Significant advances in fluorescence imaging techniques enable life scientists today to gain insights into biological systems at an unprecedented scale. The interpretation of image features in such bioimage datasets and their subsequent quantitative analysis is referred to as bioimage analysis. A substantial proportion of bioimage analyses is still performed manually by a human expert - a tedious process that is long known to be subjective. Particularly in tasks that require the annotation of image features with a low signal-to-noise ratio, like in fluorescence images of tissue samples, the inter-rater agreement drops. However, like any other scientific analysis, also bioimage analysis has to meet the general quality criteria of quantitative research, which are objectivity, reliability, and validity. Thus, the automation of bioimage analysis with computer-aided approaches is highly desirable. Albeit conventional hard-coded algorithms are fully unbiased, a human user has to set its respective feature extraction parameters. Thus, also these approaches can be considered subjective. Recently, deep learning (DL) has enabled impressive advances in computer vision research. The predominant difference between DL and conventional algorithms is the capability of DL models to learn the respective task on base of an annotated training dataset, instead of following user-defined rules for feature extraction. This thesis hypothesized that DL can be used to increase the objectivity, reliability, and validity of bioimage analyses, thus going beyond mere automation. However, in absence of ground truth annotations, DL models have to be trained on manual and thus subjective annotations, which could cause the model to incorporate such a bias. Moreover, model training is stochastic and even training on the same data could result in models with divergent outputs. Consequently, both the training on subjective annotations and the model-to-model variability could impair the quality of DL-based bioimage analyses. This thesis systematically assessed the impacts of these two limitations experimentally by analyzing fluorescence signals of a protein called cFOS in mouse brain sections. Since the abundance of cFOS correlates with mouse behavior, behavioral analyses could be used for cross-validation of the bioimage analysis results. Furthermore, this thesis showed that pooling the input of multiple human experts during model training and integration of multiple trained models in a model ensemble can mitigate the impact of these limitations. In summary, the present study establishes guidelines for how DL can be used to increase the general quality of bioimage analyses.

The liver plays a pivotal role in maintaining energy homeostasis. Hepatic carbohydrate and lipid metabolism are tightly regulated in order to adapt quickly to changes in nutrient availability. Postprandially, the liver lowers the blood glucose levels and stores nutrients in form of glycogen and triglycerides (TG). In contrast, upon fasting, the liver provides glucose, TG, and ketone bodies. However, obesity resulting from a discrepancy in food intake and energy expenditure leads to abnormal fat accumulation in the liver, which is associated with the development of hepatic insulin resistance, non-alcoholic fatty liver disease, and diabetes. In this context, hepatic insulin resistance is directly linked to the accumulation of diacylglycerol (DAG) in the liver. Besides being an intermediate product of TG synthesis, DAG serves as second messenger in response to G-protein coupled receptor signaling. Protein kinase D (PKD) family members are DAG effectors that integrate multiple metabolic inputs. However, the impact of PKD signaling on liver physiology has not been studied so far. In this thesis, PKD3 was identified as the predominantly expressed isoform in liver. Stimulation of primary hepatocytes with DAG as well as high-fat diet (HFD) feeding of mice led to an activation of PKD3, indicating its relevance during obesity. HFD-fed mice lacking PKD3 specifically in hepatocytes displayed significantly improved glucose tolerance and insulin sensitivity. However, at the same time, hepatic deletion of PKD3 in mice resulted in elevated liver weight as a consequence of increased hepatic lipid accumulation. Lack of PKD3 in hepatocytes promoted sterol regulatory element-binding protein (SREBP)-mediated de novo lipogenesis in vitro and in vivo, and thus increased hepatic triglyceride and cholesterol content. Furthermore, PKD3 suppressed the activation of SREBP by impairing the activity of the insulin effectors protein kinase B (AKT) and mechanistic target of rapamycin complexes (mTORC) 1 and 2. In contrast, liver-specific overexpression of constitutive active PKD3 promoted glucose intolerance and insulin resistance. Taken together, lack of PKD3 improves hepatic insulin sensitivity but promotes hepatic lipid accumulation. For this reason, manipulating PKD3 signaling might be a valid strategy to improve hepatic lipid content or insulin sensitivity. However, the exact molecular mechanism by which PKD3 regulates hepatocytes metabolism remains unclear. Unbiased proteomic approaches were performed in order to identify PKD3 phosphorylation targets. In this process, numerous potential targets of PKD3 were detected, which are implicated in different aspects of cellular metabolism. Among other hits, phenylalanine hydroxylase (PAH) was identified as a target of PKD3 in hepatocytes. PAH is the enzyme that is responsible for the conversion of phenylalanine to tyrosine. In fact, manipulation of PKD3 activity using genetic tools confirmed that PKD3 promotes PAH-dependent conversion of phenylalanine to tyrosine. Therefore, the data in this thesis suggests that PKD3 coordinates lipid and amino acid metabolism in the liver and contributes to the development of hepatic dysfunction.

