Discovery of Nanomolar Inhibitors for Human Dihydroorotate Dehydrogenase Using Structure-Based Drug Discovery Methods.

We used a structure-based drug discovery approach to identify novel inhibitors of human dihydroorotate dehydrogenase (DHODH), which is a therapeutic target for treating cancer and autoimmune and inflammatory diseases. In the case of acute myeloid leukemia, no previously discovered DHODH inhibitors have yet succeeded in this clinical application. Thus, there remains a strong need for new inhibitors that could be used as alternatives to the current standard-of-care. Our goal was to identify novel inhibitors of DHODH. We implemented prefiltering steps to omit PAINS and Lipinski violators at the earliest stages of this project. This enriched compounds in the data set that had a higher potential of favorable oral druggability. Guided by Glide SP docking scores, we found 20 structurally unique compounds from the ChemBridge EXPRESS-pick library that inhibited DHODH with IC50, DHODH values between 91 nM and 2.7 µM. Ten of these compounds reduced MOLM-13 cell viability with IC50, MOLM-13 values between 2.3 and 50.6 µM. Compound 16 (IC50, DHODH = 91 nM) inhibited DHODH more potently than the known DHODH inhibitor, teriflunomide (IC50, DHODH = 130 nM), during biochemical characterizations and presented a promising scaffold for future hit-to-lead optimization efforts. Compound 17 (IC50, MOLM-13 = 2.3 µM) was most successful at reducing survival in MOLM-13 cell lines compared with our other hits. The discovered compounds represent excellent starting points for the development and optimization of novel DHODH inhibitors.

Mechanisms of Pathogenicity of Hypertrophic Cardiomyopathy-Associated Troponin T (TNNT2) Variant R278C+/- During Development.

Hypertrophic cardiomyopathy (HCM) is one of the most common heritable cardiovascular diseases and variants of TNNT2 (cardiac troponin T) are linked to increased risk of sudden cardiac arrest despite causing limited hypertrophy. In this study, a TNNT2 variant, R278C+/-, was generated in both human cardiac recombinant/reconstituted thin filaments (hcRTF) and human- induced pluripotent stem cells (hiPSCs) to investigate the mechanisms by which the R278C+/- variant affects cardiomyocytes at the proteomic and functional levels. The results of proteomics analysis showed a significant upregulation of markers of cardiac hypertrophy and remodeling in R278C+/- vs. the isogenic control. Functional measurements showed that R278C+/- variant enhances the myofilament sensitivity to Ca2+, increases the kinetics of contraction, and causes arrhythmia at frequencies >75 bpm. This study uniquely shows the profound impact of the TNNT2 R278C+/- variant on the cardiomyocyte proteomic profile, cardiac electrical and contractile function in the early stages of cardiac development.

Predicting ion mobility collision cross sections using projection approximation with ROSIE-PARCS webserver.

Ion mobility coupled to mass spectrometry informs on the shape and size of protein structures in the form of a collision cross section (CCSIM). Although there are several computational methods for predicting CCSIM based on protein structures, including our previously developed projection approximation using rough circular shapes (PARCS), the process usually requires prior experience with the command-line interface. To overcome this challenge, here we present a web application on the Rosetta Online Server that Includes Everyone (ROSIE) webserver to predict CCSIM from protein structure using projection approximation with PARCS. In this web interface, the user is only required to provide one or more PDB files as input. Results from our case studies suggest that CCSIM predictions (with ROSIE-PARCS) are highly accurate with an average error of 6.12%. Furthermore, the absolute difference between CCSIM and CCSPARCS can help in distinguishing accurate from inaccurate AlphaFold2 protein structure predictions. ROSIE-PARCS is designed with a user-friendly interface, is available publicly and is free to use. The ROSIE-PARCS web interface is supported by all major web browsers and can be accessed via this link (https://rosie.graylab.jhu.edu).

Umbrella Sampling Simulations of Cardiac Thin Filament Reveal Thermodynamic Consequences of Troponin I Inhibitory Peptide Mutations.

