Le système de différenciation entre le « soi » et le « non-soi » des vertébrés permet la détection et le rejet de pathogènes et de cellules allogéniques. Il requiert la surveillance de petits peptides présentés à la surface cellulaire par les molécules du complexe majeur d’histocompatibilité de classe I (CMH I). Les molécules du CMH I sont des hétérodimères composés par une chaîne lourde encodée par des gènes du CMH et une chaîne légère encodée par le gène β2-microglobuline. L’ensemble des peptides est appelé l’immunopeptidome du CMH I. Nous avons utilisé des approches en biologie de systèmes pour définir la composition et l’origine cellulaire de l’immunopeptidome du CMH I présenté par des cellules B lymphoblastoïdes dérivés de deux pairs de fratries avec un CMH I identique. Nous avons découvert que l’immunopeptidome du CMH I est spécifique à l’individu et au type cellulaire, qu’il dérive préférentiellement de transcrits abondants, est enrichi en transcrits possédant d’éléments de reconnaissance par les petits ARNs, mais qu’il ne montre aucun biais ni vers les régions génétiques invariables ni vers les régions polymorphiques. Nous avons également développé une nouvelle méthode qui combine la spectrométrie de masse, le séquençage de nouvelle génération et la bioinformatique pour l’identification à grand échelle de peptides du CMH I, dont ceux résultants de polymorphismes nucléotidiques simples non-synonymes (PNS-ns), appelés antigènes mineurs d’histocompatibilité (AMHs), qui sont les cibles de réponses allo-immunitaires. La comparaison de l’origine génomique de l’immunopeptidome de soeurs avec un CMH I identique a révélé que 0,5% des PNS-ns étaient représentés dans l’immunopeptidome et que 0,3% des peptides du CMH I seraient immunogéniques envers une des deux soeurs. En résumé, nous avons découvert des nouveaux facteurs qui modèlent l’immunopeptidome du CMH I et nous présentons une nouvelle stratégie pour l’indentification de ces peptides, laquelle pourrait accélérer énormément le développement d’immunothérapies ciblant les AMHs. / The self/nonself discrimination system of vertebrates allows detection and rejection of pathogens and allogeneic cells. It requires the surveillance of short peptides presented by major histocompatibility class I (MHC I) molecules on the cell surface. MHC I molecules are heterodimers that consist of a heavy chain produced by MHC genes and a light chain encoded by the β2-microglobulin gene. The peptides presented by MHC I molecules are collectively referred to as the MHC I immunopeptidome. We employed systems biology approaches to define the composition and cellular origin of the self MHC I immunopeptidome presented by B lymphoblastoid cells derived from two pairs of MHC-identical siblings. We found that the MHC I immunopeptidome is subject- and cell-specific, derives preferentially from abundant transcripts, is enriched in transcripts bearing microRNA response elements and shows no bias toward invariant vs. polymorphic genomic sequences. We also developed a novel personalized approach combining mass-spectrometry, next-generation sequencing and bioinformatics for high-throughput identification of MHC I peptides including those caused by nonsynonymous single nucleotide polymorphisms (ns-SNPs), termed minor histocompatibility antigens (MiHAs), which are the targets of allo-immune responses. Comparison of the genomic landscape of the immunopeptidome of MHC-identical siblings revealed that 0.5% of ns-SNPs were represented in the immunopeptidome and that 0.3% of the MHC I-peptide repertoire would be immunogenic for one of the siblings. We discovered new factors that shape the self MHC I immunopeptidome and present a novel strategy for the identification of MHC I-associated peptides that could greatly accelerate the development of MiHA-targeted immunotherapy.

La génomique d'association cherche à déterminer des liens entre le génome et des traits ou phénotypes, notamment dans le contexte de maladies. Aujourd'hui, les études les plus fréquentes en génomique d’association sont les études génome entier, qui analysent autant de variants du génomes (SNPs) que possible, sans avoir de préjugé a priori sur leur fonction biologique. Cependant, les méthodes de génotypage utilisées pour ces études ne permettent pas toujours d'obtenir des informations précises dans une région hypervariable comme la région HLA, qui joue un rôle crucial dans l'immunité, et les variants génétiques de ces régions sont souvent prédits par des approches bioinformatiques. J'ai durant ma thèse créé un nouvel outil, HLA-Check, permettant d'évaluer, à partir des génotypes obtenus par puce de génotypage, la plausibilité de données d'allèles HLA d'un même individu, et démontré que cette technique permettait d'identifier plus précisément les individus dont les allèles HLA avaient été mal caractérisés afin de les retyper ou de les écarter de l'étude. Un article documentant cet outil a été publié dans BMC Bioinformatics. J'ai également effectué une étude d'association génome entier sur le déclenchement de la cirrhose chez les patients co-infectés par le VIH (virus de l'immunodéficience humaine) et le VHC (virus de l'hépatite C). La coinfection par ces deux maladies est fréquente en raison de modes d’infection similaires, et l'infection par le VIH stimule l'activité du VHC et accélère la fibrose du foie puis sa cirrhose, causant la mort des patients co-infectés. Mon étude porte sur 306 patients co-infectés issus de la cohorte ANRS CO-13 HEPAVIH. J'ai ainsi pu mettre en évidence trois signaux associés avec le déclenchement de la cirrhose, dont deux ont un lien pertinent avec les maladies hépatiques (gene CTNND2 et gene MIR7-3HG). L'identification de ces nouveaux variants devrait permettre une meilleure compréhension des mécanismes moléculaires de la cirrhose, et contribuer au développement de nouvelles stratégies diagnostiques ou thérapeutiques. L’article documentant cette étude est en cours de publication. / Association genomics aims at finding links between the genome and some traits or illnesses. Today, the most frequent studies in this field are genome wide association studies (GWAS), which analyze as many genome variants (mainly Single Nucleotide Polymorphisms) as possible, without any a priori on their biological function. However, genotyping methods used in these studies may be insufficient to get reliable information in higly variable regions such as the HLA which plays a crucial role in immunity, and the genetic variants of such regions are often predicted using bioinformatics approaches. During my PhD, I have created a new tool, HLA-Check, that allows to rate the plausibility of HLA alleles from the genotypes obtained from genotyping chips. I also assesses its performances and showed that it was able to point out individuals with a wrong HLA typing, in order to retype them or remove them from the study. An article documenting this tool was published in BMC Bioinformatics. I have also performed a genome-wide association study on cirrhosis outbreak in individuals coinfected with HIV (human immunodeficiency virus) and HCV (hepatitis C virus). Because of similar infection routes (blood-related), co-infection with those two viruses are frequent, and the infection by HIV enhances HCV activity and increases liver fibrosis leading to cirrhosis and death of co-infected patients. Our study has dealt with 306 co-infected patients from the ANRS CO-13 HEPAVIH cohort. I could point out three statistically significant signals, two of them being highly relevant for their involvement in liver diseases (gene CTNND2 and gene MIR7-3HG). The identification of these new variants should lead to a better understanding of the molecular mechanisms involved in cirrhosis, and should contribute to the rational developement of new diagnostic or therapeutic strategies. A publication is under way.

