TopScore: Using Deep Neural Networks and Large Diverse Data Sets for Accurate Protein Model Quality Assessment

Identification of ligands binding to MB327-PAM-1, a binding pocket relevant for resensitization of nAChRs

Desensitization of nicotinic acetylcholine receptors (nAChRs) can be induced by overstimulation with acetylcholine (ACh) caused by an insufficient degradation of ACh after poisoning with organophosphorus compounds (OPCs). Currently, there is no generally applicable treatment for OPC poisoning that directly targets the desensitized nAChR. The bispyridinium compound MB327, an allosteric modulator of nAChR, has been shown to act as a resensitizer of nAChRs, indicating that drugs binding directly to nAChRs can have beneficial effects after OPC poisoning. However, MB327 also acts as an inhibitor of nAChRs at higher concentrations and can thus not be used for OPC poisoning treatment. Consequently, novel, more potent resensitizers are required. To successfully design novel ligands, the knowledge of the binding site is of utmost importance. Recently, we performed in silico studies to identify a new potential binding site of MB327, MB327-PAM-1, for which a more affine ligand, UNC0646, has been described. In this work, we performed ligand-based screening approaches to identify novel analogs of UNC0646 to help further understand the structure-affinity relationship of this compound class. Furthermore, we used structure-based screenings and identified compounds representing four new chemotypes binding to MB327-PAM-1. One of these compounds, cycloguanil, is the active metabolite of the antimalaria drug proguanil and shows a higher affinity towards MB327-PAM-1 than MB327. Furthermore, cycloguanil can reestablish the muscle force in soman-inhibited rat muscles. These results can act as a starting point to develop more potent resensitizers of nAChR and to close the gap in the treatment after OPC poisoning.

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning

Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper we first introduced an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), which learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, we selected interpolating data points in the learned latent space that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-β 1-42 (Aβ42) monomer, our deep learning model provided a comprehensive sampling of Aβ42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. This approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.

Protein engineering for feedback resistance in 3-deoxy-D-arabino-heptulosonate 7-phosphate synthase

Abstract

The shikimate pathway delivers aromatic amino acids (AAAs) in prokaryotes, fungi, and plants and is highly utilized in the industrial synthesis of bioactive compounds. Carbon flow into this pathway is controlled by the initial enzyme 3-deoxy-D-arabino-heptulosonate 7-phosphate synthase (DAHPS). AAAs produced further downstream, phenylalanine (Phe), tyrosine (Tyr), and tryptophan (Trp), regulate DAHPS by feedback inhibition. Corynebacterium glutamicum, the industrial workhorse for amino acid production, has two isoenzymes of DAHPS, AroF (Tyr sensitive) and AroG (Phe and Tyr sensitive). Here, we introduce feedback resistance against Tyr in the class I DAHPS AroF (AroF_cg). We pursued a consensus approach by drawing on structural modeling, sequence and structural comparisons, knowledge of feedback-resistant variants in E. coli homologs, and computed folding free energy changes. Two types of variants were predicted: Those where substitutions putatively either destabilize the inhibitor binding site or directly interfere with inhibitor binding. The recombinant variants were purified and assessed in enzyme activity assays in the presence or absence of Tyr. Of eight AroF_cg variants, two yielded > 80% (E154N) and > 50% (P155L) residual activity at 5 mM Tyr and showed > 50% specific activity of the wt AroF_cg in the absence of Tyr. Evaluation of two and four further variants at positions 154 and 155 yielded E154S, completely resistant to 5 mM Tyr, and P155I, which behaves similarly to P155L. Hence, feedback-resistant variants were found that are unlikely to evolve by point mutations from the parental gene and, thus, would be missed by classical strain engineering.

Key points

• We introduce feedback resistance against Tyr in the class I DAHPS AroF

• Variants at position 154 (155) yield > 80% (> 50%) residual activity at 5 mM Tyr

• The variants found are unlikely to evolve by point mutations from the parental gene

Supplementary Information

The online version contains supplementary material available at 10.1007/s00253-022-12166-9.

