Search

The Wnt/Planar Cell Polarity (PCP) signalling is responsible for the establishment of planar polarity in epithelial tissues. Besides this function, it is also known to play multiple roles in neuronal development, however, the mechanism by which PCP signalling acts in these processes remained largely unclear. Neurons are highly polarised cells with extended neuronal protrusions in order to connect to target cells. The growth and maintenance of these protrusions crucially depend on the cytoskeleton. In recent years, many direct regulators of the neuronal cytoskeleton have been described, however it is still not completely understood how external or internal signals are conveyed to the neuronal cytoskeleton. The aim of this project was to study how core members of the PCP signalling module regulate the neuronal cytoskeleton, and to understand how they contribute to axon growth in Drosophila melanogaster. To this end, we used primary embryonic cell cultures which provide powerful subcellular readouts to determine the state of the neuronal cytoskeleton; and we studied the embryonic and the adult nervous system in order to assess the in vivo relevance of our findings. Our results showed that most PCP proteins are present both in the embryonic central and peripheral nervous system, and the loss of some of these components causes an early stall of the intersegmental nerve b (ISNb) in embryos. In addition, we found that the PCP proteins regulate axon growth and guidance in the ventral lateral neurons (LNv), a group of neurons in the adult brain that are responsible for controlling the locomotor activity and sleeping behaviour of the flies, as the LNv axons project ectopically in most of the PCP mutants. Accordingly, these pathfinding defects result in a strongly reduced LNv function and lead to severe alterations, particularly in the sleep/wake transition, since the morning activity of the PCP mutant flies is disrupted and becomes associated with an increased amount of daytime sleep. In accordance with these, we found that some PCP proteins affect axon growth in primary neuronal cell cultures as well. In addition, the lack of some of the PCP components alters the state of the actin and microtubule (MT) cytoskeleton in cultured neurons. In order to investigate the mechanism by which these PCP proteins can control the neuronal cytoskeleton and contribute to axon growth, we also studied DAAM (Dishevelled Associated Activator of Morphogenesis), a formin type of protein which has previously been linked to PCP signalling. Besides its known actinregulatory function, we found that DAAM strongly affects the MT cytoskeleton as well, in ways similar to the PCP proteins. Moreover, the loss of DAAM causes a similar ISNb stalling phenotype than that of the PCP proteins in the embryo. Together these findings suggest that the PCP proteins are necessary to properly regulate the neuronal cytoskeleton, presumably by controlling DAAM, and contribute to axon growth in vivo in the embryonic nervous system and in the adult brain. Beyond that, our studies also revealed a new physiological function of the Wnt/PCP pathway in controlling the sleeping behaviour of the adult flies.

Software systems have become an important part of our daily lives, and a multitude of different application scenarios, user requirements, and hardware requirements have emerged. To handle these different requirements, most software systems offer some degree of configurability in terms of configuration options, allowing the user to adapt the software system to functional and non-functional requirements. Among non-functional requirements, the performance of the software system plays an important role to end-users. It is often unclear which configuration options influence the performance of the system. Specifically, there is a gap in how to select configurations affecting the system's performance when no previous knowledge is available. Furthermore, little is known about how the influence of configuration options on the system's performance changes across different workloads and software evolution. To bridge this gap, performance modeling based on statistical learning has proved useful. In this thesis, we follow three objectives in which we use or improve performance modeling of configurable software systems by statistical learning. First, we propose a novel sampling strategy, distance-based sampling, to improve the configuration selection (i.e., \emph{sampling}) while also addressing the shortcomings of existing state-of-the art sampling strategies. To assess the advantages and limitations of distance-based sampling, we compare it to state-of-the-art sampling strategies on multiple real-world configurable software systems. Our results indicate that distance-based sampling outperforms other state-of-the-art sampling strategies in terms of accuracy, but also suggest that there is still room for improvement with regard to scalability. Second, to assess how the influence of configuration options on the system's performance changes during software evolution, we use performance modeling on multiple real-world configurable software systems. This investigation delivers multiple valuable insights into the frequency of performance changes with implications to other research domains. We further investigate in how many cases the performance changes are documented by developers and find indications in which cases performance regressions are documented and when they are not. Third, besides configurability and software evolution, we also assess the role of workload variability in an exploratory study of the configurable software system \textsc{FastDownward}. For this purpose, we propose a performance modeling approach to identify performance changes considering workload variability and investigate the accuracy of this approach by evaluating precision and recall. Our results show that our approach is able to identify most performance changes, but we also identify the limitations of our approach, leaving room for further improvement. Furthermore, our performance measurements proved helpful in that they enabled us to discover and report multiple performance regressions in a real-world configurable software system. Overall, we contribute to performance modeling of configurable software systems by (1) proposing a novel sampling strategy designed to cover configuration spaces with regards to performance; (2) lifting how performance changes affect software configurability in practice and show implications on other research areas; (3) demonstrating how performance modeling can be used to find performance changes while including workload variability. To the best of our knowledge, we are the first to investigate configurability, evolution, and workload variability of configurable software systems together.

The interaction of water vapor with the surface and bulk of poly(imidesiloxane) copolymers has been characterized in an attempt to determine the important factors in the copolymer's resistance to water ingression. The multi-block copolymers were synthesized from benzophenone tetracarboxylic dianhydride, bisaniline P and pre-formed amine-terminated poly(dimethylsiloxane) oligomers, with phthalic anhydride as an end-capping agent. Similar copolymers had been previously shown to have reduced water sorption, increased surface hydrophobicity, and increased adhesive durability in hot/wet environments. Inverse gas chromatography was used to conduct surface energetics studies on copolymers of different siloxane concentration and a polyirnide homopolymer. Free energies of specific interaction of water vapor, ΔGsp°, with the polymer surfaces were found to decrease with the incorporation of siloxane into the polyirnide. The dispersive components of the solid surface free energy of the siloxane- containing copolymers were equal within error to that of pure poly(dimethylsiloxane), indicating a PDMS-rich, hydrophobic surface. The ΔGsp° of the copolymers were not significantly different, suggesting that the copolymer surfaces were very similar. This indicated a minimum weight percent of siloxane incorporation required to maximize the copolymer's surface water resistance. The minimum amount for the studied system was at most ten percent. Diffusion coefficients of water vapor in the polyimide and copolymers were determined from gravimetric sorption experiments. Higher levels of siloxane incorporation caused a definite increase in the diffusion coefficient, indicating a decreased resistance to water ingression. The increase in diffusion was found to be influenced by siloxane block length and was interpreted in terms of morphological and free volume theories. The diffusion coefficient of a 10 weight percent PDMS copolymer, however, was found to be the same within error as the polyimide diffusion oefficient. The incorporation of siloxane into polyimides has been shown to increase water resistance due to the hydrophobicity of the siloxane-rich surface. However, high siloxane contents also increased the rate of water ingression in the bulk of the polymer. Increased water resistance of the surface may be achieved at lower siloxane concentrations without increasing diffusive (or decreasing mechanical) properties to undesirable levels.