Japanese encephalitis virus (JEV) causes viral encephalitis in new born and young adults that is prevalent in different parts of India and other parts of South East Asia with an estimated 6000 deaths per year. JEV is a single stranded RNA virus that belongs to the Flavivirusgenus of the family Flaviviridae. It is a neurotropic virus which infects the central nervous system (CNS). The virus follows a zoonotic life-cycle involving mosquitoes and vertebrates, chiefly pigs and ardeid birds, as amplifying hosts. Humans are dead end hosts. After entry into the host following a mosquito bite, JEV infection leads to acute peripheral leukocytosis in the brain and damage to Blood Brain Barrier (BBB). The exact role of the endothelial cells during CNS infection is still unclear. However, disruption of this endothelial barrier has been shown to be an important step in entry of the virus into the brain. Humoral and cell mediated immune responses during JEV infection have been intensively investigated. Previous studies from our lab have shown the activation of cytotoxic T-cells (CTLs) upon JEV infection. MHC molecules play pivotal role in eliciting both adaptive (T-cells) and innate (NK cells) immune response against viral invasion. Many viruses such as HIV, MCMV, HCMV, AdV and EBV have been found to decrease MHC expression upon infection. On the contrary, flaviviruses like West Nile Virus (WNV) have been found to increase MHC-I and MHC-II expression. More recently, data from our lab has shown that JEV infection can lead to upregulation of mouse non-classical MHC class Ib molecules like Qb1, Qa1 and T-10 along with classical MHC molecules. Non-classical MHC molecules are important components of the innate and adaptive immune systems. Non-classical MHC molecules differ from their classical MHC class I counterparts by their limited polymorphism, restricted tissue distribution and lower levels of cell surface expression. Human classical MHC class I molecules are HLA-A, -B and –C while non-classical MHC Class Ib molecules are HLA-E, -G and –F. HLA-E, the human homologue of the mouse non-classical MHC molecule, Qa-1b has been shown to be the ligand for the inhibitory NK, NKG2A/CD94 and may bridge innate and adaptive immune responses. In this thesis, we have studied the expression of human classical class I molecules HLA-A, -B, -C and the non-classical HLA molecule, HLA-E in immortalized human brain microvascular endothelial cells (HBMEC), human endothelial like cell line ECV304 (ECV), human glioblastoma cell line U87MG and human foreskin fibroblast cells (HFF). We observed an upregulation of classical HLA molecules and HLA-E mRNA in endothelial and fibroblast cells upon JEV infection. This mRNA increase also resulted in upregulation of cell surface classical HLA molecules and HLA-E in HFF cells but not in both the human endothelial cell lines, ECV and HBMECs. Release of soluble classical HLA molecules upon cytokine treatment has been a long known phenomenon. Recently HLA-E has also been shown to be released as a 37 kDa protein from endothelial cells upon cytokine treatments. Our study suggests that JEV mediated upregulation of classical HLA and HLA-E upregulation leads to release of both Classical HLA molecules and HLA-E as soluble forms in the human endothelial cell lines, ECV and HBMEC. This shedding of sHLA-E from human endothelial cells was found to be mediated by matrix metalloproteinase (MMP) proteolytic activity. MMP-9, a protease implicated in release of sHLA molecules was also found to be upregulated upon JEV infection only in endothelial cell lines but not in HFF cells. Our study provides evidence that the JEV mediated solubilisation of HLA-E could be mediated by MMP-9. Further, we have tried to understand the role of the MAPK pathway and NF-κB pathway in the process of HLA-E solubilisation by using specific inhibitors of these pathways during JEV infection of ECV cells. Our data suggests that release of sHLA-E is dependent on p38 and JNK pathways while ERK 1/2 and NF-κB pathway only had a minor role to play in this process. Treatment of endothelial cells with TNF-α, IL-1β and IFN-γ is known to result in release of sHLA-E. In addition to TNF-α and IFNtreatment, we observed that activating agents like poly (I:C), LPS and PMA also resulted in the shedding of sHLA-E from ECV as well as U87MG but not from HFF cells. Treatment of endothelial cells with IFN-β, a type-I interferon also led to release of sHLA-E. IFN-γ, a type II interferon and TNF-α are known to show additive increase in solubilisation of HLA-E. We studied the interaction between type I interferon, IFN-β and TNF-α with regard to shedding of sHLA- E. Both IFNand TNF, when present together caused an additive increase in the shedding of sHLA-E. These two cytokines were also found to potentiate the HLA-E and MMP-9 mRNA expression. Hence, our data suggest that these two cytokines could be working conjunctly to release HLA-E, when these two cytokines are present together as in the case of virus infection of endothelial cells. HLA-E is known to be a ligand for NKG2A/CD94 inhibitory receptors present on NK and a subset of T cells. Previous reports have suggested that NKG2A/CD94 mediated signaling events could inhibit ERK 1/2 phosphorylation leading to inhibition of NK cell activation. IL-2 mediated ERK 1/2 phosphorylation is known to play a very important role in maintenance and activation of NK cells. We studied the effects of sHLA-E that was released, either by JEV infection or IFN-γ treatment on IL-2 mediated ERK 1/2 phosphorylation in two NK cell lines, Nishi and NKL. The soluble HLA-E that was released upon JEV infection was functionally active since it inhibited IL-2 and PMA induced phosphorylation of ERK 1/2 in NKL and Nishi cells. Virus infected or IFN-γ treated ECV cell culture supernatants containing sHLA-E was also found to partially inhibit IL-2 mediated induction of CD25 molecules on NKL cells. CD25 is a component of the high affinity IL-2 receptor and hence could play an important role in proliferation and activation of NK cells. sHLA-E was also found to inhibit IL-2 induced [3H]-thymidine incorporation suggesting that, similar to cell surface expressed HLA-E, sHLA-E could also inhibit the proliferation and activation of NK cells. In summary, we found that establishment of JEV infection and production of cytokines like IFN-β, TNF-α, IL-6 along with MMP-9 in human endothelial cells. These cytokines may also indirectly lead to the reported damage and leukocyte infiltration across infected and uninfected vicinal endothelial cells. The increased surface expression of HLA-E in fibroblast and release of sHLA and sHLA-E molecules from endothelial cells may have an important immunoregulatory role. HLA-E is an inhibitory ligand for NKG2A/CD94 positive CD8+ T and NK cells. Hence our finding that sHLA-E can inhibit NK cell proliferation suggests an immune evasive strategy by JEV.