A MademoiseLLE domain binding platform links the key RNA transporter to endosomes

Spatiotemporal expression can be achieved by transport and translation of mRNAs at defined subcellular sites. An emerging mechanism mediating mRNA trafficking is microtubule-dependent co-transport on shuttling endosomes. Although progress has been made in identifying various components of the endosomal mRNA transport machinery, a mechanistic understanding of how these RNA-binding proteins are connected to endosomes is still lacking. Here, we demonstrate that a flexible MademoiseLLE (MLLE) domain platform within RNA-binding protein Rrm4 of Ustilago maydis is crucial for endosomal attachment. Our structure/function analysis uncovered three MLLE domains at the C-terminus of Rrm4 with a functionally defined hierarchy. MLLE3 recognises two PAM2-like sequences of the adaptor protein Upa1 and is essential for endosomal shuttling of Rrm4. MLLE1 and MLLE2 are most likely accessory domains exhibiting a variable binding mode for interaction with currently unknown partners. Thus, endosomal attachment of the mRNA transporter is orchestrated by a sophisticated MLLE domain binding platform.

Biosensor-based growth-coupling and spatial separation as an evolution strategy to improve small molecule production of Corynebacterium glutamicum

Evolutionary engineering is a powerful method to improve the performance of microbial cell factories, but can typically not be applied to enhance the production of chemicals due to the lack of an appropriate selection regime. We report here on a new strategy based on transcription factor-based biosensors, which directly couple production to growth. The growth of Corynebacterium glutamicum was coupled to the intracellular concentration of branched-chain amino acids, by integrating a synthetic circuit based on the Lrp biosensor upstream of two growth-regulating genes, pfkA and hisD. Modelling and experimental data highlight spatial separation as key strategy to limit the selection of 'cheater' strains that escaped the evolutionary pressure. This approach facilitated the isolation of strains featuring specific causal mutations enhancing amino acid production. We envision that this strategy can be applied with the plethora of known biosensors in various microbes, unlocking evolution as a feasible strategy to improve production of chemicals.

Computational Analyses of the AtTPC1 (Arabidopsis Two-Pore Channel 1) Permeation Pathway

Two Pore Channels (TPCs) are cation-selective voltage- and ligand-gated ion channels in membranes of intracellular organelles of eukaryotic cells. In plants, the TPC1 subtype forms the slowly activating vacuolar (SV) channel, the most dominant ion channel in the vacuolar membrane. Controversial reports about the permeability properties of plant SV channels fueled speculations about the physiological roles of this channel type. TPC1 is thought to have high Ca²⁺ permeability, a conclusion derived from relative permeability analyses using the Goldman–Hodgkin–Katz (GHK) equation. Here, we investigated in computational analyses the properties of the permeation pathway of TPC1 from Arabidopsis thaliana. Using the crystal structure of AtTPC1, protein modeling, molecular dynamics (MD) simulations, and free energy calculations, we identified a free energy minimum for Ca²⁺, but not for K⁺, at the luminal side next to the selectivity filter. Residues D269 and E637 coordinate in particular Ca²⁺ as demonstrated in in silico mutagenesis experiments. Such a Ca²⁺-specific coordination site in the pore explains contradicting data for the relative Ca²⁺/K⁺ permeability and strongly suggests that the Ca²⁺ permeability of SV channels is largely overestimated from relative permeability analyses. This conclusion was further supported by in silico electrophysiological studies showing a remarkable permeation of K⁺ but not Ca²⁺ through the open channel.

