Les rétrotransposons L1 sont les seuls éléments transposables autonomes et actifs chez l'Homme et constituent 20% de notre ADN. Ils prolifèrent via un intermédiaire ARN et un processus couplé de réverse transcription et d'intégration, appelé rétrotransposition, et médié par une particule ribonucléoprotéique (RNP). Les L1s sautent de façon active dans les cellules germinales, les cellules souches embryonnaires et l'embryon précoce, ce qui provoque parfois de nouvelles maladies génétiques. Cependant ils sont considérés comme éteints dans la plupart des tissus somatiques. Dans le but d'explorer l'importance et les conséquences de la rétrotransposition des L1s chez l'Homme, nous avons développé une approche de cartographie des L1s actifs dans le génome humain, en combinant amplification sélective des sites d'insertion et séquençage à haut-débit. Nous avons utilisé cette stratégie afin d'obtenir la cartographie différentielle des L1s dans deux lignées cellulaires humaines apparentées. Ainsi, nous avons découvert plusieurs insertions de L1 présentes uniquement dans la lignée fille mais absente dans la lignée parentale, démontrant pour la première fois que les éléments L1 endogènes humains sont capables de mobilité dans des lignées de cellules somatiques en culture. D'autre part, afin d'éclaircir les déterminants qui dictent l'intégration des L1s, nous avons développé un test direct de réverse transcription in vitro à partir de RNP L1 natives partiellement purifiées de cellules humaines. Ceci nous a permis de montrer que la réverse transcriptase du L1 participe à la sélection du site d'insertion, ajoutant une couche additionnelle de spécificité après l'endonucléase L1. En conclusion, notre travail met en lumière la flexibilité de la machinerie des L1s, une propriété qui a certainement participé à l'efficacité de l'invasion des génomes de mammifères par ces éléments génétiques mobiles. / L1 retrotransposons are the only autonomous and active transposable elements in humans and comprise as much as 20% of our DNA. They proliferate via an RNA intermediate and a coupled reverse transcription and integration process, called retrotransposition and mediated by an L1-encoded ribonucleoprotein particle (RNP). L1s are actively jumping in germ cells, embryonic stem cells and in the early embryo, occasionally leading to de novo genetic diseases, but are considered silent in most somatic tissues. To comprehensively map active L1 elements in the human genome and to further explore the importance and consequences of L1 retrotransposition in humans, we combined selective amplification of L1 insertion sites and high-throughput sequencing. We applied this strategy to obtain a differential map of L1 insertions in two related human cultured cell lines and to question the possibility that endogenous L1 elements could be jumping in somatic cultured cells. We discovered several L1 insertions only present in the daughter cell line but absent in the parental cell line, demonstrating for the first time that retrotransposition of endogenous L1s takes place in a human somatic cell line. To get insights into the determinants of L1 integration, we have also developed a novel reverse transcription assay using partially purified native L1 RNPs. This enabled us to show that the L1 reverse transcriptase participates to insertion site selection, adding a second layer of specificity beyond the L1 endonuclease. Finally our work highlights the flexibility of the L1 machinery, which certainly participates to the efficient spreading of L1 elements within mammalian genomes.

La dernière décennie a connu une révolution dans le traitement du cancer en s'éloignant des médicaments conventionnels qui ciblent directement la tumeur (comme les chimiothérapies et les thérapies moléculaires ciblées) au profit des immunothérapies, et en particulier les inhibiteurs des "checkpoint" immunitaires. Ces immunothérapies libèrent sélectivement le système immunitaire de l'hôte contre la tumeur et ont démontré une rémission durable sans précédent chez des patients atteints de cancers que l'on croyait incurables, comme le mélanome métastatique, le carcinome rénal métastatique et les stades avancés du cancer du poumon non à petites cellules. Cependant, plus de la moitié des patients ne répondent pas à ces traitements, et sont donc contraints à recevoir d’autres traitements potentiellement toxiques et coûteux. Face aux résultats prometteurs des immunothérapies anti‐PD1/PD‐L1, la recherche de nouvelles cibles/marqueurs moléculaires permettant d'améliorer l'efficacité et de permettre un traitement personnalisé s'est intensifiée. Nos travaux portent sur l’étude de la protéine Sequestosome 1 (SQSTM1/p62) qui est un substrat de l'autophagie sélective et une plateforme de signalisation impliquée dans l’agressivité tumorale. Nous mettons en évidence l’implication de SQSTM1/p62 dans l'inflammation tumorale et dans l’évasion immunitaire et ceci grâce à la stabilisation du messager de PD‐L1. Nous observons que la surexpression tumorale de SQSTM1/p62 caractérise les tumeurs infiltrées et immunosuppressives qui répondent le mieux aux anti‐PD1/PD‐L1. Ainsi, nous proposons SQSTM1/p62 en tant que biomarqueur potentiel et cible thérapeutique pour améliorer la stratification des patients susceptibles de répondre aux immunothérapies. / The past decade has witnessed a new revolution in cancer treatment by shifting away from the conventional drugs that directly target the tumor (such as chemotherapies and molecular targeted therapies) towards immune‐based therapies, and in particular the so‐called immune checkpoint inhibitors. These immunotherapies selectively release the host immune system against the tumor and have shown unprecedented durable remission for patients with cancers that were thought incurable such as metastatic melanoma, metastatic renal cell carcinoma and late stages of non‐small cell lung cancer. However, more than half of the patients fail to respond to these treatments and are therefore forced to receive other potentially toxic and costly treatments. In this era of promising anti‐PD1/PD‐L1 immunotherapies, the quest for reliable molecular markers/therapeutic targets that would enhance the effectiveness and allow a personalized adaptation of the treatments has intensified. In our work we focus on the scaffold protein and autophagy adaptor Sequestosome 1 (SQSTM1/p62), and we show that it is required for tumor inflammation and immune evasion. We find that SQSTM1/p62 tumor overexpression characterizes infiltrated and immunosuppressive tumors and predicts for response to anti‐PD1/PD‐L1 therapies. Thus, we propose SQSTM1/p62 as a potential biomarker and a therapeutic target to improve the stratification of patients to immunotherapies.