The topic of the presented thesis is the regulation of gene expression level of the three HSP70 genes in mononuclear cells. We investigated the dependence of expression regulation (induction) abiliy on selected point mutations, so-called SNPs (single nucleotide polymorphism) in the observed genes. The mononuclear cells were obtained from peripheral blood samples of healthy individuals. In order to analyze their gene expression, we selected individuals who were homozygous for at least one of the monitored point mutations. Taking into account the chosen criteria for healthy individuals we based on interviews with these individuals and their personal history. We determined the polymorphisms observed in two cell stress-inducible HSP70-1 (HSPA1A) and HSP70-2 (HSPA1B) genes and in one constitutively expressed gene HSP70-Hom (HSPA1L). Further, we have analyzed HSP70s gene expression regulation and the relation between the expression regulation and studied polymorphisms. We determined the degree of regulation of a gene expression in the studied genes in relation to two SNPs -110A/C (rs1008438), +190G /C (rs1043618) gene HSP70-1, and two SNPs +1267A/G (rs1061581), +2074G /C (rs539689 ) of the HSP70-2 gene, and the mutation of one five-nucleotide (rs9281590) HSP70-2 gene, and one SNP +2437T/C (rs2227956) of...

<p>  </p> <p>Metastatic breast cancer (MBC) is an extremely recalcitrant disease capable of overcoming targeted therapies and evading immune surveillance via the engagement of complicated signaling networks. Resistance to targeted therapies and therapeutic failure of immune checkpoint blockade (ICB) are two major challenges in treating MBC. To survive in the dynamic tumor microenvironment (TME) during metastatic progression, shared signaling nodes are required for MBC cells to regulate the signaling networks efficiently, which are potential multifunctional therapeutic targets. SH2 containing protein tyrosine phosphatase-2 (SHP2) is a druggable oncogenic phosphatase that is a key shared node in both tumor cells and immune cells. How tumor-cell autonomous SHP2 manages its signaling inputs and outputs to facilitate the growth of tumor cells, drug resistance, immunosuppression, and the limited response of ICB in MBC is not fully understood. Herein, we used inducible genetic depletion and two distinct types of pharmacological inhibitors to investigate anti-tumor effects with immune reprogramming during SHP2 targeting. </p> <p>We first focus on the signaling inputs and outputs of SHP2. We find that phosphorylation of SHP2 at Y542 predicts the survival rates of breast cancer patients and their immune profiles. Phosphorylation of SHP2 at Y542 is elevated with differential activation mechanisms under a growth-factor-induced and extracellular matrix (ECM)-rich culture environment. Phosphorylation of SHP2 at Y542 is also elevated in HER2 positive MBC cells upon acquired resistance to the HER2 kinase inhibitor, neratinib. The resistant cells can be targeted by SHP2 inhibitors. SHP2 inhibitors block ERK1/2 and AKT signaling and readily prevented MBC cell growth induced by multiple growth factors. Inhibition of SHP2 also blocks these signaling events generated from the ECM signaling. In fact, the inhibitory effects of SHP2 blockade are actually enhanced in the ECM-rich culture environment. We utilize the <em>in vitro</em> T-cell killing assays and demonstrate that pretreatment of tumor cells with FGF2 and PDGF reduces the cytotoxicity of CD8+ T cells in a SHP2-dependent manner. Both growth factors and ECM-rich culture environment transcriptionally induce PD-L1 via SHP2. SHP2 inhibition balances MAPK signaling and STAT1 signaling, which prevents growth factor-mediated suppression of INF-γ-induced expression of MHC class I. </p> <p>Next, we evaluate the efficacy of SHP2 inhibitors. Blockade of SHP2 in the adjuvant setting decreased pulmonary metastasis <em>in vivo</em> and extended the survival of systemic tumor-bearing mice. Tumor-cell autonomous depletion of SHP2 reduces pulmonary metastasis and relieves exhaustion markers on CD8+ and CD4+ cells. Meanwhile, both systemic SHP2 inhibition and tumor-cell autonomous SHP2 depletion reduce tumor-infiltrated CD4+ T cells and M2-polarized tumor associated macrophages. </p> <p>Finally, we