Development of a Biosensor Platform for Phenolic Compounds Using a Transition Ligand Strategy

The time-consuming and laborious characterization of protein or microbial strain designs limits the development of high-performance biocatalysts for biotechnological applications. Here, transcriptional biosensors emerged as valuable tools as they allow for rapid characterization of several thousand variants within a very short time. However, for many molecules of interest, no specific transcriptional regulator determining a biosensor's specificity is available. We present an approach for rapidly engineering biosensor specificities using a semirational transition ligand approach combined with fluorescence-activated cell sorting. In this two-step approach, a biosensor is first evolved toward a more relaxed-ligand specificity before using the resulting variant as the starting point in a second round of directed evolution toward high specificity for several chemically different ligands. By following this strategy, highly specific biosensors for 4-hydroxybenzoic acid, p-coumaric acid, 5-bromoferulic acid, and 6-methyl salicylic acid were developed, starting from a biosensor for the intracellular detection of trans-cinnamic acid.

F/G Region Rigidity is Inversely Correlated to Substrate Promiscuity of Human CYP Isoforms Involved in Metabolism

Of 57 human cytochrome P450 (CYP) enzymes, 12 metabolize 90% of xenobiotics. To our knowledge, no study has addressed the relation between enzyme dynamics and substrate promiscuity for more than three CYPs. Here, we show by constraint dilution simulations with the Constraint Network Analysis for the 12 isoforms that structural rigidity of the F/G region is significantly inversely correlated to the enzymes' substrate promiscuity. This highlights the functional importance of structural dynamics of the substrate tunnel.

Promiscuous Esterases Counterintuitively Are Less Flexible than Specific Ones

Understanding mechanisms of promiscuity is increasingly important from a fundamental and application point of view. As to enzyme structural dynamics, more promiscuous enzymes generally have been recognized to also be more flexible. However, examples for the opposite received much less attention. Here, we exploit comprehensive experimental information on the substrate promiscuity of 147 esterases tested against 96 esters together with computationally efficient rigidity analyses to understand the molecular origin of the observed promiscuity range. Unexpectedly, our data reveal that promiscuous esterases are significantly less flexible than specific ones, are significantly more thermostable, and have a significantly increased specific activity. These results may be reconciled with a model according to which structural flexibility in the case of specific esterases serves for conformational proofreading. Our results signify that an esterase sequence space can be screened by rigidity analyses for promiscuous esterases as starting points for further exploration in biotechnology and synthetic chemistry.

Foamy Viruses, Bet, and APOBEC3 Restriction

Non-human primates (NHP) are an important source of viruses that can spillover to humans and, after adaptation, spread through the host population. Whereas HIV-1 and HTLV-1 emerged as retroviral pathogens in humans, a unique class of retroviruses called foamy viruses (FV) with zoonotic potential are occasionally detected in bushmeat hunters or zookeepers. Various FVs are endemic in numerous mammalian natural hosts, such as primates, felines, bovines, and equines, and other animals, but not in humans. They are apathogenic, and significant differences exist between the viral life cycles of FV and other retroviruses. Importantly, FVs replicate in the presence of many well-defined retroviral restriction factors such as TRIM5α, BST2 (Tetherin), MX2, and APOBEC3 (A3). While the interaction of A3s with HIV-1 is well studied, the escape mechanisms of FVs from restriction by A3 is much less explored. Here we review the current knowledge of FV biology, host restriction factors, and FV-host interactions with an emphasis on the consequences of FV regulatory protein Bet binding to A3s and outline crucial open questions for future studies.

TopSuite Web Server: A Meta-Suite for Deep-Learning-Based Protein Structure and Quality Prediction

Proteins carry out the most fundamental processes of life such as cellular metabolism, regulation, and communication. Understanding these processes at a molecular level requires knowledge of their three-dimensional structures. Experimental techniques such as X-ray crystallography, NMR spectroscopy, and cryogenic electron microscopy can resolve protein structures but are costly and time-consuming and do not work for all proteins. Computational protein structure prediction tries to overcome these problems by predicting the structure of a new protein using existing protein structures as a resource. Here we present TopSuite, a web server for protein model quality assessment (TopScore) and template-based protein structure prediction (TopModel). TopScore provides meta-predictions for global and residue-wise model quality estimation using deep neural networks. TopModel predicts protein structures using a top-down consensus approach to aid the template selection and subsequently uses TopScore to refine and assess the predicted structures. The TopSuite Web server is freely available at https://cpclab.uni-duesseldorf.de/topsuite/.