This thesis is focused on the L1 regression, a possible alternative to the ordinary least squares regression. L1 regression replaces the least squares estimation with the least absolute deviations estimation, thus generalizing the sample median in the linear regres- sion model. Unlike the ordinary least squares regression, L1 regression enables loosening of certain assumptions and leads to more robust estimates. Fundamental theoretical re- sults, including the asymptotic distribution of regression coefficient estimates, hypothesis testing, confidence intervals and confidence regions, are derived. This method is then compared to the ordinary least squares regression in a simulation study, with a focus on heavy-tailed distributions and the possible presence of outlying observations. 1

Although PID controllers work well on Miniature Air Vehicles (MAVs), they require tuning for each MAV. Also, they quickly lose performance in the presence of actuator failures or changes in the MAV dynamics. Adaptive control algorithms that self tune to each MAV and compensate for changes in the MAV during flight are explored. However, because the autopilots on MAVs are small, many of the adaptive control algorithms like those that employ least squares estimation may take too much code space, memory, and/or computing power. In this thesis we develop several Lyapunov-based model reference adaptive control (MRAC) schemes that are both simple and efficient with the MAV autopilot resources. Most notable are the L1 controllers that have all the benefits of traditional MRACs but have reduced high frequency content to the actuators. The schemes control both roll and pitch through aileron and elevator commands. Flight test results for the schemes are also compared.

No description available.

Many studies have shown that peroxisome-proliferator-activated receptor [gamma] (PPAR[gamma]) plays an important role in adipose tissue formation by activating genes implicated in adipogenesis. PPAR[gamma] heterodimerizes with retinoid X receptor [alpha] (RXR[alpha]), in the presence of ligand, on PPAR response elements (PPREs) in the promoter of target genes involved in adipocyte differentiation. General receptor for phosphoinositides 1 (GRP1) is a corepressor of thyroid hormone receptors (TRs), a nuclear receptor like PPAR[gamma]. GRP1 decreases TRs' transcriptional activity by lowering dimerisation on DNA. Since PPARs and TRs have important structural similarities and that GRP1 interacts with PPARs in vitro, we hypothesized that GRP1 could be a coregulator of PPARs and, thus be implicated in adipogenesis. To better understand GRP1's effect on PPAR[gamma]2, transcriptional activity assays have been done and show that increasing concentrations of GRP1 decrease the transcriptional activity of PPAR[gamma]2. We also studied GRP1 expression by Western blots of total protein extracts from 3T3-L1 cells at different times during differentiation: GRP1 is present in 3T3-L1 preadipocytes and its expression decreases during adipogenesis. According to those results, GRP1 may be a PPAR[gamma] corepressor. After those observations, GRP1 effects on adipogenesis were studied by modulating its expression with lentiviral particles. Interestingly, GRP1 knock-down before inducing 3T3-L1 differentiation, almost abrogates adipogenesis and adipocytes markers, PPAR[gamma] and aP2, while its overexpression increases lipid storage without affecting PPAR[gamma] expression. On the opposite, GRP1 modulation after differentiation induction shows that expression knock-down slightly promotes adipogenesis by increasing PPAR[gamma], aP2 and lipid accumulation and that overexpression weakly decreases lipid storage. Our results suggest that GRP1 implication during adipogenesis occurs at two distinct and precise moments. It seems to be a key factor in the early stages of adipocyte differentiation and to be implicated as a PPAR[gamma] transcriptional activity modulator as a corepressor. Future experiments will help detail modulation of protein expression and underlying mechanisms to better understand the role of GRP1 in adipogenesis and, eventually, comorbidities linked to obesity like cardiovascular diseases and type 2 diabetes mellitus.