The cardiac thin filament comprises F-actin, tropomyosin, and troponin (cTn). cTn is composed of three subunits: troponin C (cTnC), troponin I (cTnI), and troponin T (cTnT). To computationally study the effect of the thin filament on cTn activation events, we employed targeted molecular dynamics followed by umbrella sampling using a model of the thin filament to measure the thermodynamics of cTn transition events. Our simulations revealed that the thin filament causes an increase in the free energy required to open the cTnC hydrophobic patch and causes a more favorable interaction between this region and the cTnI switch peptide. Mutations to the cTn complex can lead to cardiomyopathy, a collection of diseases that present clinically with symptoms of hypertrophy or dilation of the cardiac muscle, leading to impairment of the heart's ability to function normally and ultimately myocardial infarction or heart failure. Upon introduction of cardiomyopathic mutations to R145 of cTnI, we observed a general decrease in the free energy of opening the cTnC hydrophobic patch, which is on par with previous experimental results. These mutations also exhibited a decrease in electrostatic interactions between cTnI-R145 and actin-E334. After introduction of a small molecule to the wild-type cTnI-actin interface to intentionally disrupt intersubunit contacts, we successfully observed similar thermodynamic consequences and disruptions to the same protein-protein contacts as observed with the cardiomyopathic mutations. Computational studies utilizing the cTn complex in isolation would have been unable to observe these effects, highlighting the importance of using a more physiologically relevant thin-filament model to investigate the global consequences of cardiomyopathic mutations to the cTn complex.

Insights into the catalytic mechanism of a bacterial deglycase essential for utilization of fructose-lysine.

Amadori rearrangement products are stable sugar-amino acid conjugates that are formed nonenzymatically during preparation, dehydration, and storage of foods. Because Amadori compounds such as fructose-lysine (F-Lys), an abundant constituent in processed foods, shape the animal gut microbiome, it is important to understand bacterial utilization of these fructosamines. In bacteria, F-Lys is first phosphorylated, either during or after uptake to the cytoplasm, to form 6-phosphofructose-lysine (6-P-F-Lys). FrlB, a deglycase, then converts 6-P-F-Lys to L-lysine and glucose-6-phosphate. Here, to elucidate the catalytic mechanism of this deglycase, we first obtained a 1.8-Å crystal structure of Salmonella FrlB (without substrate) and then used computational approaches to dock 6-P-F-Lys on this structure. We also took advantage of the structural similarity between FrlB and the sugar isomerase domain of Escherichia coli glucosamine-6-phosphate synthase (GlmS), a related enzyme for which a structure with substrate has been determined. An overlay of FrlB-6-P-F-Lys on GlmS-fructose-6-phosphate structures revealed parallels in their active-site arrangement and guided our selection of seven putative active-site residues in FrlB for site-directed mutagenesis. Activity assays with eight recombinant single-substitution mutants identified residues postulated to serve as the general acid and general base in the FrlB active site and indicated unexpectedly significant contributions from their proximal residues. By exploiting native mass spectrometry (MS) coupled to surface-induced dissociation, we distinguished mutations that impaired substrate binding versus cleavage. As demonstrated with FrlB, an integrated approach involving x-ray crystallography, in silico approaches, biochemical assays, and native MS can synergistically aid structure-function and mechanistic studies of enzymes.

Targeting Troponin C with Small Molecules Containing Diphenyl Moieties: Calcium Sensitivity Effects on Striated Muscles and Structure-Activity Relationship.

Despite large investments from academia and industry, heart failure, which results from a disruption of the contractile apparatus, remains a leading cause of death. Cardiac muscle contraction is a calcium-dependent mechanism, which is regulated by the troponin protein complex (cTn) and specifically by the N-terminal domain of its calcium-binding subunit (cNTnC). There is an increasing need for the development of small molecules that increase calcium sensitivity without altering the systolic calcium concentration, thereby strengthening the cardiac function. Here, we examined the effect of our previously identified calcium-sensitizing small molecule, ChemBridge compound 7930079, in the context of several homologous muscle systems. The effect of this molecule on force generation in isolated cardiac trabeculae and slow skeletal muscle fibers was measured. Furthermore, we explored the use of Gaussian accelerated molecular dynamics in sampling highly predictive receptor conformations based on NMR-derived starting structures. Additionally, we took a rational computational approach for lead optimization based on lipophilic diphenyl moieties. This integrated structural-biochemical-physiological approach led to the identification of three novel low-affinity binders, which had similar binding affinities to the known positive inotrope trifluoperazine. The most potent identified calcium sensitizer was compound 16 with an apparent affinity of 117 ± 17 µM.