Targeting spectrin redox switches to regulate the mechanoproperties of red blood cells

The mechanical properties of red blood cells (RBCs) are fundamental for their physiological role as gas transporters. RBC flexibility and elasticity allow them to survive the hemodynamic changes in the different regions of the vascular tree, to dynamically contribute to the flow thereby decreasing vascular resistance, and to deform during the passage through narrower vessels. RBC mechanoproperties are conferred mainly by the structural characteristics of their cytoskeleton, which consists predominantly of a spectrin scaffold connected to the membrane via nodes of actin, ankyrin and adducin. Changes in redox state and treatment with thiol-targeting molecules decrease the deformability of RBCs and affect the structure and stability of the spectrin cytoskeleton, indicating that the spectrin cytoskeleton may contain redox switches. In this perspective review, we revise current knowledge about the structural and functional characterization of spectrin cysteine redox switches and discuss the current lines of research aiming to understand the role of redox regulation on RBC mechanical properties. These studies may provide novel functional targets to modulate RBC function, blood viscosity and flow, and tissue perfusion in disease conditions.

Cell Type-Dependent Escape of Capsid Inhibitors by Simian Immunodeficiency Virus SIVcpz

Pandemic human immunodeficiency virus type 1 (HIV-1) is the result of the zoonotic transmission of simian immunodeficiency virus (SIV) from the chimpanzee subspecies Pan troglodytes troglodytes (SIVcpzPtt). The related subspecies Pan troglodytes schweinfurthii is the host of a similar virus, SIVcpzPts, which did not spread to humans. We tested these viruses with small-molecule capsid inhibitors (PF57, PF74, and GS-CA1) that interact with a binding groove in the capsid that is also used by CPSF6. While HIV-1 was sensitive to capsid inhibitors in cell lines, human macrophages, and peripheral blood mononuclear cells (PBMCs), SIVcpzPtt was resistant in rhesus FRhL-2 cells and human PBMCs but was sensitive to PF74 in human HOS and HeLa cells. SIVcpzPts was insensitive to PF74 in FRhL-2 cells, HeLa cells, PBMCs, and macrophages but was inhibited by PF74 in HOS cells. A truncated version of CPSF6 (CPSF6-358) inhibited SIVcpzPtt and HIV-1, while in contrast, SIVcpzPts was resistant to CPSF6-358. Homology modeling of HIV-1, SIVcpzPtt, and SIVcpzPts capsids and binding energy estimates suggest that these three viruses bind similarly to the host proteins cyclophilin A (CYPA) and CPSF6 as well as the capsid inhibitor PF74. Cyclosporine treatment, mutation of the CYPA-binding loop in the capsid, or CYPA knockout eliminated the resistance of SIVcpzPts to PF74 in HeLa cells. These experiments revealed that the antiviral capacity of PF74 is controlled by CYPA in a virus- and cell type-specific manner. Our data indicate that SIVcpz viruses can use infection pathways that escape the antiviral activity of PF74. We further suggest that the antiviral activity of PF74 capsid inhibitors depends on cellular cofactors.IMPORTANCE HIV-1 originated from SIVcpzPtt but not from the related virus SIVcpzPts, and thus, it is important to describe molecular infection by SIVcpzPts in human cells to understand the zoonosis of SIVs. Pharmacological HIV-1 capsid inhibitors (e.g., PF74) bind a capsid groove that is also a binding site for the cellular protein CPSF6. SIVcpzPts was resistant to PF74 in HeLa cells but sensitive in HOS cells, thus indicating cell line-specific resistance. Both SIVcpz viruses showed resistance to PF74 in human PBMCs. Modulating the presence of cyclophilin A or its binding to capsid in HeLa cells overcame SIVcpzPts resistance to PF74. These results indicate that early cytoplasmic infection events of SIVcpzPts may differ between cell types and affect, in an unknown manner, the antiviral activity of capsid inhibitors. Thus, capsid inhibitors depend on the activity or interaction of currently uncharacterized cellular factors.