Although there have been a large number of studies on the use of L1 and L2, there seem to be few on L1 use in Chinese university EFL classrooms, especially investigating the language use of those who teach English to students at different proficiency levels or teach different types of English courses. This thesis aims to analyze four Chinese EFL teachers’ actual use of L1 and L2, to understand their attitudes and beliefs regarding this issue, and their own perceptions of and reasons for their language use, and to explore possible influencing factors. The reading-and-writing lessons and the listening-and-speaking lessons of these four teachers, who were teaching non-English major students at four different levels, were observed and recorded. All the observed lessons were subjected to quantitative analysis with the aim of providing a clear picture of the distribution of their L1 and L2 use. Some episodes selected from these lessons were subjected to further detailed analysis, in order to provide an account of the circumstances, functions, and grammatical patterns of their language use, as well as their language use across different frames of classroom discourse. The teachers were interviewed subsequently about their general beliefs on the use of L1 in L2 teaching and learning. Separately, in a stimulated recall interview, they were invited to provide comments specifically on their language use in the selected episodes that were replayed to them. The quantitative findings show that the amount of the teachers’ L1 use was not necessarily closely related to their students’ English proficiency levels, although the teacher of the students at the lowest level used the highest amount of Chinese in her lessons. However, a noteworthy finding was that all four teachers used more Chinese in the reading-and-writing lessons than in the listening-and-speaking lessons, although with substantial individual variation. The qualitative analysis of classroom data indicates that these teachers switched often at unit boundaries, but rarely at clause boundaries. They also switched frequently within units, especially within noun phrases, and the ‘Chinese determiner + English noun’ pattern is the main one they had in common. Furthermore, the teachers used Chinese as the matrix language in their mixed utterances in most cases, and these mixed utterances nearly always fitted Myers-Scotton’s Morpheme Order principle and System Morpheme principle. The teachers were also found to use Chinese in a variety of circumstances, such as talking about lesson plans or examinations, dealing with exercises, analyzing text, teaching vocabulary, checking the students’ comprehension or retention, giving the students advice on learning, telling anecdotes and assigning homework. The functions for which they used Chinese could be divided into four main categories: facilitating developing lesson content; supporting students and carrying out classroom management; delivering information related to teaching agenda or examinations; and facilitating communication beyond language learning and teaching. The most frequent function common to all four teachers was translation. Furthermore, the study used four different ‘frames’ to analyze classroom discourse, and found that the teachers used the L1 with varying frequency across these frames. Moreover, although all four teachers believed that using the L1 was beneficial to L2 learning, their attitudes towards the medium of instruction were different. While two advocated using the L1, the other two expressed a preference for speaking English-only and perceived their L1 use as a compromise or an expedient. The teachers reported many reasons for their L1 use. The factors that affected their language use consisted of both immediate classroom factors, such as functions of utterances, students’ language use, students’ perceived mood, students’ background knowledge, the difficulty of lesson content, time limitations, teachers’ awareness of their own L1 use, and teachers’ state of mind at a particular moment in a lesson, and relatively static factors, such as the university policy, students’ L2 abilities, teaching objectives, teachers’ beliefs regarding L1 use, and teachers’ L2 abilities. Through its detailed analysis of the teachers’ language use, as well as their relevant beliefs and decision-making, this thesis hopes to make a contribution to L2 teachers’ professional development and L2 teaching, especially in helping to establish a pedagogically principled approach to L1 and L2 use.

Mémoire numérisé par la Direction des bibliothèques de l'Université de Montréal.

Mémoire numérisé par la Direction des bibliothèques de l'Université de Montréal.

Background: Feature selection, also known as variable selection, is a technique that selects a subset from a large collection of possible predictors to improve the prediction accuracy in regression model. First objective of this project is to investigate in what data structure LASSO outperforms forward stepwise method. The second objective is to develop a feature selection method, Feature Selection by L1 Regularization of Subset of Parameters (LRSP), which selects the model by combining prior knowledge of inclusion of some covariates, if any, and the information collected from the data. Mathematically, LRSP minimizes the residual sum of squares subject to the sum of the absolute value of a subset of the coefficients being less than a constant. In this project, LRSP is compared with LASSO, Forward Selection, and Ordinary Least Squares to investigate their relative performance for different data structures. Results: simulation results indicate that for moderate number of small sized effects, forward selection outperforms LASSO in both prediction accuracy and the performance of variable selection when the variance of model error term is smaller, regardless of the correlations among the covariates; forward selection also works better in the performance of variable selection when the variance of error term is larger, but the correlations among the covariates are smaller. LRSP was shown to be an efficient method to deal with the problems when prior knowledge of inclusion of covariates is available, and it can also be applied to problems with nuisance parameters, such as linear discriminant analysis.