Zebrafish xenograft models have been increasingly recognized for their ability to predict patient responses to cancer therapeutics, suggesting their potential as diagnostic tools in clinical settings. However, these models require the precise microinjection of cancer cell suspensions in many small and fragile zebrafish larvae. Manual injections are so challenging that, even after months of training, variability in experimental results persists among researchers. This limits the uptake and deployment of zebrafish xenograft models for clinical use and drug discovery. To address this challenge, we have designed, built, and validated an automated microinjection robot. Combined results of injections into the vasculature, perivitelline space, and hindbrain ventricle demonstrated an average injection success rate of approximately 60%, with a larvae survival rate exceeding 70%, comparable to manual injections using a traditional micromanipulator. Notably, the full automated mode was twice as fast as manual injections. This automation of the microinjection process significantly reduces the need for extensive personnel training while it enhances reproducibility, efficiency, and accuracy, paving the way for more extensive use of zebrafish xenograft models in drug discovery and patient diagnostics.

Zebrafish xenograft models are increasingly recognized for predicting patient responses to cancer therapeutics, suggesting their potential as clinical diagnostic tools. However, precise microinjection of cancer cells into numerous small and fragile zebrafish larvae is laborious, requires extensive training for new operators, and often yields variable results, limiting their clinical and drug discovery applications. To address these challenges, we have designed, built, and validated an automated microinjection robot. The robot performs injections into the vasculature, perivitelline space, and hindbrain ventricle in both fully automated and semi-automated modes. Combined results demonstrate an average injection success rate of approximately 60% and larvae survival exceeding 70%, comparable to manual methods, with the fully automated mode being twice as fast. This automation reduces the need for extensive personnel training while enhancing reproducibility, efficiency, and accuracy, paving the way for more extensive use of zebrafish xenograft models in drug discovery and patient diagnostics.

Diffuse midline glioma (DMG) is an aggressive pediatric brain tumor driven by the H3K27M histone mutation and represents the leading cause of cancer-related death in children. These tumors are highly infiltrative and can occasionally migrate to distant CNS regions. To uncover migration dependencies, we developed a novel two-step pooled whole-genome CRISPR-migration screen in metastatic H3K27M-DMG stem cells (n=3). Genes involved in focal adhesion (ITGB1 [integrin beta-1], CRKL, PARVA, PTK2, FERMT2) significantly restricted migration across all models; notably, only ITGB1 knockout (ITGB1-KO) completely abrogated migration. In H3K27M-DMG patient samples, unlike other brain tumor types, expression of ITGB1 correlates with higher glioma grade and worse survival. ITGB1-KO models demonstrated a reduction in expression of MYC target genes, including MYC-regulated metabolic genes involved in purine biosynthesis (e.g., IMPDH2). Further integrated RNA/metabolomic analyses revealed that loss of ITGB1 downregulates purine metabolism and the citric acid (TCA) cycle. Importantly, in in vivo models, ITGB1 deficiency significantly prolonged survival (UMPED83: 100 vs. 163.5 days, p=0.0003; pSCG-SVZ: 49 vs. 68 days, p=0.0095). Spatial transcriptomic and proteomic analyses of ITGB1-KO orthotopic H3K27M tumors showed widespread reduction in MYC target genes and depletion of precursor, undifferentiated (OPC-like), and an increase in differentiated (OC-like) K27M cells in the infiltrating edge. Direct pharmacological targeting of ITGB1 (anti-ITGB1 antibody, CNS delivered) significantly extended survival in DMG models (UMPED83: 100 vs. 125.5 days, p=0.0169; pSCG-SVZ: 49 vs. 72 days, p=0.0384). However, ITGB1-deficient pSCG-SVZ tumors exhibited compensatory alternative integrins upregulation (i.e., ITGB3, ITGB5). Promisingly, co-treatment with anti-ITGB1 antibody and cilengitide (ITGB3/5 inhibitor) further improved survival and resulted in 75% long-term survivors, free of disease. Strikingly, this combinatorial strategy failed to confer any survival benefit in adult glioblastoma (H3WT) models. Overall, these findings highlight integrin targeting as a promising therapeutic avenue in H3K27M-DMG, capable of disrupting tumor-specific migration, MYC-driven purine biosynthesis, and stemness programs.

BACKGROUND H3K27-altered diffuse midline gliomas (DMGs) have poor prognosis with no standard of care therapy beyond radiation (RT). While RT prologues survival, not all patients respond. Prior work has shown somatic TP53 mutations predict radiation resistance, and poorer outcomes. We report on a multi-center retrospective cohort study to identify molecular biomarkers of RT response in DMG patients. METHODS We performed retrospective chart reviews of patients with biopsy-proven H3K27M-altered DMG between 2013-2023 from six US medical centers and the DMG Center in Zurich. Patients with tumor genomic sequencing and completion of RT were eligible. Genomic alterations were evaluated for somatic mutations, fusions, and chromosomal instability. Cox proportional hazard models were used to evaluate associations between genomic alterations and survival. Multivariate analysis included all significant variables at P <0.1 in initial analysis, including age at diagnosis, tumor location, TP53 and PIK3R1 alterations. RESULTS 297 patients (135 female; median age 8.3years; range 0.2-71.4 years) were included. Median progression-free survival (PFS) and overall survival (OS) was 7.6months (95%-CI 6.9-8.6) and 14.0months (95%-CI 12.9-15.7). Univariate analysis identified TP53 or PIK3R1 status to be associated with shorter OS (TP53-mutant 12.5months vs. TP53-wildtype 17.5months; HR=1.5; 95%-CI (1.1-2.0), P=0.005); (PIK3R1-mutant 12.6months vs. PIK3R1-wildtype 14.5months; HR=1.9, 95%-CI (1.1-3.4), P=0.03). Multivariate analysis corroborated TP53 status association with shorter OS (HR=1.5; 95%-CI (1.2-2.0), P=0.003). Subgroup univariate analysis of pontine DMG patients showed reduced OS with TP53, NF1, or TERT mutations (TP53-mutant 11.9months vs. TP53-wildtype 15.2months; HR=1.6; 95%-CI (1.1-2.3), P=0.02); (NF1-mutant 8.4months vs. NF1-wildtype 13.6months; HR=2.3; 95%-CI (1.1-5.0), P=0.03); (TERT-mutant 8.5months vs. TERT-wildtype 13.6months; HR=2.5; 95%-CI (1.1-5.7), P=0.03). Subsequent multivariate analysis corroborated the association of TERT mutations with shorter OS in pontine DMG (HR=2.5; 95%-CI (1.05, 5.9), P=0.03). CONCLUSIONS This is one of the largest molecular characterized cohorts of H3K27M-altered DMG patients. DMGs with somatic TP53 or PIK3R1 mutations demonstrate inferior OS. Pontine DMG patients with somatic TP53, NF1 or TERT mutations demonstrate inferior OS.

BACKGROUND MYC-driven Group-3 medulloblastomas (MB) are deadly and malignant pediatric brain cancers and we sought to define actionable metabolic dependencies in these tumors. METHODS To identify uniquely upregulated genes in Group-3 MB, we performed transcriptomic analysis on two previously published medulloblastoma RNA-seq datasets. To elucidate the relationship between c-MYC/IDH1/DLAT and assess impact on tumor metabolism, we performed metabolic and transcriptional profiling of Group-3 MB cell lines that were either untreated or were subjected to shRNA-mediated knockdown of DLAT or treatment with IDH1 inhibitor. We also treated Group-3 MB cell lines containing varying levels of DLAT expression with copper ionophore elesclomol and assessed its ability to induce toxicity. Finally, we established in vivo models of Group-3 MB via orthotopic implantation to assess the effect of DLAT knockdown, IDH1 inhibition, and cuproptosis induction on tumor growth and survival outcomes. RESULTS We identified upregulation of dihydrolipoyl transacetylase (DLAT), the E2-subunit of pyruvate dehydrogenase complex (PDC) in a subset of Group-3 MB. DLAT was induced by c-MYC and targeting DLAT lowered TCA-cycle metabolism and glutathione synthesis in Group-3 MB cells. We also noted upregulation of isocitrate dehydrogenase 1 (IDH1) in Group-3 MB. Remarkably, genetic and pharmacologic suppression of IDH1 epigenetically reduced c-MYC and downstream DLAT levels. DLAT is a central regulator of cuproptosis, a copper-dependent cell death mechanism induced by the copper ionophore elesclomol. DLAT expression in Group-3 MB cells correlated with increased sensitivity to cuproptosis. Elesclomol was CNS-penetrant and suppressed tumor growth in vivo in Group-3 MB animal models. CONCLUSIONS Our data uncover an IDH1/c-MYC dependent vulnerability that regulates DLAT levels and can be targeted to kill Group-3 MB by cuproptosis.

