No description available.

Les technologies récentes de séquençage de l’ARN de cellules uniques (scRNA-seq, pour single cell RNA-seq) ont permis de quantifier le niveau d’expression des gènes au niveau de la cellules, alors que les technologies standards de séquençage de l’ARN (RNA-seq, ou bulk RNA-seq) ne permettaient de quantifier que l’expression moyenne des gènes dans un échantillon de cellules. Cette résolution supérieure a permis des avancées majeures dans le domaine de la recherche biomédicale, mais a également posé de nouveaux défis, notamment computationnels. Les données qui découlent des technologies de scRNA-seq sont en effet complexes et plus bruitées que les données de bulk RNA-seq. En outre, les technologies sont nombreuses et leur nombre explose, nécessitant chacune un prétraitement plus ou moins différent. De plus en plus de méthodes sont ainsi développées, mais il n’existe pas encore de norme établie (gold standard) pour le prétraitement et l’analyse de ces données. Le laboratoire du Dr. Mader a récemment fait l’acquisition de la technologie Drop-Seq (une technologie haut débit de scRNA-seq), nécessitant une expertise nouvelle pour le traitement des données qui en découlent. Dans ce mémoire, différentes étapes du prétraitement des données issues de la technologie Drop-Seq sont donc passées en revue et le fonctionnement de certains outils dédiés à cet effet est étudié, permettant d’établir des lignes directrices pour de futures expériences au sein du laboratoire du Dr. Mader. Cette étude est menée sur les premiers jeux de données générés avec la technologie Drop-Seq du laboratoire, issus de lignées cellulaires du cancer du sein. Les méthodes d’analyses, moins spécifiques à la technologie, ne sont pas étudiées dans ce mémoire, mais une analyse exploratoire des jeux de données du laboratoire pose les bases pour une analyse plus poussée. / Recent single cell RNA sequencing technologies (scRNA-seq) have enabled the quantification of gene expression levels at the cellular level, while standard RNA sequencing technologies (RNA-seq, or bulk RNA-seq) have only been able to quantify the average gene expression in a sample of cells. This higher resolution has allowed major advances in biomedical research, but has also raised new challenges, in particular computational ones. The data derived from scRNA-seq technologies are indeed complex and noisier than bulk RNA-seq data. In addition, the number of scRNA-seq technologies is exploding, each of them requiring a rather different pre-processing. More and more methods are thus being developed, but there is still no gold standard for the preprocessing and analysis of these data. Dr. Mader’s laboratory has recently invested in the Drop-Seq technology (a high-throughput scRNAseq technology), requiring new expertise for the processing of the resulting data. In this thesis, different steps for the pre-processing of Drop-Seq data are reviewed and the behavior of some of the dedicated tools are studied, allowing to establish guidelines for future experiments in Dr. Mader’s laboratory. This study is conducted on the first data sets generated with the Drop-Seq technology of the laboratory, derived from breast cancer cell lines. Analytical methods, less specific to the technology, are not investigated in this thesis, but an exploratory analysis of the lab’s datasets lays the foundation for further analysis.

Adult hippocampal neurogenesis is orchestrated by neural stem cell (NSC) activity. Some associations exist between autophagy and neurogenesis, yet much remains unknown about autophagic regulation of adult neurogenesis. This thesis interrogates the requirement and role of Beclin1 and Atg5, two regulators of autophagy, in the formation of adult hippocampal neurons. To examine adult brain NSCs, the experiments presented in the first objective of this thesis test the ability to isolate adult NSCs using flow cytometry and a DNA-binding dye, DyeCycleViolet. While adult NSCs could not be isolated from the adult neurogenic niches using this methodology, it was effective in isolating endothelial cells. This provided valuable insight on the use of DNA-binding dyes and a new method for isolation of brain endothelial cells. The next objective determines the role of Beclin1 in adult NSCs and their progeny using an inducible model. Beclin1 loss in Nestin-expressing hippocampal NSCs resulted in reduced proliferation, autophagy, and adult neurogenesis within one month. Single-cell RNA sequencing and other methods illuminated that loss of Beclin1 resulted in mitosis reduction, disrupted mitotic regulation of chromatin maintenance, and induction of DNA damage. The final objective first tests whether Beclin1 loss results in similar deficits within GLAST-expressing NSCs and progeny. This model mirrored neurogenesis deficits and requirement of Beclin1 in mitosis and DNA maintenance. Next, to test whether this phenotype occurs with other autophagy proteins, Atg5 was removed from GLAST NSCs. This resulted in reduced autophagy and a transient decrease in neurons in the absence of any effect on NSC proliferation. Thus, proliferation deficits are unique to Beclin1 loss and do not underlie reduced adult hippocampal neurogenesis after Atg5 removal. This work demonstrates a novel discovery of mitosis regulation in adult NSCs by Beclin1, and individual roles of Beclin1 and Atg5 in neurogenesis.