Targeting Troponin C with Small Molecules Containing Diphenyl Moieties: Calcium Sensitivity Effects on Striated Muscle and Structure Activity Relationship.

Despite large investments from academia and industry, heart failure, which results from a disruption of the contractile apparatus, remains a leading cause of death. Cardiac muscle contraction is a calcium-dependent mechanism, which is regulated by the troponin protein complex (cTn) and specifically by the N-terminal domain of its calcium binding subunit (cNTnC). There is an increasing need for the development of small molecules that increase calcium sensitivity without altering systolic calcium concentration, thereby strengthening cardiac function. Here, we examined the effect of our previously identified calcium sensitizing small molecule, ChemBridge compound 7930079, in the context of several homologous muscle systems. The effect of this molecule on force generation in isolated cardiac trabeculae and slow skeletal muscle fibers was measured. Furthermore, we explored the use of Gaussian accelerated molecular dynamics in sampling highly predictive receptor conformations based on NMR derived starting structures. Additionally, we took a rational computational approach for lead optimization based on lipophilic diphenyl moieties. This led to the identification of three novel low affinity binders, which had similar binding affinities to known positive inotrope trifluoperazine. The most potent identified calcium sensitizer was compound 16 with an apparent affinity of 117 ± 17 µM .

Protein complex prediction using Rosetta, AlphaFold, and mass spectrometry covalent labeling.

Covalent labeling (CL) in combination with mass spectrometry can be used as an analytical tool to study and determine structural properties of protein-protein complexes. However, data from these experiments is sparse and does not unambiguously elucidate protein structure. Thus, computational algorithms are needed to deduce structure from the CL data. In this work, we present a hybrid method that combines models of protein complex subunits generated with AlphaFold with differential CL data via a CL-guided protein-protein docking in Rosetta. In a benchmark set, the RMSD (root-mean-square deviation) of the best-scoring models was below 3.6 Å for 5/5 complexes with inclusion of CL data, whereas the same quality was only achieved for 1/5 complexes without CL data. This study suggests that our integrated approach can successfully use data obtained from CL experiments to distinguish between nativelike and non-nativelike models.

Actives-Based Receptor Selection Strongly Increases the Success Rate in Structure-Based Drug Design and Leads to Identification of 22 Potent Cancer Inhibitors.

Computer-aided drug design, an important component of the early stages of the drug discovery pipeline, routinely identifies large numbers of false positive hits that are subsequently confirmed to be experimentally inactive compounds. We have developed a methodology to improve true positive prediction rates in structure-based drug design and have successfully applied the protocol to twenty target systems and identified the top three performing conformers for each of the targets. Receptor performance was evaluated based on the area under the curve of the receiver operating characteristic curve for two independent sets of known actives. For a subset of five diverse cancer-related disease targets, we validated our approach through experimental testing of the top 50 compounds from a blind screening of a small molecule library containing hundreds of thousands of compounds. Our methods of receptor and compound selection resulted in the identification of 22 novel inhibitors in the low µM-nM range, with the most potent being an EGFR inhibitor with an IC50 value of 7.96 nM. Additionally, for a subset of five independent target systems, we demonstrated the utility of Gaussian accelerated molecular dynamics to thoroughly explore a target system's potential energy surface and generate highly predictive receptor conformations.

Computational Exploration and Characterization of Potential Calcium Sensitizing Mutations in Cardiac Troponin C.

