GATSol, an enhanced predictor of protein solubility through the synergy of 3D structure graph and large language modeling.

BACKGROUND: Protein solubility is a critically important physicochemical property closely related to protein expression. For example, it is one of the main factors to be considered in the design and production of antibody drugs and a prerequisite for realizing various protein functions. Although several solubility prediction models have emerged in recent years, many of these models are limited to capturing information embedded in one-dimensional amino acid sequences, resulting in unsatisfactory predictive performance. RESULTS: In this study, we introduce a novel Graph Attention network-based protein Solubility model, GATSol, which represents the 3D structure of proteins as a protein graph. In addition to the node features of amino acids extracted by the state-of-the-art protein large language model, GATSol utilizes amino acid distance maps generated using the latest AlphaFold technology. Rigorous testing on independent eSOL and the Saccharomyces cerevisiae test datasets has shown that GATSol outperforms most recently introduced models, especially with respect to the coefficient of determination R2, which reaches 0.517 and 0.424, respectively. It outperforms the current state-of-the-art GraphSol by 18.4% on the S. cerevisiae_test set. CONCLUSIONS: GATSol captures 3D dimensional features of proteins by building protein graphs, which significantly improves the accuracy of protein solubility prediction. Recent advances in protein structure modeling allow our method to incorporate spatial structure features extracted from predicted structures into the model by relying only on the input of protein sequences, which simplifies the entire graph neural network prediction process, making it more user-friendly and efficient. As a result, GATSol may help prioritize highly soluble proteins, ultimately reducing the cost and effort of experimental work. The source code and data of the GATSol model are freely available at https://github.com/binbinbinv/GATSol .

Structures of wild-type and a constitutively closed mutant of connexin26 shed light on channel regulation by CO2.

Connexins allow intercellular communication by forming gap junction channels (GJCs) between juxtaposed cells. Connexin26 (Cx26) can be regulated directly by CO2. This is proposed to be mediated through carbamylation of K125. We show that mutating K125 to glutamate, mimicking the negative charge of carbamylation, causes Cx26 GJCs to be constitutively closed. Through cryo-EM we observe that the K125E mutation pushes a conformational equilibrium towards the channel having a constricted pore entrance, similar to effects seen on raising the partial pressure of CO2. In previous structures of connexins, the cytoplasmic loop, important in regulation and where K125 is located, is disordered. Through further cryo-EM studies we trap distinct states of Cx26 and observe density for the cytoplasmic loop. The interplay between the position of this loop, the conformations of the transmembrane helices and the position of the N-terminal helix, which controls the aperture to the pore, provides a mechanism for regulation.

Allosteric activation of VCP, an AAA unfoldase, by small molecule mimicry.

The loss of function of AAA (ATPases associated with diverse cellular activities) mechanoenzymes has been linked to diseases, and small molecules that activate these proteins can be powerful tools to probe mechanisms and test therapeutic hypotheses. Unlike chemical inhibitors that can bind a single conformational state to block enzyme function, activator binding must be permissive to different conformational states needed for mechanochemistry. However, we do not know how AAA proteins can be activated by small molecules. Here, we focus on valosin-containing protein (VCP)/p97, an AAA unfoldase whose loss of function has been linked to protein aggregation-based disorders, to identify druggable sites for chemical activators. We identified VCP ATPase Activator 1 (VAA1), a compound that dose-dependently stimulates VCP ATPase activity up to ~threefold. Our cryo-EM studies resulted in structures (ranging from ~2.9 to 3.7 Å-resolution) of VCP in apo and ADP-bound states and revealed that VAA1 binds an allosteric pocket near the C-terminus in both states. Engineered mutations in the VAA1-binding site confer resistance to VAA1, and furthermore, modulate VCP activity. Mutation of a phenylalanine residue in the VCP C-terminal tail that can occupy the VAA1 binding site also stimulates ATPase activity, suggesting that VAA1 acts by mimicking this interaction. Together, our findings uncover a druggable allosteric site and a mechanism of enzyme regulation that can be tuned through small molecule mimicry.