Durant le développement préimplantatoire, les défauts de ségrégation des chromosomes conduisent à l'héritage d'un nombre incorrect de chromosomes, connu sous le nom d'aneuploïdie, qui provoque l'infertilité. L’imagerie à intervalle du développement préimplantatoire est introduite pour sélectionner le meilleur embryon et des efforts sont en cours pour utiliser l'imagerie non invasive pour identifier les ovocytes euploïdes en métaphase-II comme prédicteur de la viabilité future de l'embryon. Il est déjà bien établi que les ovocytes de mammifères en métaphase-II subissent des mouvements cytoplasmiques stéréotypés qui peuvent être visualisés par imagerie non invasive à fond clair à intervalle, appelée « flux cytoplasmique ». Ici, nous avons émis l'hypothèse que le flux cytoplasmique pourrait être affecté par le statut de ploïdie de l'ovule et donc être un outil de sélection utile pour sélectionner les ovules euploïdes de manière non invasive. Nous avons développé des conditions pour générer des ovules euploïdes et aneuploïdes à partir du même bassin d'ovocytes sains. Nous avons ensuite utilisé la microscopie d'imagerie en temps réel DIC, permettant de visualiser et de mesurer le flux cytoplasmique sans manipulation de l'ovule. Les mouvements cytoplasmiques ont été liés au statut de ploïdie pour chaque ovule individuel par immunofluorescence. Nos résultats montrent qu'il n'y a pas de différence de flux cytoplasmique entre les ovules euploïdes et aneuploïdes. Nos données démontrent que l'état de la ploïdie n'a pas d'impact sur les mouvements cytoplasmiques, suggérant que l'utilisation d'une imagerie non invasive pour essayer de distinguer l'état de la ploïdie entre des ovocytes autrement sains sera difficile. / Chromosome segregation errors during early development lead to inheritance of incorrect number of chromosomes, known as aneuploidy, which causes infertility and birth defects. Time-lapse microscopy of preimplantation development is being widely introduced with the aim of selecting the best embryo and efforts to use non-invasive brightfield imaging to identify euploid oocytes at metaphase-II as a predictor of future embryo viability are underway. It is already well established that mammalian metaphase-II oocytes undergo stereotyped cytoplasmic movements that can be visualised by non-invasive brightfield timelapse imaging, termed “cytoplasmic flow”. Here, we hypothesised that this cytoplasmic flow might be affected by ploidy status of the egg and therefore be a useful selection tool to select euploid eggs non-invasively. To address this, we developed conditions to generate euploid and aneuploid eggs from the same pool of otherwise healthy oocytes. We then used DIC live-imaging microscopy, which allowed us to visualise and measure flow without any manipulation to the egg. Importantly, individual eggs were scored for their ploidy status by immunofluorescence, so that cytoplasmic movements could be related to ploidy on an egg-by-egg basis. Our results show that there is no difference in cytoplasmic flow between euploid and aneuploid eggs. Therefore, our data demonstrates that ploidy status does not impact biologically relevant stereotyped cytoplasmic movements, suggesting that using non-invasive imaging to try to distinguish ploidy status between otherwise healthy oocytes will be challenging.