QMEANDisCo-distance constraints applied on model quality estimation

Motivation

Methods that estimate the quality of a 3D protein structure model in absence of an experimental reference structure are crucial to determine a model’s utility and potential applications. Single model methods assess individual models whereas consensus methods require an ensemble of models as input. In this work, we extend the single model composite score QMEAN that employs statistical potentials of mean force and agreement terms by introducing a consensus-based distance constraint (DisCo) score.

Results

DisCo exploits distance distributions from experimentally determined protein structures that are homologous to the model being assessed. Feed-forward neural networks are trained to adaptively weigh contributions by the multi-template DisCo score and classical single model QMEAN parameters. The result is the composite score QMEANDisCo, which combines the accuracy of consensus methods with the broad applicability of single model approaches. We also demonstrate that, despite being the de-facto standard for structure prediction benchmarking, CASP models are not the ideal data source to train predictive methods for model quality estimation. For performance assessment, QMEANDisCo is continuously benchmarked within the CAMEO project and participated in CASP13. For both, it ranks among the top performers and excels with low response times.

Availability and implementation

QMEANDisCo is available as web-server at https://swissmodel.expasy.org/qmean. The source code can be downloaded from https://git.scicore.unibas.ch/schwede/QMEAN.

Supplementary information

Supplementary data are available at Bioinformatics online.

Characterization of the nucleotide-binding domain NsrF from the BceAB-type ABC-transporter NsrFP from the human pathogen Streptococcus agalactiae

Treatment of bacterial infections is a great challenge of our era due to the various resistance mechanisms against antibiotics. Antimicrobial peptides are considered to be potential novel compound as antibiotic treatment. However, some bacteria, especially many human pathogens, are inherently resistant to these compounds, due to the expression of BceAB-type ABC transporters. This rather new transporter family is not very well studied. Here, we report the first full characterization of the nucleotide binding domain of a BceAB type transporter from Streptococcus agalactiae, namely SaNsrF of the transporter SaNsrFP, which confers resistance against nisin and gallidermin. We determined the NTP hydrolysis kinetics and used molecular modeling and simulations in combination with small angle X-ray scattering to obtain structural models of the SaNsrF monomer and dimer. The fact that the SaNsrF_H202A variant displayed no ATPase activity was rationalized in terms of changes of the structural dynamics of the dimeric interface. Kinetic data show a clear preference for ATP as a substrate, and the prediction of binding modes allowed us to explain this selectivity over other NTPs.

Quantitative assessment of the determinant structural differences between redox-active and inactive glutaredoxins

Class I glutaredoxins are enzymatically active, glutathione-dependent oxidoreductases, whilst class II glutaredoxins are typically enzymatically inactive, Fe-S cluster-binding proteins. Enzymatically active glutaredoxins harbor both a glutathione-scaffold site for reacting with glutathionylated disulfide substrates and a glutathione-activator site for reacting with reduced glutathione. Here, using yeast ScGrx7 as a model protein, we comprehensively identified and characterized key residues from four distinct protein regions, as well as the covalently bound glutathione moiety, and quantified their contribution to both interaction sites. Additionally, we developed a redox-sensitive GFP2-based assay, which allowed the real-time assessment of glutaredoxin structure-function relationships inside living cells. Finally, we employed this assay to rapidly screen multiple glutaredoxin mutants, ultimately enabling us to convert enzymatically active and inactive glutaredoxins into each other. In summary, we have gained a comprehensive understanding of the mechanistic underpinnings of glutaredoxin catalysis and have elucidated the determinant structural differences between the two main classes of glutaredoxins.