Calcium-dependent heart muscle contraction is regulated by the cardiac troponin protein complex (cTn) and specifically by the N-terminal domain of its calcium binding subunit (cNTnC). cNTnC contains one calcium binding site (site II), and altered calcium binding in this site has been studied for decades. It has been previously shown that cNTnC mutants, which increase calcium sensitization may have therapeutic benefits, such as restoring cardiac muscle contractility and functionality post-myocardial infarction events. Here, we computationally characterized eight mutations for their potential effects on calcium binding affinity in site II of cNTnC. We utilized two distinct methods to estimate calcium binding: adaptive steered molecular dynamics (ASMD) and thermodynamic integration (TI). We observed a sensitizing trend for all mutations based on the employed ASMD methodology. The TI results showed excellent agreement with experimentally known calcium binding affinities in wild-type cNTnC. Based on the TI results, five mutants were predicted to increase calcium sensitivity in site II. This study presents an interesting comparison of the two computational methods, which have both been shown to be valuable tools in characterizing the impacts of calcium sensitivity in mutant cNTnC systems.

Prediction of Intrinsic Disorder Using Rosetta ResidueDisorder and AlphaFold2.

The combination of deep learning and sequence data has transformed protein structure prediction and modeling, evidenced in the success of AlphaFold (AF). For this reason, many methods have been developed to take advantage of this success in areas where inaccurate structural modeling may limit computational predictiveness. For example, many methods have been developed to predict protein intrinsic disorder from sequence, including our Rosetta ResidueDisorder (RRD) approach. Intrinsically disordered regions in proteins are parts of the sequence that do not form ordered, folded structures under typical physiological conditions. In the original implementation of RRD, Rosetta ab initio models were generated, and disordered regions were predicted based on residue scores (disordered residues typically exist in regions of unfavorable scores). In this work, we show that by (i) replacing the ab initio modeling with AF (using the same scoring and disorder assignment approach) and (ii) updating the score function, the predictiveness improved significantly. Residues were better ranked by the order/disorder, evidenced by an improvement in receiver operating characteristic area-under-the-curve from 0.69 to 0.78 on a large (229 protein) and balanced data set (relatively even ordered versus disordered residues). Finally, the binary prediction accuracy also improved from 62% to 74% on the same data set. Our results show that the combined AF-RRD approach was as good as or better than all existing methods by these metrics (AF-RRD had the highest prediction accuracy).

Umbrella Sampling Simulations Measure Switch Peptide Binding and Hydrophobic Patch Opening Free Energies in Cardiac Troponin.

The cardiac troponin (cTn) complex is an important regulatory protein in heart contraction. Upon binding of Ca2+, cTn undergoes a conformational shift that allows the troponin I switch peptide (cTnISP) to be released from the actin filament and bind to the troponin C hydrophobic patch (cTnCHP). Mutations and modifications to this complex can change its sensitivity to Ca2+ and alter the energetics of the transition from the Ca2+-unbound, cTnISP-unbound form to the Ca2+-bound, cTnISP-bound form. We utilized targeted molecular dynamics (TMD) to obtain a trajectory of this transition pathway, followed by umbrella sampling to estimate the free energy associated with the cTnISP-cTnCHP binding and the cTnCHP opening events for wild-type (WT) cTn. We were able to reproduce experimental values for the cTnISP-cTnCHP binding event and obtain cTnCHP opening free energies in agreement with previous computational measurements of smaller cTnC systems. This excellent agreement for WT cTn demonstrated the strength of computational methods in studying the dynamics and energetics of the cTn complex. We then introduced mutations to the cTn complex that cause cardiomyopathy or alter its Ca2+ sensitivity and observed a general decrease in the free energy of opening the cTnCHP. For these same mutations, we observed no general trend in the effect on the cTnISP-cTnCHP binding event. Our method sets the stage for future computational studies on this system that predict the consequences of yet uncharacterized mutations on cTn dynamics and energetics.

Serendipitous Discovery of a Competitive Inhibitor of FraB, a Salmonella Deglycase and Drug Target.