Bronchopulmonary dysplasia (BPD) is a chronic lung disease and one of the most severe complications that develop in premature infants following mechanical ventilation, exposure to supplemental oxygen, and inflammation. The hallmarks of the lung pathology are arrested lung development, including fewer and larger alveoli with less septation, thickening of alveolar septa, and impaired development of the capillary network. BPD is associated with increased mortality, respiratory morbidity, neurodevelopmental impairment, and increased healthcare costs. Significant advancements in neonatology in the last several decades, including antenatal steroids and exogenous surfactant replacement therapy, more gentle ventilation methods, and judicious oxygen use, have allowed for the survival of more preterm infants. However, the incidence of BPD still remains high and currently, there is no cure for the disease. Novel effective interventions at this stage of life are of exceptional value. Considering their great potential in promoting tissue regeneration and modulating inflammation, mesenchymal stromal cells (MSCs) represent a promising avenue for treating several disorders, including BPD. Umbilical cord-derived MSCs (UC-MSCs) offer biological advantages over other MSC sources (easily available, high proliferative capacity, and better repair efficacy). Pioneering work in our lab showed that MSCs prevent injury to the developing lung in a rat model mimicking BPD. However, there are still considerable challenges that must be overcome before MSCs can be effectively implemented in clinical trials. As such, UC-MSC heterogeneity is poorly understood, with concerns regarding variations from donors and batches. Thus, to improve the reproducibility of basic research and clinical applications, and to identify the optimal therapeutic cell product, better molecular characterization of UC-MSCs and the development of standardized BPD models will be essential in the clinical translation of MSC therapy for BPD. Moreover, considering that BPD is a disease of prematurity, the therapeutic potential of UC-MSCs isolated from preterm birth is of major interest. In the study presented here, using single-cell RNA sequencing (scRNA-seq), we characterized MSCs isolated from the UC of term and preterm pregnancies at delivery (term and preterm donors), as well as non-progenitor control cell line, human neonatal dermal fibroblasts (HNDFs). Moreover, we associated UC-MSC transcriptomic profiles with their therapeutic potential in hyperoxia-induced lung injury in neonatal rats. Finally, we developed and characterized a novel two-hit (2HIT) BPD model in neonatal mice, assessed UC-MSCs' optimal route of injection, timing, and dose, and evaluated their therapeutic effects in that model. We showed that UC-MSCs isolated from the majority of term and preterm donors, including preterm donors with pregnancy-related complications, have limited heterogeneity and possessed a transcriptome enriched in genes related to cell cycle and cell proliferation activity (termed "progenitor-like" cells). In contrast, UC-MSCs isolated from one term and two preterm donors with preeclampsia displayed a unique transcriptome comprised of many genes related to fibroblast activity, including extracellular matrix (ECM) organization (termed "fibroblast-like" cells). In addition, treatment with progenitor-like UC-MSCs, but not with fibroblast-like cells nor HNDFs, significantly improved lung structure, function, and pulmonary hypertension (PH) in hyperoxia-induced lung injury in neonatal rats. We identified marker genes for the therapeutic UC-MSCs (progenitor-like cells) and non-therapeutic cells (fibroblast-like cells and HNDFs). Among them, the high expression of major histocompatibility complex class I (MHCI) is associated with a reduced therapeutic effect. Furthermore, we developed a novel 2HIT BPD mice model with in-depth characterization of the innate immune response and lung injury. 2HIT injury caused a transient type 1 proinflammatory cytokine response and a significant decrease in type 2 anti-inflammatory cytokine lung expression and number of anti-inflammatory M2 type alveolar macrophages. Moreover, 2HIT mice showed impaired lung compliance and growth. Repeated intravenous (i.v.) injections of UC-MSCs at a dose of 20×10⁶ cells/kg body weight (BW) on postnatal day (PD) one and two improved survival, BW, lung compliance, and growth of 2HIT animals. In conclusion, scRNA-seq experimentation provided evidence that UC-MSCs isolated from different donors harbor different transcriptomes with progenitor-like or fibroblast-like characteristics. Only progenitor-like cells provided a therapeutic effect in hyperoxia-induced lung injury in neonatal rats. The development of a novel murine 2HIT BPD model allowed us to characterize the innate immune response and lung pathology and confirm the optimal dose of UCMSCs to provide therapeutic potential in that model. These results will enable better therapeutic selection of UC-MSCs and help improve treatment regimen prior to ultimate clinical translation.

No description available.

No description available.

