14 dans les métastases. Bien que les mutations SF3B1 et EIF1AX soient des facteurs pronostiques majeurs dans les mélanomes uvéaux, leurs fréquences dans les métastases étaient similaires à celle observée dans les séries historiques de tumeurs primaires. Les métastases présentaient quelques altérations de nombre de copies supplémentaires par rapport aux tumeurs primaires correspondantes. Parmi les altérations du nombre de copies les plus souvent acquises lors du processus métastatique, les gains du 8q étaient présents dans 92% des métastases. Lors de ce travail, nous avons découvert un mélanome uvéal présentant un phénotype hypermuté CpG>TpG chez une patiente ayant présenté une réponse exceptionnelle à une immunothérapie anti-PD1. Ce phénotype hypermutateur a été expliqué par une mutation germinale délétère de MBD4 (Methyl-CpG Binding Domain 4) avec une inactivation bi-allélique dans la tumeur. Deux autres tumeurs hypermutées CpG>TpG porteuses d’une mutation de MBD4 germinale, un mélanome uvéal et un glioblastome, ont été identifiées dans les bases de données publiques. La biologie des mélanomes conjonctivaux et leurs profils génomiques sont mal connus. Nous avons séquencé les génomes de 6 tumeurs, puis procédé à un séquençage ciblé de 47 autres tumeurs. Nous avons montré que ces tumeurs présentent un profil hypermuté C>T induit par l’exposition aux ultra-violets. Ces tumeurs présentaient un profil de mutations proche des mélanomes cutanés avec une fréquence moindre de mutations BRAF (33%), et des mutations plus spécifiques des mélanomes muqueux telles que des mutations activatrices de KIT et SF3B1 dans les mélanomes conjonctivaux non exposés au soleil. Nous avons également identifié des mutations de CTNNB1 dans les tumeurs développées sur des nevi conjonctivaux. L’ensemble de ces travaux illustrent comment la description moléculaire des tumeurs rares permet d’envisager de nouvelles stratégies de médecine de précision. / Ocular melanomas are rare tumors representing about 5% of all melanomas. Ocular melanomas may arise from two tissues: the uvea (~ 500 cases / year in France) and the conjunctiva (~ 30 cases / year). Uveal melanomas have a very low rate of somatic mutations. These tumors also carry specific distinctive copy number alterations (gains of 8q, 1q, 6p, losses of 3, 1p, 6q or 8p). The evolution of the genome of these tumors during metastatic progression has been poorly described.To explore the metastatic evolution of uveal melanoma, we whole-exome sequenced 14 primary tumors and 79 metastases from 24 patients. Primary tumors and metastases presented close genetic profiles with a median of 11.5 mutations in primary tumors, and 14 in metastases. Although SF3B1 and EIF1AX mutations are major prognostic factors in uveal melanomas, their frequencies in metastases were similar to those observed in historical primary tumors. The metastases showed some additional copy number alterations compared to the corresponding primary tumors. Among the alterations acquired during the metastatic process, 8q gains were present in 92% of metastases.Thanks to this work, we found a uveal melanoma with a CpG> TpG hypermutated phenotype in a patient who had an exceptional response to anti-PD1 immunotherapy. This hypermutated phenotype was explained by a deleterious germline mutation of MBD4 (Methyl-CpG Binding Domain 4) with bi-allelic inactivation in the tumor. Two other hypermuted CpG> TpG tumors with germline MBD4 mutation, a uveal melanoma and a glioblastoma, were identified in public databases.The biology of conjunctival melanomas and their genomic profiles have been scarcely described. We sequenced the genomes of 6 tumors and then target-sequenced 47 other tumors. We showed that these tumors had a C> T hypermuted profile induced by ultraviolet exposure. These tumors presented a pattern of mutations close to cutaneous melanomas with a lower frequency of BRAF mutations (33%), and mutations that were more specific of mucosal melanomas such as activating mutations of KIT and SF3B1 in conjunctival melanomas not exposed to the sun. We also identified CTNNB1 mutations in tumors developed on conjunctival nevi.All of these works illustrate how the molecular description of rare tumors opens new avenues for precision medicine.