Inhibition and transport mechanisms of the ABC transporter hMRP5.

Human multidrug resistance protein 5 (hMRP5) effluxes anticancer and antivirus drugs, driving multidrug resistance. To uncover the mechanism of hMRP5, we determine six distinct cryo-EM structures, revealing an autoinhibitory N-terminal peptide that must dissociate to permit subsequent substrate recruitment. Guided by these molecular insights, we design an inhibitory peptide that could block substrate entry into the transport pathway. We also identify a regulatory motif, comprising a positively charged cluster and hydrophobic patches, within the first nucleotide-binding domain that modulates hMRP5 localization by engaging with membranes. By integrating our structural, biochemical, computational, and cell biological findings, we propose a model for hMRP5 conformational cycling and localization. Overall, this work provides mechanistic understanding of hMRP5 function, while informing future selective hMRP5 inhibitor development. More broadly, this study advances our understanding of the structural dynamics and inhibition of ABC transporters.

Folding the human proteome using BioNeMo: A fused dataset of structural models for machine learning purposes.

Human proteins are crucial players in both health and disease. Understanding their molecular landscape is a central topic in biological research. Here, we present an extensive dataset of predicted protein structures for 42,042 distinct human proteins, including splicing variants, derived from the UniProt reference proteome UP000005640. To ensure high quality and comparability, the dataset was generated by combining state-of-the-art modeling-tools AlphaFold 2, OpenFold, and ESMFold, provided within NVIDIA's BioNeMo platform, as well as homology modeling using Innophore's CavitomiX platform. Our dataset is offered in both unedited and edited formats for diverse research requirements. The unedited version contains structures as generated by the different prediction methods, whereas the edited version contains refinements, including a dataset of structures without low prediction-confidence regions and structures in complex with predicted ligands based on homologs in the PDB. We are confident that this dataset represents the most comprehensive collection of human protein structures available today, facilitating diverse applications such as structure-based drug design and the prediction of protein function and interactions.

Structural basis for expanded substrate specificities of human long chain acyl-CoA dehydrogenase and related acyl-CoA dehydrogenases.

Crystal structures of human long-chain acyl-CoA dehydrogenase (LCAD) and the catalytically inactive Glu291Gln mutant, have been determined. These structures suggest that LCAD harbors functions beyond its historically defined role in mitochondrial ß-oxidation of long and medium-chain fatty acids. LCAD is a homotetramer containing one FAD per 43 kDa subunit with Glu291 as the catalytic base. The substrate binding cavity of LCAD reveals key differences which makes it specific for longer and branched chain substrates. The presence of Pro132 near the start of the E helix leads to helix unwinding that, together with adjacent smaller residues, permits binding of bulky substrates such as 3α, 7α, l2α-trihydroxy-5ß-cholestan-26-oyl-CoA. This structural element is also utilized by ACAD11, a eucaryotic ACAD of unknown function, as well as bacterial ACADs known to metabolize sterol substrates. Sequence comparison suggests that ACAD10, another ACAD of unknown function, may also share this substrate specificity. These results suggest that LCAD, ACAD10, ACAD11 constitute a distinct class of eucaryotic acyl CoA dehydrogenases.

Asymmetric allostery in estrogen receptor-α homodimers drives responses to the ensemble of estrogens in the hormonal milieu.

