Glioblastoma (GBM) remains the most aggressive and prevalent malignant primary brain tumor in adults. Unchanged since 2005, standard of care (SoC) consists of surgical resection, followed by radiation therapy (RT) with concurrent and adjuvant chemotherapy with temozolomide (TMZ). Despite these therapeutic efforts, patients succumb to recurrent disease with a median overall survival of 14.6 months and a five-year survival rate of 5.5-6.8%. Therapeutic failure is largely explained by ITH and the presence of treatment-resistant GBM stem-like cells (GSCs). Given the lack of understanding of recurrent GBM and absence of second line therapies patients, I hypothesize that genome-scale functional genetic interrogation will unravel recurrent GBM-specific tumor biology and inform development of novel therapeutics. First, I compared primary and recurrent GBM at the genetic, transcriptomic, proteomic and functional genetic levels. These analyses map a multilayered genetic response to drive tumor recurrence, identifying protein tyrosine phosphatase 4A2 (PTP4A2) as a novel modulator of self-renewal, proliferation and tumorigenicity at GBM recurrence. Mechanistically, genetic perturbation and a small molecule inhibitor of PTP4A2 repress axon guidance activity through a dephosphorylation axis with roundabout guidance receptor 1 (ROBO1) and exploit a genetic dependency on ROBO signaling. Importantly, engineered anti-ROBO1 single-domain antibodies also mimic the effects of PTP4A2 inhibition. Given the genetic dependency on ROBO signaling and enrichment of ROBO1 expression in GBM tissues, I undertook a campaign to evaluate ROBO1 as a therapeutic target in recurrent GBM and develop anti-ROBO1 chimeric antigen receptor T (CAR-T) cells using camelid single-domain antibodies targeting human ROBO1. I optimized the design of anti-ROBO1 CAR-T cells and tested the anti-tumor activity of these modalities in in vitro using patient-derived recurrent GBM lines and orthotopic patient-derived xenograft models. I present data to expand the repertoire of GBM-enriched antigens suitable for effective CAR-T cell therapy. Given that resistance to SoC and disease relapse are inevitable for GBM patients, pre-clinical and clinical advancement of immunotherapeutic modalities, combined with recent insights into the tumor immune microenvironment, are poised to improve clinical outcomes for this patient population. / Thesis / Doctor of Philosophy (PhD) / Glioblastoma remains the most lethal and prevalent primary brain tumor in adults. Standard of care for patients remains unchanged since 2005, consisting of surgery to remove visible tumor at diagnosis (primary tumor), followed by radiation therapy and chemotherapy to treat remaining tumor cells. Despite these therapeutic efforts, tumor relapse (recurrent tumor) is inevitable with no standardized second-line therapy. Patients succumb to recurrent disease with a median overall survival of 14.6 months and only 5.5-6.8% of patients survive five years post diagnosis. Therapy failure and tumor relapse are explained by immense diversity among tumor cells at the DNA and protein levels, giving rise to a subset of tumor cells with abilities to resist therapy and seed the recurrent tumor. Previous studies have presented evolution of tumor cells through therapy, with recurrent tumor cells harboring novel changes at the DNA and protein levels. However, the impact of these changes on tumor cell function has not been evaluated. In this thesis, we developed and applied a genetic screening technique to determine the functional role of thousands of genes in primary and recurrent tumor cells from the same patient. This analysis revealed numerous genes that exhibit differential effects on survival of primary and recurrent tumor cells, including genes that drive recurrent tumor cell growth but are dispensable in primary tumor cells. Functional remodeling of these genes and pathways revealed a new functional role of multiple proteins belonging to a process called axonal guidance in recurrent tumor cells. To evaluate the therapeutic potential of these findings, we deeply interrogated the mechanism by which axonal guidance drives recurrent tumor cells and targeted crucial molecular players using chemical and immunological therapies. Using models that predict clinical effectiveness, we engineered and tested a novel therapy that redirects immune cells to target recurrent tumor cells driven by dysfunctional axonal guidance activity. The goal of this thesis was to discover the functional differences between primary and recurrent tumor cells, thereby leveraging this information to engineer candidate therapies for treatment of recurrent glioblastoma.