investigate potential combination therapies with SHP2 inhibitors. The combination of SHP2 inhibitors and FGFR-targeted kinase inhibitors synergistically blocks the growth of MBC cells. Pharmacological inhibition SHP2 sensitizes MBC cells growing in the lung to α-PD-L1 antibody treatment via relieving T cell exhaustion induced by ICB. </p> <p>Overall, our findings support the conclusion that MBC cells are capable of simultaneously engaging several survival pathways and immune-suppressive mechanisms via SHP2 in response to multiple growth factors and ECM signaling. Inhibition of SHP2, potentially in combination with other targeted agents and ICB, holds promise for the therapeutic management of MBC.</p>

La présentation de l’antigène est un mécanisme par lequel les cellules nucléées présentent un court peptide sur la molécule de classe 1 du Complexe Majeur d’Histocompatibilité (CMH-1) codée par les gènes « antigènes d’histocompatibilité humains ». Le terme “immunopeptidomique” est utilisé pour décrire l’ensemble des peptides associés aux molécules du CMH-1. Les cellules T CD8+ patrouillent l’organisme, s’attachent à la molécule CMH-1 par leur récepteur T et détruisent les cellules affichant un peptide atypique. Ce domaine présente un grand intérêt au niveau du traitement des infections virales et dans la conception de vaccins. Compte tenu que les coronavirus ont été à l’origine de trois épidémies durant les 20 dernières années, et que de multiples souches circulent chez l’humain ainsi que dans le règne animal, il est impératif de développer des vaccins universels qui pourrait prévenir de futurs événement épidémiologiques mondiaux reliés aux coronavirus. L’immunopeptidomique souffre d'un manque de protocoles normalisés et de contrôle et d’assurance de la qualité des échantillons afin de libérer tout son potentiel dans la recherche biomédicale. Dans le cadre de cette étude, la spectrométrie de masse, la cytométrie de flux et des approches bio-informatiques ont été utilisées pour développer des protocoles de contrôle de qualité pour la caractérisation de l’immunopeptidome de cellules infectées par les coronavirus. Nous avons isolé et analysé l’immunopeptidome de cellules MRC-5 avant et après infection par le coronavirus humain OC-43. En plus d’observer une forte baisse de l’abondance des molécules HLA et de la variété des peptides présentés après l’infection, 9 peptides viraux ont été isolés à partir des molécules du CMH-1. Ces peptides pourraient être utilisés afin de contribuer à formuler un vaccin pan-coronavirus qui élicite une réponse balancée entre la réponse humorale et la réponse cytotoxique. / Antigen presentation is a mechanism by which nucleated cells present a short peptide on the Major Histocompatibility Complex (MHC-1) class 1 molecule encoded by the “human histocompatibility antigen” genes. The term "immunopeptidomics" is used to describe the set of peptides associated with MHC-1 molecules. CD8+ T cells patrol the body, attach to the MHC-1 molecule through their T receptor and destroy cells displaying an atypical peptide. This field is of great interest in the treatment of viral infections and in vaccine design. Given that coronaviruses have been responsible for three epidemics in the last 20 years, and that multiple strains circulate in humans and animals, it is imperative to develop universal vaccines that could prevent future global epidemiological events related to coronaviruses. Immunopeptidomics suffers from a lack of standardized protocols and sample quality control and assurance to unleash its full potential in biomedical research. In this study, mass spectrometry, flow cytometry, and bioinformatics approaches were used to develop quality control protocols for characterizing the immunopeptidome of coronavirusinfected cells. We isolated and analyzed the immunopeptidome of MRC-5 cells before and after infection with human coronavirus OC-43. In addition to observing a strong decrease in the abundance of HLA molecules and in the variety of peptides presented after infection, 9 viral peptides were isolated from MHC-1 molecules. These peptides could be used to help formulate a pan-coronavirus vaccine that elicits a balanced response between humoral and cytotoxic responses.