The estrogen receptor-α (ER) is thought to function only as a homodimer but responds to a variety of environmental, metazoan, and therapeutic estrogens at subsaturating doses, supporting binding mixtures of ligands as well as dimers that are only partially occupied. Here, we present a series of flexible ER ligands that bind to receptor dimers with individual ligand poses favoring distinct receptor conformations-receptor conformational heterodimers-mimicking the binding of two different ligands. Molecular dynamics simulations showed that the pairs of different ligand poses changed the correlated motion across the dimer interface to generate asymmetric communication between the dimer interface, the ligands, and the surface binding sites for epigenetic regulatory proteins. By examining the binding of the same ligand in crystal structures of ER in the agonist vs. antagonist conformers, we also showed that these allosteric signals are bidirectional. The receptor conformer can drive different ligand binding modes to support agonist vs. antagonist activity profiles, a revision of ligand binding theory that has focused on unidirectional signaling from the ligand to the coregulator binding site. We also observed differences in the allosteric signals between ligand and coregulator binding sites in the monomeric vs. dimeric receptor, and when bound by two different ligands, states that are physiologically relevant. Thus, ER conformational heterodimers integrate two different ligand-regulated activity profiles, representing different modes for ligand-dependent regulation of ER activity.

Correlations of dynamic changes in lipid and protein of salted large yellow croaker during storage.

Protein and lipid are two major components that undergo significant changes during processing of aquatic products. This study focused on the protein oxidation, protein conformational states, lipid oxidation and lipid molecule profiling of salted large yellow croaker during storage, and their correlations were investigated. The degree of oxidation of protein and lipid was time-dependent, leading to an increase in carbonyl content and surface hydrophobicity, a decrease in sulfhydryl groups, and an increase in conjugated diene, peroxide value and thiobarbituric acid reactive substances value. Oxidation caused protein structure denaturation and aggregation during storage. Lipid composition and content changed dynamically, with polyunsaturated phosphatidylcholine (PC) was preferentially oxidized compared to polyunsaturated triacylglycerol. Correlation analysis showed that the degradation of polyunsaturated key differential lipids (PC 18:2_20:5, PC 16:0_22:6, PC 16:0_20:5, etc.) was closely related to the oxidation of protein and lipid. The changes in protein conformation and the peroxidation of polyunsaturated lipids mutually promote each other's oxidation process.

Treatment of flexibility of protein backbone in simulations of protein-ligand interactions using steered molecular dynamics.

To ensure that an external force can break the interaction between a protein and a ligand, the steered molecular dynamics simulation requires a harmonic restrained potential applied to the protein backbone. A usual practice is that all or a certain number of protein's heavy atoms or Cα atoms are fixed, being restrained by a small force. This present study reveals that while fixing both either all heavy atoms and or all Cα atoms is not a good approach, while fixing a too small number of few atoms sometimes cannot prevent the protein from rotating under the influence of the bulk water layer, and the pulled molecule may smack into the wall of the active site. We found that restraining the Cα atoms under certain conditions is more relevant. Thus, we would propose an alternative solution in which only the Cα atoms of the protein at a distance larger than 1.2 nm from the ligand are restrained. A more flexible, but not too flexible, protein will be expected to lead to a more natural release of the ligand.

[Syntheses and Structure-Activity Relationship Studies of Antitumor Bicyclic Hexapeptide RA-VII Analogues].

A series of antitumor bicyclic hexapeptide RA-VII analogues modified at residue 2, 3, or 6 were prepared by the chemical transformation of the hydroxy, methoxy, or carboxy groups or the aromatic rings of natural peptides RA-II, III, V, VII, and X. Analogues with modified side chains or peptide backbones, which cannot be prepared by the chemical transformation of their natural peptides, and newly isolated peptides from Rubia cordifolia roots were synthesized by using protected cycloisodityrosines prepared by the degradation of bis(thioamide) obtained from RA-VII or the diphenyl ether formation of boronodipeptide under the modified Chan-Lam coupling reaction conditions. Studies of the conformational features of the analogues and the newly isolated peptides and their relationships with cytotoxic activities against the HCT-116, HL-60, KATO-III, KB, L1210, MCF-7, and P-388 cell lines revealed the following: the methoxy group at residue 3 is essential for the potent cytotoxic activity; the methyl group at Ala-2 and Ala-4 but not at D-Ala-1 is required to establish the bioactive conformation; the N-methyl group at Tyr-5 is necessary for the peptides to adopt the active conformation preferentially; and the orientation of Tyr-5 and/or Tyr-6 phenyl rings has a significant effect on the cytotoxic activity.