Although salmonellosis, an infectious disease, is a significant global healthcare burden, there are no Salmonella-specific vaccines or therapeutics for humans. Motivated by our finding that FraB, a Salmonella deglycase responsible for fructose-asparagine catabolism, is a viable drug target, we initiated experimental and computational efforts to identify inhibitors of FraB. To this end, our recent high-throughput screening initiative yielded almost exclusively uncompetitive inhibitors of FraB. In parallel with this advance, we report here how a separate structural and computational biology investigation of FrlB, a FraB paralog, led to the serendipitous discovery that 2-deoxy-6-phosphogluconate is a competitive inhibitor of FraB (KI ~ 3 µM). However, this compound was ineffective in inhibiting the growth of Salmonella in a liquid culture. In addition to poor uptake, cellular metabolic transformations by a Salmonella dehydrogenase and different phosphatases likely undermined the efficacy of 2-deoxy-6-phosphogluconate in live-cell assays. These insights inform our ongoing efforts to synthesize non-hydrolyzable/-metabolizable analogs of 2-deoxy-6-phosphogluconate. We showcase our findings largely to (re)emphasize the role of serendipity and the importance of multi-pronged approaches in drug discovery.

The effect of Mg2+ on Ca2+ binding to cardiac troponin C in hypertrophic cardiomyopathy associated TNNC1 variants.

Cardiac troponin C (cTnC) is the critical Ca2+ -sensing component of the troponin complex. Binding of Ca2+ to cTnC triggers a cascade of conformational changes within the myofilament that culminate in force production. Hypertrophic cardiomyopathy (HCM)-associated TNNC1 variants generally induce a greater degree and duration of Ca2+ binding, which may underly the hypertrophic phenotype. Regulation of contraction has long been thought to occur exclusively through Ca2+ binding to site II of cTnC. However, work by several groups including ours suggest that Mg2+ , which is several orders of magnitude more abundant in the cell than Ca2+ , may compete for binding to the same cTnC regulatory site. We previously used isothermal titration calorimetry (ITC) to demonstrate that physiological concentrations of Mg2+ may decrease site II Ca2+ -binding in both N-terminal and full-length cTnC. Here, we explore the binding of Ca2+ and Mg2+ to cTnC harbouring a series of TNNC1 variants thought to be causal in HCM. ITC and thermodynamic integration (TI) simulations show that A8V, L29Q and A31S elevate the affinity for both Ca2+ and Mg2+ . Further, L48Q, Q50R and C84Y that are adjacent to the EF hand binding motif of site II have a more significant effect on affinity and the thermodynamics of the binding interaction. To the best of our knowledge, this work is the first to explore the role of Mg2+ in modifying the Ca2+ affinity of cTnC mutations linked to HCM. Our results indicate a physiologically significant role for cellular Mg2+ both at baseline and when elevated on modifying the Ca2+ binding properties of cTnC and the subsequent conformational changes which precede cardiac contraction.

Protein shape sampled by ion mobility mass spectrometry consistently improves protein structure prediction.

Ion mobility (IM) mass spectrometry provides structural information about protein shape and size in the form of an orientationally-averaged collision cross-section (CCSIM). While IM data have been used with various computational methods, they have not yet been utilized to predict monomeric protein structure from sequence. Here, we show that IM data can significantly improve protein structure determination using the modelling suite Rosetta. We develop the Rosetta Projection Approximation using Rough Circular Shapes (PARCS) algorithm that allows for fast and accurate prediction of CCSIM from structure. Following successful testing of the PARCS algorithm, we use an integrative modelling approach to utilize IM data for protein structure prediction. Additionally, we propose a confidence metric that identifies near native models in the absence of a known structure. The results of this study demonstrate the ability of IM data to consistently improve protein structure prediction.

Simulation of Energy-Resolved Mass Spectrometry Distributions from Surface-Induced Dissociation.

Understanding the relationship between protein structure and experimental data is crucial for utilizing experiments to solve biochemical problems and optimizing the use of sparse experimental data for structural interpretation. Tandem mass spectrometry (MS/MS) can be used with a variety of methods to collect structural data for proteins. One example is surface-induced dissociation (SID), which is used to break apart protein complexes (via a surface collision) into intact subcomplexes and can be performed at multiple laboratory frame SID collision energies. These energy-resolved MS/MS experiments have shown that the profile of the breakages depends on the acceleration energy of the collision. It is possible to extract an appearance energy (AE) from energy-resolved mass spectrometry (ERMS) data, which shows the relative intensity of each type of subcomplex as a function of SID acceleration energy. We previously determined that these AE values for specific interfaces correlated with structural features related to interface strength. In this study, we further examined the structural relationships by developing a method to predict the full ERMS plot from the structure, rather than extracting a single value. First, we noted that for proteins with multiple interface types, we could reproduce the correct shapes of breakdown curves, further confirming previous structural hypotheses. Next, we demonstrated that interface size and energy density (measured using Rosetta) correlated with data derived from the ERMS plot (R2 = 0.71). Furthermore, based on this trend, we used native crystal structures to predict ERMS. The majority of predictions resulted in good agreement, and the average root-mean-square error was 0.20 for the 20 complexes in our data set. We also show that if additional information on cleavage as a function of collision energy could be obtained, the accuracy of predictions improved further. Finally, we demonstrated that ERMS prediction results were better for the native than for inaccurate models in 17/20 cases. An application to run this simulation has been developed in Rosetta, which is freely available for use.