Chez les mammifères, la méthylation est une marque épigénétique ciblant la cytosine principalement dans un contexte CpG pour produire une 5mC. 5mC est très sensible à une déamination spontanée ou enzymatique, conduisant à la formation d'un mésappariement G/T. La 5mCpeut également être oxydée pour former successivement la 5hmC, la 5fC et la 5caC. Ces modifications de la 5mC participent aux processus actifs de déméthylation de l’ADN. Chez les mammifères, la thymine, dans le mésappariement G/T, est clivée par TDG et MBD4. TDG est également en mesure d'exciser 5fC et 5caC. Cette thèse avait pour but de clarifier la fonction de TDG et MBD4 dans la dynamique de la 5mC. Nous avons montré que MBD4 est associée aux protéines de réparation des mésappariements. Les tests enzymatiques, in vitro, montrent que le complexe MBD4/MMR a une activité bifonctionnelle (glycosylase/lyase) spécifique pour G/T, qui est régulée par la méthylation. Pour TDG, nous avons ciblé cette enzyme dans les cellules MEF et caractérisé la distribution des cytosines modifiées. Les résultats montrent des profils de méthylation/oxydation d'ADN qui sont régulés par TDG et surviennent principalement au niveau des répétitions de CA et dans les rétroéléments spécifiques de la lignée souris. / In mammals, methylation is an epigenetic mark targeting cytosine mainly in a CpG context, producing 5mC. 5mC is highly sensitive to a spontaneous or enzymatic deamination leading to G/Tmismatch. 5mC can also be oxidized to 5- 5hmC, 5fC and 5caC. These modifications of 5mC participate in the active demethylation processes. In mammals, the thymine in G/T mismatch is cleaved by TDG and MBD4 glycosylases. TDG is able also to excise the 5fC and 5caC.This thesis was to clarify the function of TDG and MBD4 in the dynamics of 5mC. We showed that MBD4 is associated with PMS2, MLH1, MSH2 and MSH6 proteins, four proteins involved in DNA mismatch repair (MMR). The in vitro enzymatic tests show that MBD4/MMR complex has a bifunctional glycosylase/lyase activity specific for G/T and is regulated by methylation.For TDG, we targeted this enzyme in MEF cells and characterized the distribution of modified cytosines. The results show that DNA methylation/oxidation patterns are regulated by TDG and occur mainly at CA repeats and at the mouse-lineage specific retro-elements.

The base excision repair protein, MBD4 (also known as MED1) is known to be transcriptionally coupled to a mismatch repair protein MLH1. To date the significance of this coupling has not been elucidated and the significance of MBD4 within the mismatch repair system and apoptotic pathway is still being understood. Recently a novel alternatively spliced form of MBD4 has been identified and termed sMBD4. To date the significance of sMBD4 is unknown. MBD4 and sMBD4 share a common glycosylase domain and this is the domain through which MBD4 is reported to interact with MLH1. It was the aim of this study to determine if sMBD4 was also a binding partner of MLH1 to help elucidate a potential role of sMBD4 and to further characterise the binding domain between MLH1 and MBD4. Recombinant proteins were utilised in binding assays however, a specific protein – protein interaction could not be determined. Regular aspirin intake is associated with a reduction in the incidence of colorectal cancer. Aspirin has been shown to be cytotoxic to colorectal cancer cells in vitro. The molecular basis for this cytotoxicity is controversial, with a number of competing hypotheses in circulation. One suggestion is that the protective effect is related to the induction of DNA mismatch repair (MMR) proteins in DNA MMR proficient cells. As MBD4 has previously been suggested to be coupled to MLH1 expression by a post‐translational mechanism the cytotoxicy of aspirin in relation to MBD4 expression was examined. This study reports that aspirin does not up‐regulate MBD4 gene transcription in vitro in the DNA mismatch repair proficient/p53 mutant colorectal cancer cell line SW480. However, MBD4 gene transcription was up‐regulated upon treatment with the aspirin precursor, salicylic acid. The suggested involvement of the DNA repair proteins in the mechanism of action of aspirin promoted the investigation into the expression of DNA damage signalling pathways genes upon aspirin exposure. This study utilised a commercially available PCR array to analyse the expression of 84 DNA damage signalling genes in the SW480 colorectal cancer cell line upon aspirin treatment. It is reported that treatment of the SW480 cell line with aspirin caused changes in mRNA expression of several key genes involved in DNA damage signalling including a significant down‐regulation in expression of the genes encoding ATR, BRCA1 and MAPK12 and increases in the expression of XRCC3 and GADD45α genes. Regulation of these genes could potentially have profound effects on colorectal cancer cells and may play a role in the observed chemo‐protective effect of aspirin in vivo.Further to this, protein expression was analysed to determine if correlation could be established with the changes in mRNA expression observed. Although a correlation was not seen between transcript and protein levels of ATR, BRCA1 and GADD45α, an increase in XRCC3 protein expression upon aspirin treatment in SW480 cells was observed by immunoblotting, immunofluorescence and immunohistochemical analysis. This study indicates that alterations in gene expression seen in microarray studies need to be verified at the protein level. Furthermore, this study reports the novel discovery of XRCC3 gene and protein expression being susceptible to exposure to the non‐steroidal anti‐inflammatory drug, aspirin.