Calcium-induced structural transitions are central to the folding, function, and processing of serratiopeptidase zymogen into mature form.

Serratia marcescens is an emerging health-threatening, gram-negative opportunistic pathogen associated with a wide variety of localized and life-threatening systemic infections. One of the most crucial virulence factors produced by S. marcescens is serratiopeptidase, a 50.2-kDa repeats-in-toxin (RTX) family broad-specificity zinc metalloprotease. RTX family proteins are functionally diverse exoproteins of gram-negative bacteria that exhibit calcium-dependent structural dynamicity and are secreted through a common type-1 secretion system (T1SS) machinery. To evaluate the impact of various divalent ligands on the folding and maturation of serratiopeptidase zymogen, the protein was purified and a series of structural and functional investigations were undertaken. The results indicate that calcium binding to the C-terminal RTX domain acts as a folding switch, triggering a disordered-to-ordered transition in the enzyme's conformation. Further, the auto-processing of the 16-amino acid N-terminal pro-peptide results in the maturation of the enzyme. The binding of calcium ions to serratiopeptidase causes a highly cooperative conformational transition in its structure, which is essential for the enzyme's activation and maturation. This conformational change is accompanied by an increase in solubility and enzymatic activity. For efficient secretion and to minimize intracellular toxicity, the enzyme needs to be in an unfolded extended form. The calcium-rich extracellular environment favors the folding and processing of zymogen into mature serratiopeptidase, i.e., the holo-form required by S. marcescens to establish infections and survive in different environmental niches.

eIF4G as a switch for heat shock mRNA translation.

The heat shock response is crucial for cell survival. In this issue of Molecular Cell, Desroches Altamirano et al.1 demonstrate that a temperature-induced conformational change in the translation initiation factor eIF4G is a key mechanism regulating translation during the heat shock response.

Freeprotmap: waiting-free prediction method for protein distance map.

BACKGROUND: Protein residue-residue distance maps are used for remote homology detection, protein information estimation, and protein structure research. However, existing prediction approaches are time-consuming, and hundreds of millions of proteins are discovered each year, necessitating the development of a rapid and reliable prediction method for protein residue-residue distances. Moreover, because many proteins lack known homologous sequences, a waiting-free and alignment-free deep learning method is needed. RESULT: In this study, we propose a learning framework named FreeProtMap. In terms of protein representation processing, the proposed group pooling in FreeProtMap effectively mitigates issues arising from high-dimensional sparseness in protein representation. In terms of model structure, we have made several careful designs. Firstly, it is designed based on the locality of protein structures and triangular inequality distance constraints to improve prediction accuracy. Secondly, inference speed is improved by using additive attention and lightweight design. Besides, the generalization ability is improved by using bottlenecks and a neural network block named local microformer. As a result, FreeProtMap can predict protein residue-residue distances in tens of milliseconds and has higher precision than the best structure prediction method. CONCLUSION: Several groups of comparative experiments and ablation experiments verify the effectiveness of the designs. The results demonstrate that FreeProtMap significantly outperforms other state-of-the-art methods in accurate protein residue-residue distance prediction, which is beneficial for lots of protein research works. It is worth mentioning that we could scan all proteins discovered each year based on FreeProtMap to find structurally similar proteins in a short time because the fact that the structure similarity calculation method based on distance maps is much less time-consuming than algorithms based on 3D structures.

Interplay of structural preorganization and conformational sampling in UDP-glucuronic acid 4-epimerase catalysis.