Dwarf galaxies are the most common type of galaxy in the Universe and have emerged as powerful testbeds for our standard cosmological model LambdaCDM on small scales, where several tensions persist. This thesis focuses specifically on the phase-space distribution of dwarf satellites around massive hosts. Several nearby hosts, including the Milky Way, appear to have flattened, kinematically correlated satellite configurations that are uncommon among LambdaCDM analogs---the planes-of-satellites challenge. Robust tests require large, contamination-controlled satellite samples across diverse environments, which are challenging to obtain beyond the Local Group because dwarfs are intrinsically faint and generally have low surface brightness. This thesis contributes to the extension of phase-space studies beyond the Local Group by (i) evaluating and refining methods to quantify phase-space distributions, in particular lopsided satellite distributions and planes-of-satellites, and conducting statistical tests on such distributions in existing survey data (e.g., MATLAS, ELVES); (ii) expanding line-of-sight velocity coverage via VLT/MUSE spectroscopy of MATLAS candidates and establishing host membership; and (iii) developing an automated pipeline, combining classical methods with deep learning, to build a survey-scale catalog of dwarf galaxy candidates in the wide-field UNIONS survey. Across 68 host systems in MATLAS/ELVES, ~21% show significant lopsidedness with the signal strongest at larger projected radii, consistent with recent accretions and in line with LambdaCDM expectations. Follow-up on reported candidate planes generally revealed no significant tension with LambdaCDM when revisited with new data. An exception to this trend is the NGC 4490 group, where such a highly correlated structure was identified as uncommon in simulated analogs. In the MUSE program, we confirmed 75% of the MATLAS dwarfs in the sample as satellites; non-members tend to be late-types, supporting morphology as a membership prior. Our pipeline yielded dwarf probability scores for 1.5 million selected objects, producing the Galaxies OBserved as Low-luminosity Identified Nebulae (GOBLIN) catalog. GOBLIN contains ~43,000 high probability (>=0.8) dwarf candidates, which represents a significant increase in the number of high-confidence candidates in the local Universe. Taken together, the majority of our investigations into phase-space distributions of dwarfs revealed consistency with LambdaCDM given current data. There are, however, noteworthy exceptions, and several systems are still far from complete in terms of distance and velocity estimates. Such measurements are necessary to draw definitive conclusions. Our publicly available GOBLIN catalog contains a large sample of high-probability dwarf candidates, laying the foundation for targeted follow-up campaigns and future phase-space studies.

The kinetic inertness and low solubility of Au3(pyrazolate)3 complexes have made it difficult to integrate them into supramolecular assemblies. In this thesis, an investigation into the dynamic combinatorial chemistry of these complexes is presented, followed by strategies for their incorporation into molecular cages and a study of the host - guest chemistry of the resulting assemblies. The first research chapter, Chapter 2, describes the constitutional dynamic chemistry of Au3(pyrazolate)3 complexes. Ligand exchange reactions between"free" pyrazole ligands and Au3(pyrazolate)3 complexes were performed, and a pronounced autocatalytic behavior was observed. This led to the understanding that pyrazole ligands can act as catalysts for ligand scrambling between two Au3(pyrazolate)3 complexes. In addition, four crystal structures of heteroleptic Au3(pyrazolate)2(pyrazolate') are presented. Building on this foundation, Chapter 3 describes the synthesis of a dodecanuclear Au(I) cage. The cage consists of four Au3(pyrazolate)3 units, connected through an organic linker. The cage was first synthesized by a ligand exchange reaction, using an Au3(triazolate)3 complex as a gold source. Surprisingly, the same cage was also obtained in a direct synthesis from a reaction with AuCl(SMe2). The results highlight the importance of ligand design for the successful formation of Au3(pyrazolate)3-based cages. In Chapters 4 and 5, we demonstrate how the kinetic inertness of Au cyclic trinuclear complexes (CTCs) can be utilized in order to use them as pre-formed building blocks for cage synthesis. In Chapter 4, a stepwise approach, in which the use of metalloligands enabled the construction of a Fe24Au24Pd8 cage, is presented. With a molecular weight of 21 kDa and a diameter of approximately 4.1 nm, it is among the largest [Pd6L8]12+ cages reported. In addition, we demonstrate the use of a heteroleptic Au3(pyrazolate)2(pyrazolate') complex for the construction of a smaller Fe8Au12Pd2 cage. In Chapter 5, the preparation of three Au3(pyrazolate)3-based cages by combining coordination chemistry and dynamic covalent chemistry is described. First, two Au3(pyrazolate)3 complexes bearing peripheral aldehyde groups were synthesized. These complexes were used as building blocks for an imine condensation reaction, yielding one tetrahedral cage, containing four CTC units, and two prismatic cages, containing two CTC units each. The crystal structures of all three assemblies are reported. In addition, the host-guest chemistry with pi-acidic guests was studied: the tetrahedral cage was found to bind fullerenes with high affinity, while the prismatic cages encapsulate halogenated aromatic compounds. In Chapter 6, the combination of coordination chemistry and dynamic covalent chemistry for the synthesis of a new family of zirconium-based cages is presented. Pre-formed zirconium clusters were employed as rigid, preorganized building blocks, leading to the synthesis of four new cages: a compact [1+1] species, a [2+3] architecture, and two large tetrahedral [4+4] cages.

