Le glioblastome multiforme (GBM) est la tumeur cérébrale la plus commune et létale chez l’adulte. Malgré les avancés fulgurantes dans la dernière décennie au niveau des thérapies contre le cancer, le pronostique reste inchangé. Le manque de spécificité des traitements est la cause première de la récurrence de cette tumeur. Une meilleure compréhension au niveau des mécanismes moléculaires et biologiques de cette tumeur est impérative. La découverte des cellules souches cancéreuses (CD133+) au niveau du GBM offre une nouvelle opportunité thérapeutique contre cette tumeur. Effectivement, les cellules CD133+ seraient responsables de l’établissement, le maintien et la progression du GBM. De plus, elles sont également la cause de la résistance du GBM faces aux traitements de radiothérapies. Ces cellules représentent une cible de choix dans le but d’éradiquer le GBM. L’oncogène BMI1 a été associé à plusieurs types de tumeurs et est également essentielle au maintien de différentes populations de cellules souches normales et cancéreuses. Une forte expression de BMI1 est observée au niveau du GBM et plus précisément, un enrichissement préférentiel de cette protéine est noté au niveau des cellules CD133+. L’objectif principal de cette thèse est d’évaluer le rôle potentiel de BMI1 dans le maintien et la radiorésistance des cellules souches cancéreuses (CSC), CD133+ du GBM. La fonction principale de BMI1 est la régulation négative du locus INK4A/ARF. Ce locus est impliqué dans l’activation de deux voies majeurs anti-tumorales : P53 et RB. Or, la perte de BMI1 induit in vitro une diminution des capacités prolifératives, une augmentation de la différentiation et de l’apoptose, ainsi qu’une augmentation de la radiosensibilité des CSC du GBM indépendamment de la présence du locus INK4A/ARF. Effectivement, deux tumeurs sur trois possèdent une délétion de ce locus, ce qui suggère que BMI1 possède d’autre(s) cible(s) transcriptionnelle(s). Parmi ces nouvelles cibles ont retrouve la protéine P21, un régulateur négatif du cycle cellulaire. De plus, la perte de BMI1 inhibe l’établissement d’une tumeur cérébrale lors d’études de xénogreffe chez la souris NOD/SCID. Également, une nouvelle fonction de BMI1 indépendante de son activité transcriptionnel a été démontrée. Effectivement, suite à l’induction d’un bris double brin (BDB) de l’ADN, BMI1 est rapidement recruté au niveau de la lésion et influence le recrutement des protéines de reconnaissance du dommage à l’ADN. La perte de BMI1 mène à un défaut au niveau de la reconnaissance et la réparation de l’ADN, alors que sa surexpression induit plutôt une augmentation de ces mécanismes et procure une radiorésistance. Ces résultats décrivent pour la première fois l’importance de BMI1 au niveau du maintien, de l’auto-renouvellement et la radiorésistance des CSC du GBM. Ainsi, ces travaux démontrent que la protéine BMI1 représente une cible thérapeutique de choix dans le but d’éradiquer le GBM, une tumeur cérébrale létale. / Glioblastoma multiform (GBM) is the most common and lethal primary brain tumor found in adults. Despite the advances made in the field of cancer therapy in the last decade, the median survival rate remains less than a year. Therefore, a better understanding of the molecular biology of GBM will reveal the mechanisms responsible for the initiation and progression of the tumor, and allow the development of new therapeutic strategies. GBM contains a minority cell population, characterized by tumor initiating cells expressing the stem cell marker, CD133. The CD133+ GBM cells are responsible for tumor initiation, maintenance, progression and resistance to chemo/radiotherapy. The CD133+ cells represent a valuable and specific therapeutic target against GBM. The Polycomb (PcG) group family of transcriptional repressors have been involved in a vast range of cancers. The PcG protein and oncogene BMI1 is the best-characterized PcG protein. The implication of BMI1 in normal and cancer stem cell survival, self-renewal and maintenance has been thoroughly investigated. BMI1 is highly expressed in GBM and more precisely; it is enriched specifically in CD133+ cell populations. The main goal of this thesis was to elucidate the potential role of BMI1 in GBM CD133 + cancer stem cell (CSC) maintenance and radioresistance. The main function of BMI1 is to repress the expression of the genes encoded by the INK4A/ARF locus, which is implicated in the activation of two major tumor suppressor pathways, P53 and RB. However, BMI1 depletion in vitro induces a reduction in proliferation potential, as well as an increase in differentiation, apoptosis, and radiosensitivity regardless of INK4A/ARF status. Indeed, two-thirds of all tumors posses a deletion of this locus, suggesting that BMI1 regulates other targets. P21, a cell cycle regulator, was identified as a new BMI1 target. Moreover, we have observed that the loss of BMI1 inhibits the establishment of a cerebral tumor in a xenograft mouse model. In addition to transcription related activity, we identified a new transcription independent function of BMI1. After the induction of a DNA double-strand-break, BMI1 is rapidly recruited to the damage site and influences the recruitment of DNA damage response proteins. Furthermore, defects in DNA damage recognition and repair are observed after BMI1 knockdown. Consistent with these results, BMI1 overexpression induces DNA damage response and increases radioresistance potential. These results emphasize for the first time the requirement of BMI1 for the maintenance, self-renewal, and radioresistance in GBM CSC, thus providing a potential target for future therapeutic strategies against GBM.

