In natural ecosystems, spatially structured growth of microorganisms is ubiquitous, from biofilms in aquatic natural environments to aggregates in dairy products. These spatial structures influence how microorganisms behave; they shape colonisation patterns, mediate interactions such as nutrient exchange or competition, and provide protection that helps communities remain robust in changing environments. To better understand spatially structured growth and translate the insights into applications, researchers have developed laboratory systems that mimic these environments. Apart from revealing how spatial organisation affects microbial fitness and interactions, such systems also enable the coupling of individual cells or communities to their performance. Among these tools, microdroplet-based platforms have gained prominence. Microdroplets provide isolated environments for microorganisms to grow, interact, and evolve. Microdroplet-based platforms also offer a high-throughput approach to discover and select unique microbial properties that are challenging to access with conventional methods. This thesis explores how compartmentalised microbial growth in microdroplets can be harnessed for application in biotechnology and ecology.

Microdroplets are generated by mixing an aqueous phase, which contains nutrients for microbial growth, with an oil phase containing an oil-soluble surfactant that prevents coalescence of droplets. The oil phase serves as a diffusion barrier for nutrients between microdroplets, ensuring that each droplet functions as an isolated compartment. In addition, millions of microdroplets are generated in a small volume, facilitating high-throughput cultivation. The introduction (Chapter 1) of this thesis summarises the biotechnological and ecological applications of microdroplet-based technologies for high-throughput cultivation, selection, and screening of microbial cells and communities. This chapter highlights studies where selecting certain microbial phenotypes is particularly challenging to achieve using conventional techniques, but where microdroplets offer unique solutions. Chapter 1 discusses technical and practical aspects, including microdroplet types, microdroplet generation, encapsulation, and distribution of cells between microdroplets, and constraints for growth in microdroplets.

Alongside the ability to generate millions of microdroplets, they offer the advantage of requiring small volumes of aqueous phase—often less than a millilitre—thereby reducing costs of large-scale screening efforts. Chapter 2 proposes a microdroplet based approach for low-volume, label-free, and high-throughput screening of Saccharomyces cerevisiae strains resistant to costly antimetabolites, a common strategy for selecting metabolite overproducers and carbon-de-repressed enzymes. Using this method, 16,000-fold enrichment of an antibiotic-resistant strain and a 600-fold enrichment of a glucose-analogue-resistant strain were achieved in a single round of selection.

Microdroplets can also be applied to select for increased biomass yield on glucose. Such selection for high-biomass-yield strategies can provide insights into metabolic strategies and constraints imposed by cellular regulations and metabolism. By physically separating cells into individual droplets, microdroplet cultivation eliminates competition for glucose between cells in different droplets, enabling the selection of cells with higher cell-number yield on glucose. A higher cell-number yield can, but not necessarily, translate to a higher biomass yield on glucose. In S. cerevisiae, increased biomass yield on glucose often corresponds to a metabolic shift from less energetically efficient (respiro)fermentation toward more efficient respiration. In a previous study, S. cerevisiae mutants were propagated in emulsion with glucose as the sole carbon source, resulting in enrichment of mutants with increased biomass yield. In Chapter 3, these S. cerevisiae mutants were further characterised. Experiments showed that, in addition to an increased cell-number yield, the mutants also showed increased biomass yields on glucose. These increased yields coincided with a decrease in ethanol yield on glucose, which is consistent with a shift in metabolism from (respiro)fermentation toward respiration. Among other mutations, these emulsion-propagated isolates showed amino-acid substitutions in hexokinase 2, which catalyses the first reaction of glycolysis. In Chapter 3, these mutations were reverse-engineered into the parental strain of the emulsion-propagated isolates as well as into another commonly used laboratory strain of S. cerevisiae. Despite strain-dependent differences in phenotype, the reverse-engineered strains exhibited higher biomass yields and lower ethanol yields on glucose. Further experiments showed that this phenotype is most likely due to a loss of catalytic activity caused by the mutations.

Building on the above mentioned work with single-strain selection in microdroplets, the last two chapters of the thesis shift focus to microbial communities. In nature, microbial growth often occurs in communities, where interactions shape the overall community function and stability. Deciphering these interactions is crucial for understanding microbial communities. Microdroplets are promising for this purpose as they enable high-throughput parallel cultivation of sub-communities. However, existing microdroplet-based approaches require cumbersome or sometimes infeasible steps, such as isolation of community members and labelling cell types with fluorescent markers via genetic engineering. Chapter 4 proposes a microdroplet-based, label-free, and isolation-independent approach to decipher microbial pairwise interactions. In this approach, sub-communities were sorted based on growth, and through a combination of experimental data and probabilistic modelling of experimental steps, it was demonstrated that pairwise interactions can be deciphered in three-member Lactococcus cremoris consortia. Furthermore, this approach was computationally validated to decipher pairwise interactions in larger consortia.

Understanding microbial interactions opens possibilities to design robust microbial co-cultures for industrial applications. Currently, industrial biotechnological processes are largely dominated by monocultures with rare examples of co-cultures. Chapter 5 of this thesis reviews studies that explore potential advantages of co-cultures over monocultures, with special attention to studies that provide quantitative data to enable reliable comparisons with monocultures. This chapter also outlines the key challenges that currently limit the implementation of co-cultures in industrial processes.

The thesis is concluded by reflecting on the future broader implications of microdroplet-based approaches for biotechnological applications and the study of microbial ecology.

Large-scale microbial-biotechnology processes for production of chemicals almost exclusively rely on pure cultures of microbial strains. Especially for extensively engineered pure cultures, process performance can be negatively affected, which can be caused by issues such as pathway imbalance, deterioration of productivity caused by genetic instability and enzyme promiscuity. An increasing number of studies demonstrate that, under ‘academic’ laboratory conditions, the use of defined co-cultures (i.e. deliberate mixtures of known microbial strains) offers unique possibilities for mitigating such drawbacks. These advantages differ for dissimilatory products, whose synthesis from one or more carbon substrates provides cells with free energy, and assimilatory products, whose synthesis requires a net input of free energy. Based on advances in experimental and theoretical research, this paper highlights how defined co-cultures can address several limitations of mono-cultures for production of low-molecular-weight compounds. From this largely academic perspective, we outline the key challenges for scaling these systems to industry, which underscore the need for innovative solutions and continued research in this area.

Microorganisms form communities, and their interactions shape the function and stability of these communities. Understanding these interactions can aid in revealing ecosystem dynamics, enhancing community function, and informing the design of synthetic consortia for industrial applications. Deciphering microbial interactions is challenging due to the difficulty of culturing natural microorganisms and the exponential increase in experiments with expanding consortium size. One approach to improving culturing throughput is the use of microcompartments such as agarose microbeads. Microbead-based techniques enable the generation of large numbers of picolitre-sized compartments, facilitating high-throughput, parallel studies of microbial sub-communities. However, the existing microbead-based techniques for deciphering microbial interactions are dependent on single-culture isolates of consortium members and/or labelling of consortium members with fluorescent markers via genetic engineering. We developed a microbead-based, label-free method that eliminates the requirement of single-cell isolates to predict microbial interactions. Our method involves an isolation-independent manner of microbead inoculation with different sub-communities and microbead sorting to separate sub-communities based on growth. Using a probabilistic model, we predict interactions based on cell concentrations and relative abundances in the inoculum and after microbead sorting. We successfully predicted pairwise interactions in two three-member consortia. Additionally, we computationally showcased the validity of our approach for predicting pairwise interactions in larger consortia.