Über die letzten Jahre haben sich Einzelzelldaten als Trend in der Bioinformatik etabliert, was zu umfangreichen Datensätzen führte. Die Entwicklung von Analysemethoden für solche Daten hat jedoch nicht mit deren Produktion Schritt gehalten. Diese Arbeit befasst sich mit einigen Problemen, die beim Analysieren von Einzelzelldaten auftreten. Das zweite Kapitel enthält eine Analyse von scRNA-seq-Daten von Darmkrebspatienten und Organoiden. Es werden Entwicklungstrajektorien des Darms beschrieben. Anschließend werden Signalgradienten dieser Achsen charakterisiert, insbesondere MAPK- und WNT-Signale. Weiter wird gezeigt, wie RNA velocity basierend auf metabolischen labeling ähnliche Trajektorien in Organoiden aufzeigen kann. Schließlich werden Auswirkungen von Signalwegenhemmung auf die Entwicklungstrajektorien im Detail beschrieben. Das dritte Kapitel bietet einen noch nie dagewesenen Einblick in frühe Stadien von COVID-19 in der Lunge. Verwendete scRNA-seq Daten stammen aus Lungengewebeproben etablierter Hamstermodelle, die mehrere Spezies, SARS-CoV-2-Dosen und Zeitpunkte umfassen, und die ich mit entsprechenden Daten von menschlichen Patienten vergleiche. Für die Analyse zentraler Zelltypen, die den unterschiedlichen Krankheitsverläufen zugrunde liegen, wende ich post-hoc Interpretation auf sonst unzugängliche latent spaces von diffusionmaps an, die neue Einblicke in die zelluläre Pathogenese von COVID-19 bieten. Im letzten Kapitel stelle ich scperturb vor, die größte Sammlung von perturbierten Einzelzelldaten. Ich zeige, wie E-Statistik verwendet werden kann, um solche Daten auf statistisch fundierte Weise zu analysieren. Verzerrungen werden mittels eines neuen Term zur Korrektur der E-Distanz beseitigt. Anschließlich untersuche ich Robustheit der E-Statistiken für einige Analyseszenarien, wie die COVID-19 Daten aus vorherigen Kapitel. Schließlich leite ich Richtlinien für die experimentelle Planung von Einzelzell-Perturbationsstudien mit robuster Statistik ab. / In recent years, single-cell data has emerged as leading trend in bioinformatics, resulting in the generation of substantial datasets. However, development of analysis methods for single-cell data has not kept pace with its production, presenting challenges for analysts. This thesis addresses some of the most pressing issues encountered during the analysis of single-cell data. After a short introductory chapter, the second chapter deals with the problem of arranging single-cell transcriptomes based on biological trajectories, and how these correlate with signaling pathways, specifically those relevant as targets for potential treatments. This thesis demonstrates how RNA velocity based on metabolic labeling can recover similar trajectories in organoids, identifying WNT and MAPK as underlying signaling pathways for development in normal and colon cancer organoids. The third chapter provides an unprecedented view into early stages of COVID-19 in the lungs. For the analysis of key cell types underlying divergent COVID-19 outcomes I apply post-hoc interpretation methods to otherwise inaccessible latent spaces of diffusion maps, revealing new insights into the cellular pathogenesis of COVID-19. Used scRNA-seq data is derived from lung tissue samples of established hamster models, encompassing multiple species, varying SARS-CoV-2 doses, and time points, which I then compare to data from human patients. In the fourth chapter, I present scperturb, the largest collection of single-cell perturbation data. I show how E-statistics can be used to analyze such data in a statistically sound way. After introducing a new bias-correction term to the calculation of E-distances, I investigate the robustness of resulting E-statistics for various analysis scenarios, such as the COVID-19 data from the previous chapter. Finally, I derive guidelines for the experimental design of single-cell perturbation studies such that robust statistics can be achieved.