Le glioblastome multiforme (GBM) est la tumeur cérébrale la plus commune et létale chez l’adulte. Malgré les avancés fulgurantes dans la dernière décennie au niveau des thérapies contre le cancer, le pronostique reste inchangé. Le manque de spécificité des traitements est la cause première de la récurrence de cette tumeur. Une meilleure compréhension au niveau des mécanismes moléculaires et biologiques de cette tumeur est impérative. La découverte des cellules souches cancéreuses (CD133+) au niveau du GBM offre une nouvelle opportunité thérapeutique contre cette tumeur. Effectivement, les cellules CD133+ seraient responsables de l’établissement, le maintien et la progression du GBM. De plus, elles sont également la cause de la résistance du GBM faces aux traitements de radiothérapies. Ces cellules représentent une cible de choix dans le but d’éradiquer le GBM. L’oncogène BMI1 a été associé à plusieurs types de tumeurs et est également essentielle au maintien de différentes populations de cellules souches normales et cancéreuses. Une forte expression de BMI1 est observée au niveau du GBM et plus précisément, un enrichissement préférentiel de cette protéine est noté au niveau des cellules CD133+. L’objectif principal de cette thèse est d’évaluer le rôle potentiel de BMI1 dans le maintien et la radiorésistance des cellules souches cancéreuses (CSC), CD133+ du GBM. La fonction principale de BMI1 est la régulation négative du locus INK4A/ARF. Ce locus est impliqué dans l’activation de deux voies majeurs anti-tumorales : P53 et RB. Or, la perte de BMI1 induit in vitro une diminution des capacités prolifératives, une augmentation de la différentiation et de l’apoptose, ainsi qu’une augmentation de la radiosensibilité des CSC du GBM indépendamment de la présence du locus INK4A/ARF. Effectivement, deux tumeurs sur trois possèdent une délétion de ce locus, ce qui suggère que BMI1 possède d’autre(s) cible(s) transcriptionnelle(s). Parmi ces nouvelles cibles ont retrouve la protéine P21, un régulateur négatif du cycle cellulaire. De plus, la perte de BMI1 inhibe l’établissement d’une tumeur cérébrale lors d’études de xénogreffe chez la souris NOD/SCID. Également, une nouvelle fonction de BMI1 indépendante de son activité transcriptionnel a été démontrée. Effectivement, suite à l’induction d’un bris double brin (BDB) de l’ADN, BMI1 est rapidement recruté au niveau de la lésion et influence le recrutement des protéines de reconnaissance du dommage à l’ADN. La perte de BMI1 mène à un défaut au niveau de la reconnaissance et la réparation de l’ADN, alors que sa surexpression induit plutôt une augmentation de ces mécanismes et procure une radiorésistance. Ces résultats décrivent pour la première fois l’importance de BMI1 au niveau du maintien, de l’auto-renouvellement et la radiorésistance des CSC du GBM. Ainsi, ces travaux démontrent que la protéine BMI1 représente une cible thérapeutique de choix dans le but d’éradiquer le GBM, une tumeur cérébrale létale. / Glioblastoma multiform (GBM) is the most common and lethal primary brain tumor found in adults. Despite the advances made in the field of cancer therapy in the last decade, the median survival rate remains less than a year. Therefore, a better understanding of the molecular biology of GBM will reveal the mechanisms responsible for the initiation and progression of the tumor, and allow the development of new therapeutic strategies. GBM contains a minority cell population, characterized by tumor initiating cells expressing the stem cell marker, CD133. The CD133+ GBM cells are responsible for tumor initiation, maintenance, progression and resistance to chemo/radiotherapy. The CD133+ cells represent a valuable and specific therapeutic target against GBM. The Polycomb (PcG) group family of transcriptional repressors have been involved in a vast range of cancers. The PcG protein and oncogene BMI1 is the best-characterized PcG protein. The implication of BMI1 in normal and cancer stem cell survival, self-renewal and maintenance has been thoroughly investigated. BMI1 is highly expressed in GBM and more precisely; it is enriched specifically in CD133+ cell populations. The main goal of this thesis was to elucidate the potential role of BMI1 in GBM CD133 + cancer stem cell (CSC) maintenance and radioresistance. The main function of BMI1 is to repress the expression of the genes encoded by the INK4A/ARF locus, which is implicated