Finding biotechnological solutions to industrial processes is a labor- and timeintensive task but offers the opportunity to introduce more environmentally friendly and cost-effective approaches to chemical manufacturing. In fact, the products of industrial biotechnology claim a continuously increasing market share. The development of these processes relies on efficient procedures for the evolution of biocatalysts toward the desired application. At the same time, continuous improvement of the manufacturing process is performed to reach the highest possible titers and yields, minimizing costs and the impact on the environment. In this PhD thesis, we aimed to facilitate the engineering of industrial hydrolases and support the generation of cost-effective manufacturing procedures based on the improvement of enzyme secretion in Bacillus subtilis, one of the most used bacterial production strain in industrial biotechnology. Directed evolution is an interplay of generating high genetic variability (i.e. large cell libraries) followed by phenotypic characterization of the variants via high throughput analytical platforms. The implementation of powerful high throughput screening setups has immediate impact on the success of directed evolution campaigns, as an increase in throughput directly leads to higher chances of identifying improved biocatalyst, while increased accuracy of the phenotypic characterization reduces the number of clones that need to be analyzed in detail in follow up assays. The implementation of such high throughput assays is a field of intense research efforts, in part due to the introduction of microfluidic droplet technologies to the field, which allow efficient miniaturization of reaction compartments. In our work, we designed novel assays for secreted industrial hydrolases based on the nanoliter reactor (NLR) high throughput screening. This technology allows the encapsulation of single cell variants in hydrogel beads of up to 65 nL volume to both achieve the compartmentalization of single variants and the generation of a semi-confined space that can be engineered to allow the study of secreted products, still allowing exchange of nutrients with the environment. The NLRs show very high cell compatibility and can be analyzed via flow cytometry, allowing screening campaigns of up to 106 clones per day. As the analysis of these compartments relies on fluorescence, we developed customized protocols to correlate the desired phenotypic feature to a fluorescent read-out. In Chapter 2 we focus on engineering of a peptidase for its use beyond the food sector toward novel pharmaceutical applications. We addressed one of the most common food-related illnesses: the coeliac disease, which is currently affecting about 1% of the Western population. Coeliac disease consists of an inflammatory response to the malabsorption of peptides derived from the digestion of gluten. Therapeutical strategies foresee the use of prolyl endopeptidases as an adjuvant to a gluten-free diet, helping in reducing the concentration of immunogenic peptides in the food. In addition, such enzymes may help food and beverage production to maintain gluten-free manufacturing processes. With this goal, we designed a novel NLR-based high throughput screening (HTS) assay to engineer secreted prolyl endopeptidases for improved clearance of highly immunogenic peptides. Two substrates, among the most resistant to proteolysis by gastrointestinal proteases, were selected and a strategy for sensitive detection in NLRs was developed. Through random mutagenesis on the gene coding for the target protease and the application of the developed screening procedure, we were able to identify strains secreting engineered variants that showed improved activity toward both substrates. In Chapter 3, we tackled improvements in the secretion levels of a recombinant amylase. We focused on the rational design of the N-terminus of the target protein, which is the region recognized by the secretory pathway for its translocation across the membrane. We designed a workflow that combines high throughput screening in NLRs and machine learning to efficiently analyze large libraries of signal peptides (SPs) with the goal of understanding the most relevant physicochemical parameters in the SP design that can influence secretion. We tested and validated this workflow starting with a large library of signal peptides fused to an industrially relevant hydrolase and analyzed the secretion levels by monitoring enzyme activity in NLRs. All variants were binned based on their secretion efficiency and used to train a machine-learning model aimed at predicting the performance of any signal peptide fused to the target protein. In addition, the model was interpreted to understand the relevance of defined features of the signal peptide sequence on secretion. The model was evaluated against a set of newly designed signal peptides and showed high precision in predicting their performance. iii Chapter 4 investigates different approaches for the screening of efficient production strains in the NLR platform based on direct quantification of secreted proteins, without relying on enzymatic activity of the target protein. Non-enzymatic proteins represent a growing market in the food and cosmetic industry, for example in the development of biotechnological processes for proteins normally derived from animals and plants. For these proteins, we added small peptide tags to the protein of interest, which we considered a minor modification to the molecule that should neither drastically impact the capacity for secretion nor the function of the protein. As part of this work, we first investigated the widely distributed hexahistidine tag and achieved protein retention in NLRs via the co-encapsulation of its corresponding Ni2+-NTA affinity resin. The detection of the immobilized His6-tagged protein was successfully obtained by developing a competitive assay in NLRs between the secreted protein and a similarly labeled low affinity fluorescent probe. The setup showed some limitations as the dissociation kinetics of the protein affected the stability of the system over time. In light of these limitations, we selected an alternative peptide tag, the sortag. This peptide tag is recognized by the transpeptidase sortase A that catalyzes the covalent ligation of this tag to a polyglycine peptide and thus overcomes the aforementioned stability issues. The approach relied on the secretion of the protein with a C-terminal sortag and its retention in NLRs via a customized resin presenting the polyglycine peptide on the surface (with purified sortase A added in the cultivation medium). This new strategy was preliminarily validated with purified proteins and the amount of immobilized protein was quantified with the addition of a labelled antibody specific for the tagged protein. Overall, the projects described in this PhD thesis show the relevance and need for rapid screening tools for the engineering and directed evolution of biocatalysts, with the overall goal of applying these enzymes to reduce the ecological footprint of industrial processes and offer novel products to improve the quality of life of the final consumer.