Christine Peter

Microtubules are dynamic polymers essential for intracellular organization, with basic constituents being αβ-tubulin dimers. Microtubule growth and shrinkage is tightly regulated by GTP hydrolysis in β-tubulin, yet how the information about the bound nucleotide propagates across the tubulin dimer, beyond its direct binding site, remains unclear. Here we combine extensive all-atom molecular dynamics simulations of αβ-tubulin dimers with GTP or GDP in the β-subunit to investigate longitudinal allosteric signaling. We featurize conformations using residue-wise minimum distances between the two subunits, apply EncoderMap dimensionality reduction and HDBSCAN clustering to identify conformational substates, and perform dynamic network analysis on these clusters. We find that the tubulin dimer relaxes from the straight starting structures into a broad ensemble of bent conformations, and that the bound nucleotide does not by itself separate this ensemble. The simulations retain a comparatively long-lived memory of their starting structure that makes the subtler nucleotide signal harder to detect. Network analysis identifies residues that are most important for information flow, as well as preferred communication routes across the longitudinal interface. Substate-specific analysis indicates that the transition from the GTP- to GDP-bound state is associated with changes in the preferred allosteric routes, enhancing our understanding of longitudinal communication in tubulin. Beyond tubulin, our approach illustrates how combining dimensionality reduction with substate-specific network analysis can uncover hidden allosteric signatures in protein assemblies.

Riboswitches are dynamic RNA structures that regulate gene expression in response to ligand binding. The guanidine-II riboswitch consists of two hairpin loops, connected by a flexible linker, which can dimerize in the presence of a cognate ligand. Both processes, ligand binding and dimerization, are governed by the first three bases of each of the tetraloops, while the role of the extrahelical fourth base of the tetraloop, conserved to a purine, is less well understood. An influence of base identity in this position on ligand binding affinity and cooperativity has been shown in experiment, but no molecular-level explanation of this influence has yet been attempted. To investigate the interactions underlying this observation, we performed molecular dynamics simulations of the guanidine-II riboswitch with varied base identity of these extrahelical bases. The simulations show base-identity dependence in the formation of a recently described conformational change that involves the extrahelical bases and interlocks both hairpins. Tuning an established network-analysis-based approach to be sensitive to delicate changes in hydrogen bonding patterns enables us to classify the conformational states involved and show how base identity affects interlocking stability. These interactions may underpin the influence of the extrahelical bases on the structural equilibrium of the riboswitch and suggest a rationale for base-identity preference at this position.

Efficient analysis methods are needed that are able to cope with the huge size of modern molecular simulation datasets. Typical analysis workflows involve dimensionality reduction algorithms that improve the comprehensibility of the dataset by lowering its dimensions. Other common elements of the analysis are clustering algorithms that divide the dataset into groups according to a predefined characteristic, often with an emphasis on structural homogeneity. The application of such a clustering algorithm can help identify important metastable states of the system. In this paper, we revisit a clustering workflow described by Hunkler et al. [J. Chem. Phys. 158, 144109 (2023)] and provide a refined and automated version. We apply this workflow to a dataset of atomistic simulations of the 76 amino-acid residue protein ubiquitin (Ub) to illustrate its strengths and characteristics. The automated clustering workflow combines two dimensionality reduction algorithms, cc_analysis and EncoderMap, with the clustering algorithm HDBSCAN and a root mean square deviation-based sorting criterion. Due to an iterative approach, it is especially suitable for highly efficient categorization of large datasets of structures into structurally homogeneous clusters of different densities and sizes. Small adaptations to the original clustering workflow ensure considerable improvements in the quality and homogeneity of the obtained clusters.

ZUSAMMENFASSUNG: Aminoacyl‐tRNA‐Synthetasen vermitteln die Aktivierung und Übertragung von Aminosäuren auf ihre jeweilige tRNA und stellen damit eines der initialen Ereignisse der Proteinbiosynthese dar. Obwohl verschiedene Wirkmechanismen für die Katalyse dieser Enzyme vorgeschlagen wurden, bleibt ihr vollständiger katalytischer Zyklus bislang unvollständig verstanden. Hier verwendeten wir Elektronenspinresonanz‐Spektroskopie (EPR) in vitro und in Zellen in Kombination mit Molekulardynamik‐Simulationen, um die Rolle von Aminosäure‐Interaktionen im katalytischen Zyklus der Pyrrolysyl‐tRNA‐Synthetasen (PylRS) zu untersuchen, einem weit verbreiteten System zur Erweiterung des genetischen Codes. Experimente mit der paramagnetischen nicht‐kanonischen Aminosäure SLK1 zeigten das Vorhandensein und die Belegung sekundärer Aminosäure‐Bindungsstellen in PylRS, die an der Intermonomer‐Grenzfläche liegen und sich vom katalytischen Bindungsort unterscheiden. Auf Grundlage unserer Ergebnisse schlagen wir ein Modell vor, das einen alternierenden Wirkmodus der beiden PylRS‐Monomere für den katalytischen Zyklus von PylRS annimmt.

Aminoacyl‐tRNA synthetases mediate the activation and transfer of amino acids to their cognate tRNA, which constitutes one of the initial events in protein biosynthesis. Even though different mechanisms of action have been proposed for the catalysis of these enzymes, their entire catalytic cycle remains elusive. Here, we used electron paramagnetic resonance spectroscopy in vitro and in cells in combination with molecular dynamics simulations to study the role of amino acid interactions in the catalytic cycle of pyrrolysyl‐tRNA synthetases (PylRS), a widely used tool for genetic code expansion. Experiments using the paramagnetic non‐canonical amino acid SLK‐1 revealed the presence and occupation of secondary amino acid binding sites in PylRS located at the intermonomer interface, distant from the catalytic binding site. Based on our results, we propose a model that assumes an alternating mode of action of the two PylRS monomers for the catalytic cycle of PylRS.