in the activation of two major tumor suppressor pathways, P53 and RB. However, BMI1 depletion in vitro induces a reduction in proliferation potential, as well as an increase in differentiation, apoptosis, and radiosensitivity regardless of INK4A/ARF status. Indeed, two-thirds of all tumors posses a deletion of this locus, suggesting that BMI1 regulates other targets. P21, a cell cycle regulator, was identified as a new BMI1 target. Moreover, we have observed that the loss of BMI1 inhibits the establishment of a cerebral tumor in a xenograft mouse model. In addition to transcription related activity, we identified a new transcription independent function of BMI1. After the induction of a DNA double-strand-break, BMI1 is rapidly recruited to the damage site and influences the recruitment of DNA damage response proteins. Furthermore, defects in DNA damage recognition and repair are observed after BMI1 knockdown. Consistent with these results, BMI1 overexpression induces DNA damage response and increases radioresistance potential. These results emphasize for the first time the requirement of BMI1 for the maintenance, self-renewal, and radioresistance in GBM CSC, thus providing a potential target for future therapeutic strategies against GBM.

Insulin like Growth Factor Binding Proteins (IGFBPs) 1 to 6 have important physiological functions of regulating half life and bioavailability of Insulin like Growth Factors (IGFs). Consequently, these have been known to play important roles in embryonic development, postnatal growth and disease conditions like cancer. However, the physiological roles of IGFBPs are diverse and not restricted only to the IGF regulation. These molecules are found to be tumor suppressors or promoters depending on the physiological contexts. IGFBP-2 has been established as a tumor promoter and found to be unregulated in several cancers including breast, ovarian, prostate cancer and glioblastoma (GBM). Various in vitro and in vivo studies have convincingly demonstrated the role of IGFBP-2 in inducing tumor cell proliferation, migration, invasion and chemoresistance. Increased plasma and tissue levels of IGFBP-2 have been associated with poor clinical outcome with respect to patients’ response to the therapy, relapse and overall survival. Various studies so far have demonstrated the role of IGFBP-2 in promoting glioma cell proliferation, migration, invasion, chemoresistance and determining stamens of GICs (Glioma Initiating Cells). However, the exact mechanisms underlying these functions remain unknown. Apart from being a diagnostic and prognostic indicator, IGFBP-2 has also been proposed as a therapeutic target. Therefore it is essential to understand mechanistic insights into pro-tumorigenic functions of IGFBP-2. Apart from the conventional function of regulating IGFs, IGFBP-2 has been shown to have several IGF independent functions. In a previous study, we reported IGFBP-2 as an upstream regulator of β-catenin signaling pathway in breast cancer. Interestingly, this study linked the association of higher expression of IGFBP-2 and β-catenin with the lymph node metastasis status of breast cancer. β-catenin signaling has been considered as one of the most important pro-tumorigenic pathways in several cancers including glioblastoma. Considering the importance of IGFBP-2 and β-catenin signaling pathways in glioblastoma, it becomes important to evaluate regulation of β-catenin activity by IGFBP-2 in glioma and address its clinical relevance. With this aim, the objectives of this study are,  To study mechanism of IGFBP-2 mediated regulation of β-catenin signaling in glioma cells and prognostic significance of IGFBP-2 and β-catenin expression in GBM tissues.  Isolation of human single chain variable fragment (scFv) against IGFBP-2 and its characterization as an inhibitor for IGFBP-2 pro-tumorigenic functions. Towards this, we established stable IGFBP-2 knockdown U251 cell line and IGFBP-2 over expressing LN229 and U87 cell lines. IGFBP-2 modulation in these glioma cell lines did not alter the rate of proliferation but there was a significant effect on cellular migration and invasion. In case of U251 cell line, there was a significant decrease in the intracellular levels of β-catenin while in IGFBP-2 over expressing cell lines there was a marked increase in intracellular β-catenin suggesting that IGFBP-2 is involved in the regulation of β-catenin in these cells. It was observed that this regulation of β-catenin was not because of its transcriptional regulation or regulation of canonical Wnt ligands Wnt1, Wnt2 and Wnt3a. To further delineate the pathway and understand the mechanism behind regulation of β-catenin, upstream regulators of β-catenin were analyzed. GSK3β is an important negative regulator of β-catenin which primes it for ubiquitination and proteasomal degradation. Phosphorylation of GSK3β at Ser9 position renders this enzyme inactive. In our study, it was observed that there was a significant downregulation of p-GSK3β in U251 cells with IGFBP-2 knockdown and upregulation in IGFBP-2 over expressing cell lines. Overexpression of IGFBP-2 in LN229 and U87 cell lines resulted in considerable decrease in the GSK3β mediated phosphorylation of β-catenin. This study unequivocally established that regulation of β-catenin by IGFBP-2 is via inactivation of GSK3β. Furthermore, regulation of GSK3β was found to be due to action of FAK following binding of IGFBP-2 to integrins. The expression pattern of IGFBP-2 and β-catenin protein in the tumor tissues of 112 GBM patients was studied and its correlation with patient survival was analysed. In this analysis it was observed that co-expression of IGFBP-2 and β-catenin is a strong predictor of patient prognosis. These results further implied the importance of understanding IGFBP-2 and β-catenin association in GBM pathology. One of the interesting observations in our study is that, not only full length IGFBP-2 protein but also C-terminal domain of IGFBP-2 was sufficient to regulate β-catenin and other IGFBP-2 mediated functions. This strongly asserts the importance of C-terminal region of IGFBP-2 as a tumor promoter. Towards an attempt to develop an inhibitor for IGFBP-2 actions, we screened a human single chain variable fragment (scFv) library using phage display technique. From this screening, one scFv (B7J) was identified which was a binder of full length IGFBP-2 as well as C-terminal domain of IGFBP-2. This scFv showed inhibition of IGFBP-2-cell surface interaction and also efficiently inhibited IGFBP-2-induced signaling pathways like ERK, FAK and GSK3β/β-catenin. B7J treatment also neutralized regulation of IGFBP-2 transcriptional targets like MMP2 and CD24. Gelatin zymography indicated the ability of B7J to decrease matrix metalloprotease activity in the conditioned medium of glioma cells. These effects ultimately reflected on the IGFBP-2-induced cellular migratory and invasive behaviour which was largely abrogated by B7J scFv treatment. Considering the therapeutic importance of scFvs because of their small size, better tumor penetration and tumor retention capacity than full length antibody molecules, such kind of strategy could be of great importance in the management of GBM. Altogether, this study provides a mechanistic insight of IGFBP-2 mediated actions involving integrin/FAK/GSK3β/β-catenin pathways and the possible role of this crosstalk in the aggressiveness of glioblastoma. This study also provides a proof of principle that an inhibitor like anti IGFBP-2 scFv could be of importance for controlling invasive glioblastoma.

Artificial neural networks are fascinating machine learning algorithms. They used to be considered unreliable and computationally very expensive. Now it is known that modern neural networks can be quite useful, but their computational expensiveness unfortunately remains. Statistical boosting is considered to be one of the most important machine learning ideas. It is based on an ensemble of weak models that together create a powerful learning system. The goal of this thesis is the comparison of these machine learning models on three use cases. The first use case deals with modeling the probability of burglary in the city of Chicago. The second use case is the typical example of customer churn prediction in telecommunication industry and the last use case is related to the problematic of the computer vision. The second goal of this thesis is to introduce an open-source machine learning platform called H2O. It includes, among other things, an interface for R and it is designed to run in standalone mode or on Hadoop. The thesis also includes the introduction into an open-source software library Apache Hadoop that allows for distributed processing of big data. Concretely into its open-source distribution Hortonworks Data Platform.