Understanding enzyme catalysis as connected to protein motions is a major challenge. Here, based on temperature kinetic studies combined with isotope effect measurements, we obtain energetic description of C-H activation in NAD-dependent UDP-glucuronic acid C4 epimerase. Approach from the ensemble-averaged ground state (GS) to the transition state-like reactive conformation (TSRC) involves, alongside uptake of heat ( Δ H ‡ = 54 kJ mol-1), significant loss in entropy ( - T Δ S ‡ = 20 kJ mol-1; 298 K) and negative activation heat capacity ( Δ C p ‡ = -0.64 kJ mol-1 K-1). Thermodynamic changes suggest the requirement for restricting configurational freedom at the GS to populate the TSRC. Enzyme variants affecting the electrostatic GS preorganization reveal active-site interactions important for precise TSRC sampling and H-transfer. Collectively, our study captures thermodynamic effects associated with TSRC sampling and establishes rigid positioning for C-H activation in an enzyme active site that requires conformational flexibility in fulfillment of its natural epimerase function.

GSScore: a novel Graphormer-based shell-like scoring method for protein-ligand docking.

Protein-ligand interactions (PLIs) are essential for cellular activities and drug discovery. But due to the complexity and high cost of experimental methods, there is a great demand for computational approaches to recognize PLI patterns, such as protein-ligand docking. In recent years, more and more models based on machine learning have been developed to directly predict the root mean square deviation (RMSD) of a ligand docking pose with reference to its native binding pose. However, new scoring methods are pressingly needed in methodology for more accurate RMSD prediction. We present a new deep learning-based scoring method for RMSD prediction of protein-ligand docking poses based on a Graphormer method and Shell-like graph architecture, named GSScore. To recognize near-native conformations from a set of poses, GSScore takes atoms as nodes and then establishes the docking interface of protein-ligand into multiple bipartite graphs within different shell ranges. Benefiting from the Graphormer and Shell-like graph architecture, GSScore can effectively capture the subtle differences between energetically favorable near-native conformations and unfavorable non-native poses without extra information. GSScore was extensively evaluated on diverse test sets including a subset of PDBBind version 2019, CASF2016 as well as DUD-E, and obtained significant improvements over existing methods in terms of RMSE, $R$ (Pearson correlation coefficient), Spearman correlation coefficient and Docking power.

HighFold: accurately predicting structures of cyclic peptides and complexes with head-to-tail and disulfide bridge constraints.

In recent years, cyclic peptides have emerged as a promising therapeutic modality due to their diverse biological activities. Understanding the structures of these cyclic peptides and their complexes is crucial for unlocking invaluable insights about protein target-cyclic peptide interaction, which can facilitate the development of novel-related drugs. However, conducting experimental observations is time-consuming and expensive. Computer-aided drug design methods are not practical enough in real-world applications. To tackles this challenge, we introduce HighFold, an AlphaFold-derived model in this study. By integrating specific details about the head-to-tail circle and disulfide bridge structures, the HighFold model can accurately predict the structures of cyclic peptides and their complexes. Our model demonstrates superior predictive performance compared to other existing approaches, representing a significant advancement in structure-activity research. The HighFold model is openly accessible at https://github.com/hongliangduan/HighFold.

Principles of peptide selection by the transporter associated with antigen processing.

Our ability to fight pathogens relies on major histocompatibility complex class I (MHC-I) molecules presenting diverse antigens on the surface of diseased cells. The transporter associated with antigen processing (TAP) transports nearly the entire repertoire of antigenic peptides into the endoplasmic reticulum for MHC-I loading. How TAP transports peptides specific for MHC-I is unclear. In this study, we used cryo-EM to determine a series of structures of human TAP, both in the absence and presence of peptides with various sequences and lengths. The structures revealed that peptides of eight or nine residues in length bind in a similarly extended conformation, despite having little sequence overlap. We also identified two peptide-anchoring pockets on either side of the transmembrane cavity, each engaging one end of a peptide with primarily main chain atoms. Occupation of both pockets results in a global conformational change in TAP, bringing the two halves of the transporter closer together to prime it for isomerization and ATP hydrolysis. Shorter peptides are able to bind to each pocket separately but are not long enough to bridge the cavity to bind to both simultaneously. Mutations that disrupt hydrogen bonds with the N and C termini of peptides almost abolish MHC-I surface expression. Our findings reveal that TAP functions as a molecular caliper that selects peptides according to length rather than sequence, providing antigen diversity for MHC-I presentation.