For decades, cancer biology research has primarily focused on intrinsic tumor mechanisms driven by genetic lesions. It is now clear that tumor cell behavior is shaped by additional layers of regulation, including epigenetic modifications and metabolic reprogramming, which confer high plasticity and render cancer cells highly responsive to cues from their surrounding microenvironment. Continuous cross-talk between tumor cells and the diverse components of the tumor microenvironment dictates tumorigenesis and is often skewed toward pro-tumoral signals, thereby fueling cancer progression. In B-cell lymphomas, a heterogeneous group of malignancies arising from mature B cells, immune-tumor interactions are particularly important, especially in early disease and indolent subtypes. The increasingly precise understanding of these interactions has recently driven the development of immune-based therapies, which have achieved transformative results in both lymphomas and solid tumors. Nevertheless, tumors continuously adapt and evade immune control, and relapsed or refractory cancers remain largely incurable despite recent advances. Overcoming these limitations requires novel therapeutic strategies and preclinical models that reproduce the complexity of human disease.

In this thesis, I describe a strategy to enhance anti-cancer immunity by targeting long-chain fatty acid (LCFA) elongation. By combining high-throughput drug screening with proteomics, lipidomics, and genetic engineering, we identified Phago Booster 1 (PB1), a small-molecule inhibitor of HSD17B12, a key enzyme in LCFA elongation. Targeting HSD17B12 selectively altered membrane organization in cancer cells, leading to the redistribution of surface receptors, including immune checkpoints and lipid transporters. In contrast, HSD17B12 inhibition enhanced the cytotoxic capacity of immune cells by increasing their glycolytic activity, thereby promoting activation. This divergent effect proved therapeutically relevant in vivo, leading to slower tumor growth and improved survival without notable toxicity. In parallel, I contributed to the development of a patient-derived lymphoma tissue explant platform designed to preserve native tumor architecture, stromal components, and immune cell diversity. We demonstrated that this system, termed lymphomoids, can support personalized medicine approaches by enabling the ex vivo assessment of patient-specific therapeutic responses. In a pilot cohort of 8 patients, drug sensitivity in lymphomoids successfully matched clinical outcomes in 89% of cases, highlighting its potential as a predictive tool for guiding treatment selection.

Together, these approaches illustrate how combining innovative therapeutic discovery with physiologically relevant disease models can uncover actionable strategies to overcome immune resistance and advance personalized cancer treatment.

