A multilayered structural proteomics approach to detect and model dynamic changes of proteins and protein complexes in situ
EMBARGOED UNTIL 2027-12-20
Loading...
Author / Producer
Date
2024
Publication Type
Doctoral Thesis
ETH Bibliography
yes
Citations
Altmetric
EMBARGOED UNTIL 2027-12-20
Data
Rights / License
Abstract
Proteins build the functional core of all biological systems. Their three-dimensional structure determines their function and dynamically adapts to meet the requirements of changing cellular environments. Those changes can translate into different enzyme activity states or into changed affinities towards distinct binding partners that can result in rewiring of protein-protein interaction (PPI) networks and altogether define a specific cellular phenotype. Elucidating the structure of proteins and their adaptation to different environments is thus essential to understanding their molecular functioning.
While experimental protein structure determination is time and labor intensive, recent advances in machine learning allow highly accurate protein structure prediction from their sequence without the need of any experimental data. Although powerful, all models were trained on mostly static snapshots of proteins isolated from their cellular environment and thus highest scoring predictions do not necessarily reflect the state most relevant to a given condition. Even though efforts are made to predict the conformational space of proteins by modelling ensembles or by molecular dynamics simulations, it is currently not possible to accurately predict protein dynamics and alternative confirmations in situ and thus experimental data reporting on the structural state of a protein in a given cellular environment remains crucial. Our knowledge of the dynamics of protein complexes is equally sparse. Methods to characterize PPIs exist and have led to our current understanding that proteins form densely interconnected networks that rewire across conditions. However, all of those methods are time and labor intensive and do not allow to monitor interaction network rewiring across conditions easily. To address those limitations, my thesis aimed to bridge the gap between our current static view of protein structures and PPIs and their known dynamic behavior. Specifically, I aimed at complementing current structure predictors and tools to probe PPI networks with approaches that combine in situ structural proteomics and computational biology to monitor dynamic changes of proteins and protein complexes globally and at high throughput.
We first conducted a targeted analysis of the potential existence of a dynamic glycolytic metabolon in Escherichia coli and its assembly changes across metabolic conditions. We found crosslinking during cell lysis to be essential to detect assemblies of glycolytic enzymes by size exclusion chromatography coupled to mass spectrometry (SEC-MS), suggesting that these enzymes form rather labile and dynamic heteromeric complexes. Further, we found that the majority of those assemblies is dependent on RNA and that only assemblies of enzymes catalyzing pathway-specific reactions rearranged between glycolytic and gluconeogenic conditions with gluconeogenesis-specific enzymes doing so in an RNA-independent manner. By crosslinking coupled to mass spectrometry (XL-MS), we found that pathway-specific enzymes physically interact in vitro and proposed structural models of protein complexes by molecular docking that suggest adenosine phosphates to regulate complex formation. Finally, we corroborated identification of some of those interactions in situ. Thus, our data supports the existence of a glycolytic metabolon in E. coli with pathway-specific enzymes dynamically changing across metabolic conditions and highlights the importance of PPI changes to adapt to different environments.
Next, to monitor PPI dynamics globally and at high-throughput, we developed FLiP-MS, a novel structural proteomics workflow that uses filtration to separate protein assemblies by size and tests for protease susceptibility changes between different assembly states of the same protein retained in different filter fractions by limited proteolysis coupled to mass spectrometry (LiP-MS). Comparing LiP patterns across fractions allows to identify protein binding interfaces and other protein regions that change upon complex formation. Applied to Saccharomyces cerevisiae, this created a PPI marker library for more than 1,000 proteins enriched in protein-protein binding interfaces. In combination with LiP-MS analyses of whole cell extracts, the library allows to pinpoint proteins likely to change their interactions as shown by our approach correctly recapitulating PPI changes required for P-body formation under DNA replication stress. We further proposed a computational pipeline integrating prior knowledge about PPIs and network analysis to generate a protein complex-centric view of the rearrangements. We captured known and novel protein complex rearrangements,
supporting the key role played by Spt-Ada-Gcn5 acetyltransferase activity in the stress response. Our data suggest several protein complex rearrangements that depend directly or indirectly on Gcn5 activity and identified a link between Gcn5 activity and the regulation of P-bodies. To conclude, our novel structural proteomics pipeline enables tracking of PPIs globally, at high throughput, in situ and for any condition of interest and complements static PPI networks with dynamic readouts.
To more generally elucidate the dynamics of structural proteome rearrangements including both dynamics of single proteins and protein complexes, we generated the to date largest multilayered structural proteomics dataset comparing yeast grown under glycolytic and gluconeogenic conditions. We combined LiP-MS reporting on structural changes via changes in protease susceptibility and XL-MS probing structural changes via differences in distances between residues. We further evaluated whether monolinks, a side product of the crosslinking reaction, could be used as a readout reporting on differences in surface accessibilities and integrated the FLiP-MS pipeline to simultaneously track PPI changes. We found that although a large portion of the proteome is regulated by abundance, a similar number of proteins is regulated by structure only, with enzymes of upper glycolysis being prime examples. Mapping crosslinking data to experimental structures revealed that some proteins adopt conformations closer to an active state characterized in a homologous protein, highlighting the importance of integrating experimental data to select the structure most relevant to a given cellular environment. Further, our structural readouts correctly recapitulated known changes of glycolytic enzymes undergoing the glycolytic shift and proved valuable in generating specific hypothesis about the type and location of structural changes. Importantly, our data allowed integrative three-dimensional modelling of environment-specific protein conformations for 630 proteins and distinct conformations in glycolytic and gluconeogenic conditions for 304 of these proteins. We showed that our condition-specific structural models correctly recapitulated a known switch from an active to an inactive enzymatic state for pyruvate decarboxylase. Further, our approach allowed detecting a potentially new ADP binding site in 3-phosphoglycerate kinase in an alternative modelled enzyme conformation, which indeed showed altered accessibility at increasing ADP concentrations by a LiP-based titration experiment. Thus, our dataset generated the first proteome-wide landscape of protein structure dynamics that complements static experimental structures, enabling to generate mechanistic hypotheses and further allowed for integrative modelling of environment- and condition-specific protein conformations.
To conclude, in my thesis I underlined the potential existence of a dynamic glycolytic metabolon in E. coli, I developed a novel pipeline enabling to track PPIs globally and at high throughput and generated the largest integrative structurome rearrangement model supported by distinct structural proteomics readouts. We envision, that applying those techniques across conditions and organisms will finally bridge the gap between our current static view of protein structures and PPIs and their actual dynamics. Protein structure modelling could further benefit from this type of in situ datasets since they contain valuable information about unstructured regions that most experimental models are blind to and integrate effects of the cellular environment. Applied to a disease context, our approaches have the potential to detect structurally dysregulated proteins and model their disease-specific conformations for better selection of druggable protein candidates and drug lead identification.
Permanent link
Publication status
published
External links
Editor
Contributors
Examiner : Picotti, Paola
Examiner : Leitner, Alexander
Examiner : Stelling, Jörg
Examiner : Weis, Karsten
Examiner : Correia, Bruno
Book title
Journal / series
Volume
Pages / Article No.
Publisher
ETH Zurich
Event
Edition / version
Methods
Software
Geographic location
Date collected
Date created
Subject
Proteomics; Cross-linking mass spectrometry; limited proteolysis-coupled mass spectrometry; metabolon; Systems Biology; Modelling
Organisational unit
03927 - Picotti, Paola / Picotti, Paola
Notes
Funding
Related publications and datasets
Is variant form of: