Proteins regulate cellular processes in health and disease. Each protein’s function is determined by its structure. The structures, and therefore functions, of some proteins change to adapt to cellular perturbations or in disease. Hence, the ability to monitor alterations of protein structures is central to understanding how cells adapt to perturbations and how they are dysregulated in disease. High resolution structures or structure predictions are now available for a majority of proteins but little is known about dynamics of protein structures. Learning about protein structure dynamics is also crucial for understanding of dynamic protein assemblies, in which proteins can undergo major structural rearrangements. Dynamic protein assemblies range from small labile protein complexes, to macromolecular structures like condensates and proteins undergoing liquid-liquid phase separation. Membraneless condensates such as stress granules are associated with neurodegenerative diseases. For example, proteins like FUS and TDP-43 implicated in neurodegenerative diseases undergo liquid-liquid phase separation and aggregation during disease progression, and complexes of multiple enzymes are known to regulate central metabolic pathways. However, despite their central roles in diseases and metabolism, dynamic protein assemblies are insufficiently characterized because they dissociate easily, for example upon cell lysis, and also because studying them in the cell is difficult. We lack techniques to characterize these structures in the cell and to study their dynamics. Even if their components are known, it is difficult to characterize molecular rearrangements upon phase separation directly in the cellular environment. For this reason, dynamic assemblies are studied by fluorescence microscopy in cells or by reconstituting assemblies from single components. Both approaches require knowledge of the participating molecules, fluorescence microscopy provides poor structural information, and reconstitution does not allow accounting for the effects of crowding and all cellular components. Hence, so far it was neither possible to confidently determine which proteins participate in stress granule formation, nor to provide structural evidence for predicted multi-enzyme complexes in glycolysis. We lack techniques to study protein structures and their dynamics within the cell and to detect altered protein structures upon specific perturbations in an unbiased manner. 6 This would be particularly important to study proteins involved in condensates and phase separation because they are affected by cell lysis. The aim of the thesis was to develop approaches to study the dynamics of protein structures in situ and on a global scale and to use these methods to attempt the characterization of dynamic protein assemblies. To study proteome-wide structural changes with peptide-level resolution, we developed a methodology to apply limited proteolysis-coupled mass spectrometry (LiP-MS) within cells. In LiP-MS, proteins are cleaved by a broad-specificity protease for a short time. Previous work had demonstrated that LiP-MS provides information on structure dynamics, but so far the method had only been applied in native cell lysates. To enable use of LiP-MS in cells, we first tested different strategies to introduce broad-specificity proteases into cells. Expression and photocaging of the proteases proteinase K, subtilisin, and papain did not lead to sufficient proteolytic activity for proteome-wide cleavage. Similarly, delivery of proteinase K via extracellular vesicles resulted only in low-level cleavages. However, sufficient levels of proteolytic cleavage were achieved by introducing proteinase K via electroporation. Electroporation permeabilized the cells and resulted in reproducible proteome-wide and precisely timed cleavages. We optimized conditions for introduction of proteinase K to maximize proteome and sequence coverage of structure-specific cleavage. LiP-MS in cells captured not only the known binding of rapamycin to FKBP1A within the cell but also downstream effects of pathway activation. By comparing LiP-MS applied before and after cell lysis, we identified human proteins that are susceptible to structural change upon cell lysis. In particular RNA and DNA binding proteins were structurally altered upon cell lysis and should be studied preferably by intracellular methods. As a proof of principle, we applied LiP-MS to human cells treated with sodium arsenite which is known to induce the formation of stress granules, dynamic molecular condensates. We used this approach to probe the dynamics of stress granule formation at the molecular level and within the cell. Using this novel in-cell LiP-MS technique, we found global protein structural alterations upon sodium arsenite treatment and captured the intracellular structural dynamics of hundreds of proteins with peptide-level resolution. Both known and novel structural alterations of proteins involved in stress granules were detected. This dataset provides a resource describing the structural changes of human proteins in response to cellular stress and pinpoints structurally altered regions with a resolution of single functional sites. Interestingly, we 7 found that nuclear speckles, another type of biomolecular condensate, rearrange upon arsenite treatment too, and validated this alteration by fluorescence microscopy. Moreover, using a combination of size-separation and crosslinking techniques, we assessed the possibility that enzymes in glycolysis and gluconeogenesis assemble into supramolecular complexes also called metabolons. Specifically, we studied enzyme structural rearrangements when switching between glucose and acetate supplemented media. We observed that many glycolytic enzymes participated in labile RNA-mediated assemblies without major rearrangements between glycolysis and gluconeogenesis. Only pathway-specific enzymes and their direct neighbors changed their assembly state, and gluconeogenesis-specific enzymes did so independently of RNA. We show that gluconeogenic enzymes fbp, fbaA, and tpiA and glycolytic enzymes pfkA, fbaA, and tpiA interact in vitro, and fbaA forms defined heteromeric complexes with pfkA and fbp. Using lysine-specific crosslinks as restraints for docking of enzyme structures, we present a model of complex formation in upper glycolysis and gluconeogenesis. In summary, my research led to the development of a novel method that allows study of structural dynamics of proteins in cells upon various perturbations with peptide-level resolution on a proteome-wide scale. Application of this method provided insights into the formation of stress granules upon arsenite stress. Our method also allowed identification of previously unknown structural rearrangements, as shown for nuclear speckle proteins. Moreover using a combination of biochemical, imaging, and mass spectrometric techniques, we provide evidence for labile multi-enzyme complexes in glycolysis and gluconeogenesis. These datasets and methods can be used in the future to better understand dynamic protein assemblies and other cellular processes directly in their native environment.