The Structural and Molecular Mechanisms of Mycobacterium tuberculosis Translational Elongation Factor Proteins.

Targeting translation factor proteins holds promise for developing innovative anti-tuberculosis drugs. During protein translation, many factors cause ribosomes to stall at messenger RNA (mRNA). To maintain protein homeostasis, bacteria have evolved various ribosome rescue mechanisms, including the predominant trans-translation process, to release stalled ribosomes and remove aberrant mRNAs. The rescue systems require the participation of translation elongation factor proteins (EFs) and are essential for bacterial physiology and reproduction. However, they disappear during eukaryotic evolution, which makes the essential proteins and translation elongation factors promising antimicrobial drug targets. Here, we review the structural and molecular mechanisms of the translation elongation factors EF-Tu, EF-Ts, and EF-G, which play essential roles in the normal translation and ribosome rescue mechanisms of Mycobacterium tuberculosis (Mtb). We also briefly describe the structure-based, computer-assisted study of anti-tuberculosis drugs.

Energy condensation and dipole alignment in protein dynamics.

The possibility that distant biomolecules in a cell interact via electromagnetic (e.m.) radiation was proposed many years ago to explain the high rate of encounters of partners in some enzymatic reactions. The results of two recent experiments designed to test the propensity of protein bovine serum albumin (BSA) to interact via e.m. radiation with other proteins were interpreted in a theoretical framework based on three main assumptions: (i) in order to experience this kind of interaction the protein must be in an out-of-equilibrium state; (ii) in this state there is a condensation of energy in low-frequency vibrational modes; and (iii) the hydration layers of water around the protein sustain the energy condensation. In the present paper we present the results of molecular dynamics simulations of BSA in four states: at equilibrium and out-of-equilibrium in water, and at room and high temperature in vacuum. By comparing physical properties of the system in the four states, our simulations provide a qualitative and quantitative assessment of the three assumptions on which the theoretical framework is based. Our results confirm the assumptions of the theoretical model showing energy condensation at low frequency and electretlike alignment between the protein's and the water's dipoles; they also allow a quantitative estimate of the contribution of the out-of-equilibrium state and of the water to the observed behavior of the protein. In particular, it has been found that in the out-of-equilibrium state the amplitude of the oscillation of the protein's dipole moment greatly increases, thereby enhancing a possible absorption or emission of e.m. radiation. The analysis of BSA's dynamics outlined in the present paper provides a procedure for checking the propensity of a biomolecule to interact via e.m. radiation with its biochemical partners.

Cryo2StructData: A Large Labeled Cryo-EM Density Map Dataset for AI-based Modeling of Protein Structures.

The advent of single-particle cryo-electron microscopy (cryo-EM) has brought forth a new era of structural biology, enabling the routine determination of large biological molecules and their complexes at atomic resolution. The high-resolution structures of biological macromolecules and their complexes significantly expedite biomedical research and drug discovery. However, automatically and accurately building atomic models from high-resolution cryo-EM density maps is still time-consuming and challenging when template-based models are unavailable. Artificial intelligence (AI) methods such as deep learning trained on limited amount of labeled cryo-EM density maps generate inaccurate atomic models. To address this issue, we created a dataset called Cryo2StructData consisting of 7,600 preprocessed cryo-EM density maps whose voxels are labelled according to their corresponding known atomic structures for training and testing AI methods to build atomic models from cryo-EM density maps. Cryo2StructData is larger than existing, publicly available datasets for training AI methods to build atomic protein structures from cryo-EM density maps. We trained and tested deep learning models on Cryo2StructData to validate its quality showing that it is ready for being used to train and test AI methods for building atomic models.