Presque toutes les cellules nucléées expriment des peptides associés au CMH I (HLA I chez l’humain)(MAP) qui sont échantillonnés à partir du protéome cellulaire et transportés vers la surface cellulaire pour inspection par les lymphocytes T CD8. En tant que tel, la collection de MAP à la surface des cellules, ou immunopeptidome, informe les lymphocytes T CD8 de l’état cellulaire interne. L’immunosurveillance du cancer repose sur la capacité des lymphocytes T CD8 à reconnaître les MAP anormaux sur les cellules tumorales et à les éliminer tout en épargnant les cellules saines. Par conséquent, l’existence du cancer indique que bien souvent, les lymphocytes T CD8 spécifiques à la tumeur sont impuissants, dysfonctionnels ou incapables d’exercer leur fonction. Les vaccins anticancéreux peuvent actionner la destruction des tumeurs en stimulant la reconnaissance des MAP anormaux. Toutefois, le développement de vaccins anticancéreux efficaces est entravé par le manque de MAP exploitables, ou antigènes tumoraux (TA), exprimés exclusivement sur les cellules tumorales. La recherche et l’identification de TA ont été largement limitées aux MAP dérivés de mutations non synonymes situées dans des exons canoniques codant pour des protéines. Ces régions génomiques ne représentent que 2% du génome humain. Le fait que les MAP puissent potentiellement dériver de la traduction non canonique de toutes les régions génomiques n’a été pleinement compris que récemment. Ici, nous avons utilisé la protéogénomique pour découvrir des TA exploitables dérivés de produits de traduction canoniques et non canoniques partagés au sein ou entre divers types de cancers humains. Premièrement, nous avons utilisé des cellules souches pluripotentes induites (iPSC) pour identifier les MAP associés à la pluripotence (paMAP) étant partagés par les cellules cancéreuses. Les antigènes pluripotents sont exprimés dans les tissus embryonnaires et absents des tissus adultes sains, mais anormalement réexprimés par les cellules cancéreuses. Ainsi, bien qu'ils ne soient pas mutés, les paMAP constituent des cibles idéales et spécifiques au cancer. Nous avons identifié un ensemble de 48 paMAP dérivés de transcrits codants et non codants (48 %) impliqués dans le maintien de la pluripotence et exprimés de manière aberrante dans plusieurs types de cancer. Ainsi, bien qu’elles proviennent de différents types de cellules et de tissus, des tumeurs 4 distinctes convergent vers un programme transcriptionnel associé à la pluripotence. En effet, l’expression des paMAP dans les cancers est corrélée à l’hypométhylation récurrente de leurs gènes sources, la présence d’aberrations génomiques courantes et l’adoption par les tumeurs de stratégies d’évasion immunitaire communes. Enfin, comme plusieurs paMAP sont immunogènes, leur utilisation comme cibles dans des vaccins anticancéreux pourrait entrer en synergie avec les inhibiteurs disponibles des voies d'évasion immunitaire et améliorer le traitement de plusieurs cancers agressifs. Ensuite, nous avons évalué l’ensemble des TA ayant un potentiel thérapeutique dans deux types de tumeurs présentant une charge mutationnelle particulièrement élevée, le mélanome et le cancer du poumon non à petites cellules (NSCLC). Nous avons constaté que les TA mutés (mTSAs) représentent une minorité (1 %) des TA exploitables dans ces deux types de cancer. Cela peut s'expliquer par une faible expression d'ARN de la plupart des mutations non synonymes ainsi que par leur localisation en dehors des régions génomiques les plus efficaces pour la génération de MAP. En revanche, 99 % des TA dérivent de séquences génomiques non mutées spécifiques au cancer (aeTSA), surexprimées dans le cancer (TAA) ou spécifiques à la lignée cellulaire d'origine (LSA, exprimés par les mélanocytes ou par les cellules épithéliales pulmonaires, pour le mélanome et le NSCLC, respectivement). Tout comme les paMAP, environ 50 % des aeTSA identifiés dans le mélanome et le NSCLC proviennent de séquences non canoniques et sont régulés de manière épigénétique. Alors que les mTSA sont exclusivement