Stem cell therapies hold tremendous potential for tissue and organ regeneration. Yet, there remains significant need for better ex vivo culture and manipulation methods. On the one hand, many tissue-specific stem cells cannot be propagated without causing rapid deviations in cellular phenotype. On the other hand, outside of a developing organism, it remains very difficult to differentiate stem cells into mature functional cell types. Such protocols could help to build more effective and predictive pre-clinical models that would reduce or even replace animal studies. Manipulating a stem cell in vitro requires a thorough understanding of its microenvironment, or niche, and how niche factors influence stem cell fate. A customizable microarray platform was therefore developed to examine the interplay of biochemical and mechanical signals on stem cell behavior. This array was implemented to study combinations of 12 different proteins with varying substrate stiffness and their effect on the adipogenic differentiation of human mesenchymal stem cells (MSCs). These cells and their differentiated progenies are of high therapeutic relevance for skeletal tissue regeneration and have also been implied as niche-modulating neighbors of hematopoietic stem cells (HSCs) in the bone marrow (BM). Using this combinatorial approach, substrate stiffness was found to be the major determinant of the frequency of lipid accumulation in MSCs, while the lipid content was found to be strongly driven by the biochemical context. To further understand how niche cells together with stem cells self-organize to form a functional tissue, three-dimensional (3D) organ-mimicking structures from spheroids of pluri- and multi-potent have recently been developed. These systems, termed organoids, are expected to provide instrumental insights into tissue or organ development and they could be of great use for disease-specific drug screenings and to generate clinically transplantable tissues. However, current protocols for organoid generation remain highly inefficient and irreproducible. To perform the high-throughput aggregation of one or multiple cell types and to allow their long-term culture, a high-density array of U-shaped microwells was developed. This platform was used to initiate 3D co-cultures of the hematopoiesis-supportive cell line OP9 together with HSCs. To expand on the potential of this co-culture model to study more relevant primary BM cells, uniform clonal mesensphere cultures from human MSCs were established. However, restricted knowledge about the phenotypic identity of both niche and stem cell is a major limitation to initiate these co-cultures in vitro. Therefore, in this thesis a cell-targeting approach through covalent SNAP-tag-based cell-pairing is proposed. Specifically, SNAP-tag fusion proteins expressed on cell membranes were targeted to form covalent bonds with substrates that harbor the SNAP-tag substrate benzylguanine (BG). Immobilizing sufficient amounts of BG on cellular membranes is a mean to enable cell-cell pairing as demonstrated by the covalent aggregation of microbeads carrying SNAP-tag and BG at high concentrations. The artificial niche models developed in this study are broadly applicable to desired cellular systems. These approaches help to gain insight into the multifactorial composition and architectural organization of specific niches and to enable their reconstruction in vitro.