DNA methylation is a major form of epigenetic modification and involves the addition of a methyl group covalently to the 5-position of the cytosine pyrimidine ring, mostly within the context of CpG dinucleotides in vertebrate somatic cells. Methylation of CpG dinucleotides at promoter regions is generally associated with transcriptional repression. In this context, the methyl-CpG binding proteins (MeCPs) that are capable of recognition of methylated CpG dinucleotides are proposed to play a central role in DNA methylation associated transcriptional repression. Methyl-CpG binding domain protein 4 (MBD4) is an MeCP that possesses a glycosylase domain at its C-terminal, which can excise and repair both G:T and G:U mutations derived from DNA deamination at CpG dinucleotides, in addition to its Nterminal MBD binding domain. MBD4 has been associated with a number of pathways including DNA repair, apoptosis, transcriptional repression, and possibly DNA demethylation processes. However, the precise contribution of MBD4 to these processes remains unclear. To explore the functional repertoire of MBD4 I decided to undertake multiple protein interaction studies to identify potential partner proteins. I performed yeast 2-hybrid screens with an 11.5 day mouse embryonic cDNA library and multiple mass spectrometry of immunoprecipitates of tagged versions of MBD4 that were over-expressed in human cell lines. I detected ~380 potential interacting candidates with these assays. A significant number of candidates were detected in both assay systems. Chosen candidates were further validated by reciprocal co-IP of expressed partners and by immunofluorescence (IF) microscopy to determine their potential co-localisation in mouse and human cell lines. Subsequently, I identified the intervening domain of MBD4 as a novel protein interaction region for tested candidates. My analysis suggests that MBD4 can have a role in regulation of post-replication methyl-error repair/methylation machinery through its direct interaction with DNMT1 (previously shown), UHRF1 (novel) and USP7 (novel), as well as possible cross-talk to histone modification and chromatin remodelling pathways, through partners such as PRMT5 and ACF1. Interestingly the transcription regulatory components KAP1 and CFP1 not only interact with but also dramatically influence the stability of exogenously expressed MBD4 in human cells. In general positive validation by IP and IF demonstrates the robustness of the initial screens, and implies that MBD4 may impact upon several transcriptional and epigenetic networks along with a number of nuclear pathways that include transcriptional repression, DNA repair and RNA processing. To test for transcriptional aberration in the absence of Mbd4 function I profiled two independent mouse cell lines that lack MBD4 activity using Illumina MouseWG-6 v2.0 Expression BeadChip arrays. A number of genes were identified that are significantly up- or down- regulated in both Mbd4-/- MEFs. This included mis-expression of insulin-like growth factor-binding proteins and two paternally imprinted genes Dio3 and H19. The cohort of genes that were mis-expressed in the Mbd4-/- MEFs overlap with genes that responsed to tamoxifen exposure in an ER-positive ZR-75-1 xenograft model. In response to this observation I identified a potential interaction between MBD4 and estrogen receptor α (ERα) by co-IP and IF co-localisation. This suggests that MBD4 might potentiate transcription of estrogen regulated genes via a direct interaction with ERα, supporting a possible link between replication repair remodelling and steroid/thyroid hormone receptor transcriptional regulation. Additionally I performed a pathway analysis by which several developmental genes including Sox9, Klf2 and Klf4, were prioritised as possible MBD4 targets. On this basis I propose a role for MBD4 in acquired diseases such as cancers and autoimmune diseases via transcriptional regulation. I also performed a comparison of MBD4 DNA binding activity with MBD4 homologues from the Medaka fish (Oryzias latipes) and the amphibian, Xenopus laevis. I could show that DNA binding specificity to a series of methylated and mismatched probes is conserved regardless of the poor sequence conservation of the MBD domain of MBD4 between the species. I conclude that MBD4 is integrated in multiple pathways in the nucleus that includes DNA repair, chromatin remodelling, transcriptional regulation and genome stability.