spécifiques à chaque patient patient, les aeTSA sont partagés entre les échantillons tumoraux. De plus, leur absence dans les tissus normaux, leur abondance et leur capacité à activer les lymphocytes T CD8 en font des cibles idéales pour traiter les mélanomes et les NSCLC. En conclusion, cette thèse fournit un aperçu de la biogenèse de différents types de TA dans diverses cohortes de patients et ouvre la voie au développement d’immunothérapies ciblées et efficaces contre une grande variété de cancers. / Nearly all nucleated cells express MHC I (HLA I in humans)-associated peptides (MAPs) which are sampled from the cellular proteome and transported to the cell surface for inspection by CD8 T cells. As such, the collection of cell-surface MAPs, or the immunopeptidome, informs CD8 T cells on the inner cell state. Cancer immunosurveillance relies on the capacity of CD8 T cells to recognize abnormal MAPs on tumor cells and eliminate them while sparing healthy cells. Hence, the existence of cancer indicates that tumor-specific CD8 T cells are underpowered, dysfunctional or inhibited from exerting their function. Anti-cancer vaccines can boost tumor killing by stimulating the recognition of abnormal MAPs. The development of effective anti-cancer vaccines is limited by the identification of actionable MAPs, or tumor antigens (TAs), expressed exclusively on tumor cells. The TA search space has been largely limited to MAPs derived from non-synonymous mutations in canonical protein-coding exons which represent a mere 2% of the human genome. That MAPs can derive from the non-canonical translation of potentially all genomic regions has only recently been fully appreciated. Herein, we used proteogenomics to discover actionable TAs derived from canonical and non-canonical translation products shared within or across different types of human cancer. First, we used induced pluripotent stem cells (iPSCs) to identify pluripotency-associated MAPs (paMAPs) shared by cancer cells. Pluripotency antigens are restricted to embryonic tissues and absent from healthy adult tissues but abnormally re-expressed by cancer cells, which makes them ideal tumor-specific targets despite being unmutated. We identified a set of 46 paMAPs derived from coding and allegedly non-coding (48%) transcripts involved in pluripotency maintenance and aberrantly expressed in multiple cancer types. Thus, despite originating from different cell types and tissues, distinct tumor types converged towards a pluripotency-associated transcriptional program. Indeed, the expression of paMAPs across cancers correlated with recurrent source gene hypomethylation, genomic aberrations, and immune evasion properties. Several paMAPs were immunogenic, thus their targeting could synergize with available inhibitors of immune evasion pathways to improve the outcome of multiple aggressive cancers. 7 Next, we evaluated the actionable TA landscape of two tumor types with particularly high mutational load, melanoma and non-small cell lung cancer (NSCLC). We found that mutated TAs (mTSAs) represent a minority (1%) of actionable TAs in both cancer types, which can be explained by a low RNA expression of most non-synonymous mutations and their localization outside genomic regions proficient for MAP generation. By contrast, 99% of TAs derived from unmutated genomic sequences specific to cancer (aeTSAs), overexpressed in cancer (TAAs), or specific to the cell lineage of origin (LSAs, expressed by melanocytes or by lung epithelial cells, for melanoma and NSCLC LSAs, respectively). As for paMAPs, around 50% of aeTSAs in melanoma and NSCLC were non-canonical and were epigenetically regulated. Whereas mTSAs were exclusively patient-specific, aeTSAs were shared among tumor samples and exhibited all characteristics of targetable TAs, including tumor-specificity, high abundance, and immunogenicity. Altogether, this thesis provides insights into the biogenesis of different TA types in various patient cohorts and paves the way for the development of effective TA-based immunotherapies against a large variety of cancers.