Touch sensation is the ability to perceive mechanical cues which is required for essential behaviors. These encompass the avoidance of tissue damage, environmental perception, and social interaction but also proprioception and hearing. Therefore research on receptors that convert mechanical stimuli into electrical signals in sensory neurons remains a topical research focus. However, the underlying molecular mechanisms for mechano-metabotropic signal transduction are largely unknown, despite the vital role of mechanosensation in all corners of physiology. Being a large family with over 30 mammalian members, adhesion-type G protein-coupled receptors (aGPCRs) operate in a vast range of physiological processes. Correspondingly, diverse human diseases, such as developmental disorders, defects of the nervous system, allergies and cancer are associated with these receptor family. Several aGPCRs have recently been linked to mechanosensitive functions suggesting, that processing of mechanical stimuli may be a common feature of this receptor family – not only in classical mechanosensory structures. This project employed Drosophila melanogaster as the candidate to analyze the aGPCR Latrophilin/dCIRL function in mechanical nociception in vivo. To this end, we focused on larval sensory neurons and investigated molecular mechanisms of dCIRL activity using noxious mechanical stimuli in combination with optogenetic tools to manipulate second messenger pathways. In addition, we made use of a neuropathy model to test for an involvement of aGPCR signaling in the malfunctioning peripheral nervous system. To do so, this study investigated and characterized nocifensive behavior in dCirl null mutants (dCirlKO) and employed genetically targeted RNA-interference (RNAi) to cell-specifically manipulate nociceptive function. The results revealed that dCirl is transcribed in type II class IV peripheral sensory neurons – a cell type that is structurally similar to mammalian nociceptors and detects different nociceptive sensory modalities. Furthermore, dCirlKO larvae showed increased nocifensive behavior which can be rescued in cell specific reexpression experiments. Expression of bPAC (bacterial photoactivatable adenylate cyclase) in these nociceptive neurons enabled us to investigate an intracellular signaling cascade of dCIRL function provoked by light-induced elevation of cAMP. Here, the findings demonstrated that dCIRL operates as a down-regulator of nocifensive behavior by modulating nociceptive neurons. Given the clinical relevance of this results, dCirl function was tested in a chemically induced neuropathy model where it was shown that cell specific overexpression of dCirl rescued nocifensive behavior but not nociceptor morphology.

The animal diet of the carnivorous Venus flytrap, Dionaea muscipula, contains a sodium load that enters the capture organ via an HKT1-type sodium channel, expressed in special epithelia cells on the inner trap lobe surface. DmHKT1 expression and sodium uptake activity is induced upon prey contact. Here, we analyzed the HKT1 properties required for prey sodium osmolyte management of carnivorous Dionaea. Analyses were based on homology modeling, generation of model-derived point mutants, and their functional testing in Xenopus oocytes. We showed that the wild-type HKT1 and its Na\(^+\)- and K\(^+\)-permeable mutants function as ion channels rather than K\(^+\) transporters driven by proton or sodium gradients. These structural and biophysical features of a high-capacity, Na\(^+\)-selective ion channel enable Dionaea glands to manage prey-derived sodium loads without confounding the action potential-based information management of the flytrap.

Background: In the phase III AVAGAST trial, the addition of bevacizumab to chemotherapy improved progression-free survival (PFS) but not overall survival (OS) in patients with advanced gastric cancer. We studied the role of Angiopoietin-2 (Ang-2), a key driver of tumour angiogenesis, metastasis and resistance to antiangiogenic treatment, as a biomarker. Methods: Previously untreated, advanced gastric cancer patients were randomly assigned to receive bevacizumab (n = 387) or placebo (n = 387) in combination with chemotherapy. Plasma collected at baseline and at progression was analysed by ELISA. The role of Ang-2 as a prognostic and a predictive biomarker of bevacizumab efficacy was studied using a Cox proportional hazards model. Logistic regression analysis was applied for correlations with metastasis. Results: Median baseline plasma Ang-2 levels were lower in Asian (2143 pg ml\(^-\)\(^1\)) vs non-Asian patients (3193 pg ml\(^-\)\(^1\)), P<0.0001. Baseline plasma Ang-2 was identified as an independent prognostic marker for OS but did not predict bevacizumab efficacy alone or in combination with baseline VEGF. Baseline plasma Ang-2 correlated with the frequency of liver metastasis (LM) at any time: Odds ratio per 1000 pg ml\(^-\)\(^1\) increase: 1.19; 95% CI 1.10-1.29; P<0.0001 (non-Asians) and 1.37; 95% CI 1.13-1.64; P = 0.0010 (Asians). Conclusions: Baseline plasma Ang-2 is a novel prognostic biomarker for OS in advanced gastric cancer strongly associated with LM. Differences in Ang-2 mediated vascular response may, in part, account for outcome differences between Asian and non-Asian patients; however, data have to be further validated. Ang-2 is a promising drug target in gastric cancer.