In dieser Dissertation modellieren und analysieren wir scRNA-Seq-Daten, um Mechanismen, die biologischen Prozessen zugrunde liegen, zu verstehen In scRNA-Seq-Experimenten wird biologisches Rauschen mit technischem Rauschen vermischt. Mittels eines vereinfachten scRNA-Seq-Modells leiten wir eine analytische Verteilungsfunktion für die beobachtete Verteilung unter Kenntnis einer Ausgangsverteilung her. Charakteristiken und sogar ein allgemeines Moment der Ausgangsverteilung können aus der beobachteten Verteilung berechnet werden. Unsere Formeln stellen den Ausgangspunkt zur Quantifizierung von Zellvariabilität dar. Wir haben eine vollständig lineare Analyse von Transkriptomdaten entwickelt, die zeigt, dass sich Zellen während des Zellzyklus auf einer ebenen zirkulären Trajektorie im Transkriptomraum bewegen. In immortalisierten Zelllinien stellen wir fest, dass die Transkriptomdynamiken des Zellzyklus hauptsächlich unabhängig von den Dynamiken anderer Zellprozesse stattfinden. Unser Algorithmus (“Revelio”) bringt eine einfache Methode mit sich, um unsynchronisierte Zellen nach der Zeit zu ordnen und ermöglicht das exakte Entfernen von Zellzykluseffekten. Die Form der Zellzyklus-Trajektorie zeigt, dass der Zellzyklus sich dazu entwickelt hat, Änderungen der transkriptionellen Aktivitäten und der damit verbundenen regulativen Anstrengungen zu minimieren. Dieses Konstruktionsprinzip könnte auch für andere Prozesse relevant sein. Durch die Verwendung von metabolischer Molekülmarkierung erweitern wir Modelle zur mRNA-Kinetik, um dynamische mRNA-Ratenparameter für Transkription, Splicing und Degradation zu erhalten und die Lösungen auf den Zellzyklus anzuwenden. Wir zeigen, dass unser Modell zwischen Genen mit ähnlicher Genexpression aber unterschiedlicher Genregulation unterscheiden kann. Zwar enthalten scRNA-Seq-Daten aktuell noch zu viel technisches Rauschen, unser Modell wird jedoch für das zukünftige Errechnen von dynamischen mRNA-Ratenparametern von großem Nutzen sein. / In this dissertation, we model and analyse scRNA-seq data to understand mechanisms underlying biological processes. In scRNA-seq experiments, biological noise gets convoluted with various sources of technical noise. With the help of a simplified scRNA-seq model, we derive an analytical probability distribution function for the observed output distribution given a true input distribution. We find that characteristics and even general moments of the input distribution can be calculated from the output distribution. Our formulas are a starting point for the quantification of cell-to-cell variability. We developed a fully linear analysis of transcriptome data which reveals that cells move along a planar circular trajectory in transcriptome space during the cell cycle. Additionally, we find in immortalized cell lines that cell cycle transcriptome dynamics occur largely independently from other cellular processes. Our algorithm (“Revelio”) offers a simple method to order unsynchronized cells in time and enables the precise removal of cell cycle effects from the data. The shape of the cell cycle trajectory indicates that the cell cycle has evolved to minimize changes of transcriptional activity and their related regulatory efforts. This design principle may be of relevance to other cellular processes. By considering metabolic labelling, we extend existing mRNA kinetic models to obtain dynamic mRNA rate parameters for transcription, splicing and degradation and apply our solutions to the cell cycle. We can distinguish genes with similar expression values but different gene regulation strategies. While current scRNA-seq data contains too much technical noise, the model will be of great value for inferring dynamic mRNA rate parameters in future research.

<p dir="ltr">The genetic perturbations caused by spaceflight on biological systems tend to have a system-wide effect which is often difficult to deconvolute into individual signals with specific points of origin. Single cell multi-omic data can provide a profile of the perturbational effects but does not necessarily indicate the initial point of interference within a network. The objective of this project is to take advantage of large scale and genome-wide perturbational or Perturb-Seq datasets by using them to pre-train a generalist machine learning model that is capable of predicting the effects of unseen perturbations in new data. Perturb-Seq datasets are large libraries of single cell RNA sequencing data collected from CRISPR knock out screens in cell culture. The advent of generative machine learning algorithms, particularly transformers, make it an ideal time to re-assess large scale data libraries in order to grasp cell and even organism-wide genomic expression motifs. By tailoring an algorithm to learn the downstream effects of the genetic perturbations, we present a pre-trained generalist model capable of predicting the effects of multiple perturbations in combination, locating points of origin for perturbation in new datasets, predicting the effects of known perturbations in new datasets, and annotation of large-scale network motifs. We demonstrate the utility of this model by identifying key perturbational signatures in RNA sequencing data from spaceflown biological samples from the NASA Open Science Data Repository.</p>