Network methods and molecular dynamics (MD) simulations have become essential tools for studying protein dynamics. However, applying network methods to MD simulations of flexible proteins is a major challenge, since the high conformational heterogeneity in such multistate systems can lead to vastly different network topologies across an ensemble. To address this, tools that can disentangle conformational ensembles on a network level are needed. Here, we propose a graph-based clustering framework that provides state-specific insight into the residue interactions of flexible proteins. The framework hinges on using the set of graph-theoretic closeness centralities of all amino acid residues as a structural fingerprint and input for unsupervised machine learning algorithms to perform dimensionality reduction and clustering. The resulting clusters─states with shared network topology─are subsequently fed back into the upstream workflow and characterized at every representation level. Based on the example of FAT10─a protein with intrinsically disordered regions and two folded domains connected by a flexible linker─we demonstrate how this approach can be used to understand the protein's residue interactions on different, interconnected levels and to characterize its most populated states. Due to the modularity of the framework, it can be easily adapted, which makes it a suitable method to support network-based analyses of MD simulations for a wide variety of proteins.

EncoderMap is a dimensionality reduction method that is tailored for the analysis of molecular simulation data. It relies on a neural network autoencoder architecture augmented with an additional multidimensional scaling (MDS)-like loss term. Due to this additional cost function between the high-dimensional input and the latent space, EncoderMap emerges as a method that has advantages over other dimensionality reduction methods and straightforward autoencoders alike. In particular, the low-dimensional projections created by EncoderMap are guided by the MDS-like cost function adopted from the sketch-map algorithm, which produces projections with a better correlation between high- and low-dimensional similarities and dissimilarities. EncoderMap has been successfully applied to a wide range of molecular dynamics data. Here, we present a new version of the EncoderMap package. Porting EncoderMap to version 2 of the TensorFlow library not only ensures that it can be used on modern computers but most importantly allowed for the introduction of new features and a wide range of customization options (e.g., user-defined custom loss functions). We introduce new visualization capabilities and encoding/decoding features and make EncoderMap modular so that researchers can better understand the training process. Most of these additions can be used on general high-dimensional data, while some specifically aid in working with biomolecular simulation data. Furthermore, we added the possibility to provide sparse inputs to EncoderMap's autoencoder, which can be used to train the same network on topologically different proteins. We demonstrate these new features with the help of three different molecular dynamics data sets in the context of the ubiquitin system.

Network methods and molecular dynamics (MD) simulations have become essential tools for studying protein dynamics. However, applying network methods to MD simulations of flexible proteins is a major challenge, since the high conformational heterogeneity in such multi-state systems can lead to vastly different network topolo- gies across an ensemble. To address this, tools that can disentangle conformational ensembles on a network level are needed. Here, we propose a graph-based clustering framework that provides state-specific insight into the residue interactions of flexible proteins. The framework hinges on using the set of graph-theoretic closeness centralities of all amino acid residues as a structural fingerprint and input for unsupervised machine learning algorithms to perform dimensionality reduction and clustering. The resulting clusters - states with shared network topology - are subsequently fed back into the upstream workflow and characterized at every representation level. Based on the example of FAT10 - a protein with intrinsically disordered regions and two folded domains connected by a flexible linker - we demonstrate how this approach can be used to understand the protein’s residue interactions on different, interconnected levels and to characterize its most populated states. Due to the modularity of the framework, it can be easily adapted, which makes it a suitable method to support network-based analyses of MD simulations for a wide variety of proteins.

Riboswitches are structured ribonucleic acid (RNA) segments that act as specific sensors for small molecules in bacterial metabolism. Due to the flexible nature of these highly charged macromolecules, molecular dynamics simulations are instrumental to investigating the mechanistic details of their regulatory function. In the present study, the guanidine‐I riboswitch serves as an example of how atomistic simulations can shed light on the effect of ions on the structure and dynamics of RNA and on ligand binding. Relying on two orthologous crystal structures from different bacterial species, it is demonstrated how the ion setup crucially determines whether the simulation yields meaningful insights into the conformational stability of the RNA, functionally relevant residues and RNA‐ligand interactions. The ion setup in this context includes diffuse ions in solution and bound ions associated directly with the RNA, in particular a triad of 2 Mg²⁺ ions and a K⁺ ion in close proximity to the guanidinium binding site. A detailed investigation of the binding pocket reveals that the K⁺ from the ion triad plays a decisive role in stabilizing the ligand binding by stabilizing important localized interactions, which in turn contribute to the overall shape of the folded state of the RNA.

Riboswitches are structured RNA segments that act as specific sensors for small molecules in bacterial metabolism. Due to the flexi- ble nature of these highly charged macromolecules, molecular dynamics simulations are instrumental to investigating the mechanistic details of their regulatory function. In the present study, a guanidinium sensing riboswitch (the Gd-I riboswitch) serves as example how atomistic simulations can shed light on the role of ions on structure and dynamics of the RNA and on ligand binding. Rely- ing on two crystal structures from orthologous forms from different bacterial species, it is demonstrated how the ion setup crucially determines whether the simulation yields meaningful insights into the conformational stability of the RNA, functionally relevant residues and RNA-ligand interactions. Ion setup in this context includes the free ions in solution and ions associated directly with the RNA, in particular a triad of 2 Mg2+ ions and a K+ ion in close proximity to the guanidinium binding site. A detailed investi- gation of the binding pocket reveals that K+ from the ion triad plays a decisive role in stabilizing the ligand binding by stabilizing important localized interactions which in turn contribute to the overall shape of the folded state of the RNA.