Das Bronchialkarzinom ist die häufigste Todesursache bei den Krebserkrankungen und weist eine schlechte Prognose auf. Die Behandlung besteht aus einer Chemotherapie mit platinbasierten Medikamenten, doch der Erfolg ist unbefriedigend. In den letzten Jahren wurden zielgerichtete Therapien gegen Proteine wie den EGFR entwickelt. Klinische Studien zeigten, dass nur Subpopulationen von den Medikamenten Erlotinib und Cetuximab profitieren. Eine bessere (Vor-)Selektion der Patienten ist wünschenswert, um unnötige Behandlungen zu vermeiden. Für diese Analysen bedarf es relevanter präklinischer Modelle. Im Rahmen dieser Arbeit wurden 25 Xenograftmodelle des Lungenkarzinoms vergleichend charakterisiert. Ein Schwerpunkt bestand im Vergleich der Xenografts mit ihren Patiententumoren. Die Analyse der Histologie, der Proliferationsmarker als auch der Genexpressionsprofile fand übereinstimmende Ergebnisse in den Patiententumoren und ihren abgeleiteten Xenografts. Mit Hilfe von mRNA-, Protein- und SNP-Profilen ressistenzassoziierter Marker der Chemotherapie konnte die Bedeutung der Modelle zur Charakterisierung von prädiktiven und prognostischen Markern aufklärt werden. Diese Arbeit untersuchte auch Marker der anti-EGFR-Therapien. mRNA- und Proteinprofile der ERBB-Rezeptoren sowie der Liganden wurden erstellt und stimmten mit publizierten klinischen Daten überein. Genexpressionsstudien in Erlotinib Respondern und Non-Respondern zur Therapieoptimierung identifizierten den Wachstumsfaktor VEGFA als Ziel für eine Kombinationsbehandlung mit dem Angiogeneseinhibitor Bevacizumab. Die Kombination von Bevacizumab mit Erlotinib führte zu einem reduzierten Tumorwachstum. Die Ergebnisse dieser Arbeit machten deutlich, dass die individuellen Tumoreigenschaften in den patientenabgleiteten Xenografts auf Gen- und Proteinebene erhalten bleiben und diese als Modelle zur Markeranalyse sowie zur Therapieoptimierung eingesetzt werden können. / Lung cancer is still one of the most frequent cancers worldwide. The treatment option is classical chemotherapy that is based upon the combination of platin-based drugs. But no further improvement seems to be possible. For some years targeted drugs against single proteins like the EGFR were developed. The clinical trials showed that only subpopulations of patients benefit from the treatment. A better selection of patients to avoid treatment would be helpful. Therefore, pre-clinical models that are suitable for analysis and that represent clinical populations of patients are required. In this work 25 patient derived xenografts from lung cancer were intensely studied. First, the xenografts were compared with their corresponding patient tumor. The analysis of the histology and the expression of proliferation and epithelial or mesenchymal markers showed concordance of the patient tumor and the derived xenograft. The gene expression profiles were also maintained. Further analysis should elucidate the relevance of the xenografts as models for the characterisation and validation of predictive and prognostic markers. SNP, mRNA and protein expression profiles of resistance markers for chemotherapy were generated and showed similarities with clinical data. As marker for the anti-EGFR targeted therapies the ERBB receptors and the ligands of the EGFR were analysed. The mRNA and protein expression profiles resemble clinical data sets. An optimisation of the therapy should be achieved with gene expression studies. The vascular endothelial growth factor A was identified for a combination treatment with the anti-angiogenic drug bevacizumab in erlotinib resistant tumors. The combination of erlotinib and bevacizumab reduced the tumor growth in selected models. In summary, the analysis could show that the individual characteristics of the patient tumor were maintained in the xenograft. The models are a reliable tool for studies designed to improve treatment strategies.

Protein misfolding and aggregation underlie several neurodegenerative proteinopathies including amyotrophic lateral sclerosis (ALS). Superoxide dismutase 1 (SOD1) was the first gene found to be associated with familial ALS. Overexpression of human mutant or wild type SOD1 in transgenic mouse models induces motor neuron (MN) degeneration and an ALS-like phenotype. SOD1 mutations, leading to the destabilization of the SOD1 protein is associated with ALS pathogenesis. However, how misfolded SOD1 toxicity specifically affects human MNs is not clear. The aim of this thesis was to develop patient-derived, cellular models of ALS to help understand the pathogenic mechanisms underlying SOD1. To understand which cellular pathways impact on the level of misfolded SOD1 in human cells, we established a model using patient-derived fibroblasts and quantified misfolded SOD1 in relation to disturbances in several ALS-related cellular pathways. Misfolded SOD1 levels did not change following reduction in autophagy, inhibition of the mitochondrial respiratory chain, or induction of endoplasmic reticulum (ER)-stress. However, inhibition of the ubiquitin-proteasome system (UPS) lead to a dramatic increase in misfolded SOD1 levels. Hence, an age-related decline in proteasome activity might underlie the late-life onset that is typically seen in SOD1 ALS. To address whether or not SOD1 misfolding is enhanced in human MNs, we used mixed MN/astrocyte cultures (MNCs) generated in vitro from patient-specific induced pluripotent stem cells (iPSCs). Levels of soluble misfolded SOD1 were increased in MNCs as well as in pure iPSC-derived astrocytes compared to other cell types, including sensory neuron cultures. Interestingly, this was the case for both mutant and wild type human SOD1, although the increase was enhanced in SOD1 FALS MNCs. Misfolded SOD1 was also found to exist in the same form as in mouse SOD1 overexpression models and was identified as a substrate for 20S proteasome degradation. Hence, the vulnerability of motor areas to ALS could be explained by increased SOD1 misfolding, specifically in MNs and astrocytes. To investigate factors that might promote SOD1 misfolding, we focussed on the stability of SOD1 mediated by a crucial, stabilizing C57-C146 disulphide bond and its redox status. Formation of disulphide bond is dependent on oxidation by O2 and catalysed by CCS. To investigate whether low O2 tension affects the stability of SOD1 in vitro we cultured fibroblasts and iPSC-derived MNCs under different oxygen tensions. Low oxygen tension promoted disulphide-reduction, SOD1 misfolding and aggregation. This response was much greater in MNCs compared to fibroblasts, suggesting that MNs may be especially sensitive to low oxygen tension and areas with low oxygen supply could serve as foci for ALS initiation. SOD1 truncation mutations often lack C146, and cannot adopt a native fold and are rapidly degraded. We characterized soluble misfolded and aggregated SOD1 in patient-derived cells carrying a novel SOD1 D96Mfs*8 mutation as well as in cells fom an unaffected mutation carrier. The truncated protein has a C-terminal fusion of seven non-native amino acids and was found to be extremely prone to aggregation in vitro. Since not all mutation carriers develop ALS, our results suggested this novel mutation is associated with reduced penetrance. In summary, patient derived cells are useful models to study factors affecting SOD1 misfolded and aggregation. We show for the first time that misfolding of a disordered and disease associated protein is enhanced in disease-related cell types. Showing that misfolded SOD1 exists in human cells in the same form as in transgenic mouse models strengthens the translatability of results obtained in the two species. Our results demonstrate disulphide-reduction and misfolding/aggregation of SOD1 and suggest that 20S proteasome could be an important therapeutic target for early stages of disease. This model provides a great opportunity to study pathogenic mechanisms of both familial and sporadic ALS in patient-derived models of ALS. / Varje år insjuknar omkring 5300 personer i världen i motorneuronsjukdomen Amyotrofisk lateralskleros (ALS). Sjukdomen kännetecknas av degeneration av motorneuron i hjärnan och ryggmärgen, de nervceller som styr kroppens muskler, vilket leder till musklerförtvining och gradvis förlamning. ALS-patienter avlider oftast till följd av andningssvikt när sjukdomen når andningsmuskulaturen. I de allra flesta fall uppkommer ALS sporadiskt (SALS), det vill säga utan känd genetisk orsak, medan ärftliga fall (FALS) drabbar omkring 10 % och beror på mutationer i ett antal kända gener. Upp till 6 % av alla ALS fall kan härledas till mutationer i genen superoxid dismutas 1 (SOD1). SOD1 är ett enzym som ansvarar för att omvandla och oskadliggöra fria syreradikaler som bildas vid normal ämnesomsättning. 206 olika SOD1 mutationer har identifierats, alla orsakar inte ALS men många leder till att den tredimensionella proteinstrukturen förändras, vilket ökar proteinets benägenhet att felveckas. Initialt trodde man att SOD1 mutationer förhindrade proteinets normalfunktion och följaktligen orsakade ALS. Studier har emellertid visat att den enzymatiska funktionen ofta bevaras, även hos muterade proteiner. Däremot kan små mängder felveckat SOD1 störa andra viktiga cellulära funktioner. Felveckat SOD1 har en benägenhet att klumpa ihop sig och bilda aggregat i det centrala nervsystemet (CNS). Dessa aggregat återfinns hos patienter med såväl FALS som SALS vilket tyder på att även vildtyps-SOD1 kan felveckas och vara involverat i sjukdomsutvecklingen. De flesta studier är baserade på transgena musmodeller som uttrycker extremt stora mängder av muterat humant SOD1. Det är dock oklart hur väl studier i möss överensstämmer med sjukdomsutvecklingen hos ALS-patienter, där mängden SOD1 är betydligt lägre. En central fråga som fortfarande står obesvarad är varför just motorneuron degenererar i ALS, trots att SOD1 uttrycks i alla kroppens celler. Det övergripande syftet med den här avhandlingen har varit att karakterisera felveckat SOD1 i patientceller för att studera dess roll i ALSrelaterade sjukdomsmekanismer med fysiologiskt relevanta nivåer av SOD1. Samtliga studier är gjorda in vitro med celler från friska donatorer med vildtyps-SOD1, celler från patienter med SOD1-FALS, FALS som bär andra ALS-associerade gener, samt SALS. I de allra flesta fallen har vi analyserat både lösligt felveckat SOD1 samt aggregerade former av SOD1 proteinet.