Glioblastoma multiforme (GBM) is the most common and aggressive primary brain tumor. The current standard-of-care treatment including surgery, radiation and temozolomide (TMZ) chemotherapy does not prolong the survival satisfactorily. Here we have tested the feasibility, efficacy and safety of a potential gene therapy approach using AAV as gene delivery vehicle for treatment of GBM. Interferon-beta (IFNβ) is a cytokine molecule also having pleiotropic anticancerous properties. Previously it has been shown by our group that AAV mediated local (intracranial) gene delivery of human IFNβ (hIFNβ) could be an effective treatment for non-invasive human glioblastoma (U87) in orthotopic xenograft mouse model.But as one of the major challenges to treat GBM effectively in clinics is its highly invasive property, in the current study we first sought to test the efficacy of our therapeutic model in a highly invasive human GBM (GBM8) xenograft mouse model. One major limitation of using the xenograft mouse model is that these mice are immune-compromised. Moreover, as IFNβ does not interact with cross-species receptors, the influence of immune systems on GBM remains largely untested. Therefore to test the therapeutic approach in an immune-competent mouse model, we next treated a syngeneic mouse GBM model (GL261) in an immune-competent mouse (C57B6) with the gene encoding the species-matched IFNβ (mIFNβ). We also tested if combination of this IFNβ gene therapy with the current standard chemotherapeutic drug (TMZ) is more effective than any one of the therapeutic modes alone. Finally, we tested the long term safety of the AAV-mIFNβ local gene therapy in healthy C57B6 mice. Next, we hypothesized that global genetic engineering of brain cells expressing secretory therapeutic protein like hIFNβ could be more beneficial for treatment of invasive, migratory and distal multifocal GBM. We tested this hypothesis using systemic delivery of AAV9 vectors encoding hIFNβ gene for treatment of GBM8 tumor in nude mice. Using in vivo bioluminescence imaging of tumor associated firefly luciferase activity, long term survival assay and histological analysis of the brains we have shown that local treatment of AAV-hIFNβ for highly invasive human GBM8 is therapeutically beneficial at an early growth phase of tumor. However, systemic delivery route treatment is far superior for treating multifocal distal GBM8 tumors. Nonetheless, for both delivery routes, treatment efficacy is significantly reduced when treated at a later growth phase of the tumor. In syngeneic GL261 tumor model study, we show that local AAV-mIFNβ gene therapy alone or in combination with TMZ treatment can provide significant survival benefit over control or only TMZ treatment, respectively. However, the animals eventually succumb to the tumor. Safety study in the healthy animals shows significant body weight loss in some treatment groups, whereas one group shows long term survival without any weight loss or any noticeable changes in the external appearances. However, histological analysis indicates marked demyelinating neurotoxic effects upon long term exposures to mIFNβ over-expressions in brain. Overall, we conclude from this study that AAV-IFNβ gene therapy has great therapeutic potential for GBM treatment in future, but the therapeutic window is small and long term continuous expression could have severe deleterious effects on health.

No description available.

No description available.

Glioblastoma (GBM) is the most common high-grade brain tumor diagnosed in patients who are more than 50 years of age. The standard of care treatment is surgery, followed by radiotherapy and chemotherapy. The median life expectancy of patients is only between 12 to 15 months after receiving current treatment regimes. Hence, identification of new therapeutic compounds and gene targets are highly warranted. This thesis describes four interlinked studies to attain this goal. In study 1, we explored drug combination effects in a material of 41 patient-derived GBM cell (GC) cultures. Synergies between three compounds, pterostilbene, gefitinib, and sertraline, resulted in effective killing of GC and can be predicted by biomarkers. In study 2, we performed a large-scale screening of FDA approved compounds (n=1544) in a larger panel of GCs (n=106). By combining the large-scale drug response data with GCs genomics data, we built a novel computational model to predict the sensitivity of each compound for a given GC. A notable finding was that GCs respond very differently to proteasome inhibitors in both in-vitro and in-vivo. In study 3, we explored new gene targets by RNAi (n=1112) in a panel of GC cells. We found that loss of transcription factor ZBTB16/PLZF inhibits GC cell viability, proliferation, migration, and invasion. These effects were due to downregulation of c-MYC and Cyclin B1 after the treatment. In study 4, we tested the genomic stability of three GCs upon multiple passaging. Using molecular and mathematical analyses, we showed that the GCs undergo both systematic adaptations and sequential clonal takeovers. Such changes tend to affect a broad spectrum of pathways. Therefore, a systematic analysis of cell culture stability will be essential to make use of primary cells for translational oncology. Taken together, these studies deepen our knowledge of the weak points of GBM and provide several targets and biomarkers for further investigation. The work in this thesis can potentially facilitate the development of targeted therapies and result in more accurate tools for patient diagnostics and stratification.

No description available.