Glutaredoxins play a central role in numerous biological processes including cellular redox homeostasis and Fe-S cluster biogenesis. Here the authors establish the molecular basis for glutaredoxin redox catalysis through comprehensive biochemical and structural analyses.

Structural and dynamic insights revealing how lipase binding domain MD1 of Pseudomonas aeruginosa foldase affects lipase activation

Folding and cellular localization of many proteins of Gram-negative bacteria rely on a network of chaperones and secretion systems. Among them is the lipase-specific foldase Lif, a membrane-bound steric chaperone that tightly binds (K_D = 29 nM) and mediates folding of the lipase LipA, a virulence factor of the pathogenic bacterium P. aeruginosa. Lif consists of five-domains, including a mini domain MD1 essential for LipA folding. However, the molecular mechanism of Lif-assisted LipA folding remains elusive. Here, we show in in vitro experiments using a soluble form of Lif (sLif) that isolated MD1 inhibits sLif-assisted LipA activation. Furthermore, the ability to activate LipA is lost in the variant sLif_Y99A, in which the evolutionary conserved amino acid Y99 from helix α1 of MD1 is mutated to alanine. This coincides with an approximately three-fold reduced affinity of the variant to LipA together with increased flexibility of sLif_Y99A in the complex as determined by polarization-resolved fluorescence spectroscopy. We have solved the NMR solution structures of P. aeruginosa MD1 and variant MD1_Y99A revealing a similar fold indicating that a structural modification is likely not the reason for the impaired activity of variant sLif_Y99A. Molecular dynamics simulations of the sLif:LipA complex in connection with rigidity analyses suggest a long-range network of interactions spanning from Y99 of sLif to the active site of LipA, which might be essential for LipA activation. These findings provide important details about the putative mechanism for LipA activation and point to a general mechanism of protein folding by multi-domain steric chaperones.

A simple neural network implementation of generalized solvation free energy for assessment of protein structural models

Rapid and accurate assessment of protein structural models is essential for protein structure prediction and design. Great progress has been made in this regard, especially by recent application of "knowledge-based" potentials. Various machine learning based protein structural model quality assessment methods are also quite successful. However, performance of traditional "physics-based" models has not been as effective. Based on our analysis of the fundamental computational limitation behind unsatisfactory performance of "physics-based" models, we propose a generalized solvation free energy (GSFE) framework, which is intrinsically flexible for multi-scale treatments and is amenable for machine learning implementation. Finally, we implemented a simple example of backbone-based residue level GSFE with neural network, which was found to have competitive performance when compared with highly complex latest "knowledge-based" atomic potentials in distinguishing native structures from decoys.

The Membrane-Integrated Steric Chaperone Lif Facilitates Active Site Opening of Pseudomonas aeruginosa Lipase A

Lipases are essential and widely used biocatalysts. Hence, the production of lipases requires a detailed understanding of the molecular mechanism of its folding and secretion. Lipase A from Pseudomonas aeruginosa, PaLipA, constitutes a prominent example that has additional relevance because of its role as a virulence factor in many diseases. PaLipA requires the assistance of a membrane-integrated steric chaperone, the lipase-specific foldase Lif, to achieve its enzymatically active state. However, the molecular mechanism of how Lif activates its cognate lipase has remained elusive. Here, we show by molecular dynamics simulations at the atomistic level and potential of mean force computations that Lif catalyzes the activation process of PaLipA by structurally stabilizing an intermediate PaLipA conformation, particularly a β-sheet in the region of residues 17-30, such that the opening of PaLipA's lid domain is facilitated. This opening allows substrate access to PaLipA's catalytic site. A surprising and so far not fully understood aspect of our study is that the open state of PaLipA is unstable compared to the closed one according to our computational and in vitro biochemical results. We thus speculate that further interactions of PaLipA with the Xcp secretion machinery and/or components of the extracellular matrix contribute to the remaining activity of secreted PaLipA. © 2019 Wiley Periodicals, Inc.