Millions of people around the world suffer from neurodegenerative diseases. While specific treatments may be able to relieve some of the symptoms of neurodegenerative diseases, in most cases, these complex neurological disorders do not have a cure as we do not completely understand the cause of the disease. The current theory for the mechanism of neurodegenerative diseases is that the causative protein misfolds into a beta-sheet conformation triggering the formation of amyloid fibrils that aggregate into distinctive cellular inclusions. One such protein is alpha-synuclein (aSyn) which accumulates into fibrils in Parkinson's disease (PD), dementia with Lewy bodies (DLB), and multiple system atrophy (MSA). When aSyn amyloid fibrils clump together in diseases such as PD and DLB, they form Lewy bodies. As amyloid fibrils are thought to be the causative agent of the disease, many therapeutic strategies are being developed to inhibit the formation of the fibrils. To facilitate structure-based drug design, several methods have emerged in recent years to generate disease fibrils in vitro and obtain the atomic resolution 3D structure using cryo-electron microscopy. Of the many structures which are emerging from these studies, one major question is if these in vitro-generated structures are the same as what is seen in human disease. In this research, I will use multiple electron microscopy approaches to understand the physiological structure of aSyn fibrils in DLB, which is the second most common form of dementia after Alzheimer's disease. My first aim is to determine the in vitro structure of aSyn fibrils extracted from DLB patients. Cryo-EM is a tool that allows us to get the high-resolution structure of the fibrils in native conditions by plunge freezing in liquid ethane. Several software packages have been developed for electron microscopy structure determination, including RELION (REgularised LIkelihood Optimization); My second aim is to use room-temperature correlative light and electron microscopy (CLEM) method on resin-embedded brain sections to obtain the global morphology of aSyn inclusions in DLB brain donors. The inclusions will be first localized in the brain slices by fluorescence microscopy, and the structure will be analyzed by electron microscopy and electron tomography. I will investigate the inclusions in different brain regions and multiple patients. This analysis will inform us about different inclusion types present in DLB brains and define the targets for high-resolution cryo-CLEM; My final aim is to obtain the high-resolution structure directly within post-mortem human brain samples by cryo-CLEM. This will provide a direct representation of the pathological fibril structure, which can be used for drug discovery and mouse models. Cryo-CLEM and CLEM will be used to study the fibrillar structures in different parts of the post-mortem human brain with DLB. By correlating the information from post-mortem human brains with in vitro fibril structures, I can identify the specific fibril strain which makes up the pathology seen in DLB and PD. The results of this study may lead to the discovery of new therapeutic approaches for synucleinopathies and, in particular, for the prevention or slowing of neurodegeneration associated with Lewy Body diseases.

This thesis systematically evaluates wind farm power efficiency under cyclic yaw control (CYC) across various layouts and inflow conditions via extensive wind tunnel testing. The core mechanism governing the efficacy of CYC in wind farm power improvement is investigated. Inspired by the kinematic similarity between yawing turbines and floating offshore wind turbines (FOWTs) in motions, the work expands to a second part on the power performance and wake characteristics of a FOWT under prescribed platform motions.

The first part of the thesis on CYC includes four studies. The first study, on a single and three-turbine farm, found that CYC accelerates wake recovery, leaving more power available downstream and thus increasing wind farm power production. However, the power of the wind farm does not increase monotonically with the yaw frequency (or Strouhal number) or amplitude. Intriguingly, CYC can even reduce turbulence intensity in the core wake region compared to the baseline.

The second study broadened the scope to different farm layouts and inflow turbulence intensities, also evaluating power fluctuations (a key grid power quality metric). Results show the optimal Strouhal number depends strongly on farm length rather than inflow turbulence intensity and spanwise spacing. The percentage of power gain due to CYC decreases as inflow turbulence increases, highlighting its key role in wind farm control. Notably, within a certain Strouhal range, CYC simultaneously increases mean power production and reduces power fluctuations.

The third study provided the mechanistic insight: periodic wake meandering induced by CYC is the primary driver for power improvement. The periodic wake dynamics sustain throughout the farm at low Strouhal numbers but are suppressed at high frequencies. Inflow conditions, especially turbulence intensity and length scales, can significantly affect the evolution of these CYC-generated wake dynamics. CYC can also alter interactions between wind farm wakes and the atmospheric boundary layer flow.

The fourth study validated an analytical approach based on phase-averaged wake analysis to predict the wake of a wind turbine under CYC. The model accounts for the impacts of CYC on wake velocity and deflection. It is validated experimentally for moderate Strouhal numbers, providing a tool for dynamic wake control scenarios.

The second part expands the scope of this thesis to FOWTs. Three studies on prescribed pitch, surge, and roll motions (at moderate Strouhal numbers) reveal distinct impacts on different motions. Pitch motion influences power and enhances wake recovery, with its generated wake structures affecting downstream turbines. In contrast, surge motion has minimal effect on power or wake characteristics but primarily impacts turbine power fluctuations. Roll motion accelerates wake recovery, increases turbulence intensity, and enhances lateral meandering. Under strong dynamics, roll can also affect thrust and cause wake deflection, presenting a challenge for FOWT analytical wake modeling.