Validated determination of NRG1 Ig-like domain structure by mass spectrometry coupled with computational modeling.

High resolution hydroxyl radical protein footprinting (HR-HRPF) is a mass spectrometry-based method that measures the solvent exposure of multiple amino acids in a single experiment, offering constraints for experimentally informed computational modeling. HR-HRPF-based modeling has previously been used to accurately model the structure of proteins of known structure, but the technique has never been used to determine the structure of a protein of unknown structure. Here, we present the use of HR-HRPF-based modeling to determine the structure of the Ig-like domain of NRG1, a protein with no close homolog of known structure. Independent determination of the protein structure by both HR-HRPF-based modeling and heteronuclear NMR was carried out, with results compared only after both processes were complete. The HR-HRPF-based model was highly similar to the lowest energy NMR model, with a backbone RMSD of 1.6 Å. To our knowledge, this is the first use of HR-HRPF-based modeling to determine a previously uncharacterized protein structure.

Accounting for Neighboring Residue Hydrophobicity in Diethylpyrocarbonate Labeling Mass Spectrometry Improves Rosetta Protein Structure Prediction.

Covalent labeling mass spectrometry allows for protein structure elucidation via covalent modification and identification of exposed residues. Diethylpyrocarbonate (DEPC) is a commonly used covalent labeling reagent that provides insight into structure through the labeling of lysine, histidine, serine, threonine, and tyrosine residues. We recently implemented a Rosetta algorithm that used binary DEPC labeling data to improve protein structure prediction efforts. In this work, we improved on our modeling efforts by accounting for the level of hydrophobicity of neighboring residues in the microenvironment of serine, threonine, and tyrosine residues to obtain a more accurate estimate of the hydrophobic neighbor count. This was incorporated into Rosetta functionality, along with considerations for solvent-exposed histidine and lysine residues. Overall, our new Rosetta score term successfully identified best scoring models with less than 2 Å root-mean-squared deviations (RMSDs) for five of the seven benchmark proteins tested. We additionally developed a confidence metric to measure prediction success for situations in which a native structure is unavailable.

Applications of machine learning in computer-aided drug discovery.

Machine learning (ML) has revolutionised the field of structure-based drug design (SBDD) in recent years. During the training stage, ML techniques typically analyse large amounts of experimentally determined data to create predictive models in order to inform the drug discovery process. Deep learning (DL) is a subfield of ML, that relies on multiple layers of a neural network to extract significantly more complex patterns from experimental data, and has recently become a popular choice in SBDD. This review provides a thorough summary of the recent DL trends in SBDD with a particular focus on de novo drug design, binding site prediction, and binding affinity prediction of small molecules.

Protein structure prediction using residue-resolved protection factors from hydrogen-deuterium exchange NMR.

Hydrogen-deuterium exchange (HDX) measured by nuclear magnetic resonance (NMR) provides structural information for proteins relating to solvent accessibility and flexibility. While this structural information is beneficial, the data cannot be used exclusively to elucidate structures. However, the structural information provided by the HDX-NMR data can be supplemented by computational methods. In previous work, we developed an algorithm in Rosetta to predict structures using qualitative HDX-NMR data (categories of exchange rate). Here we expand on the effort, and utilize quantitative protection factors (PFs) from HDX-NMR for structure prediction. From observed correlations between PFs and solvent accessibility/flexibility measures, we present a scoring function to quantify the agreement with HDX data. Using a benchmark set of 10 proteins, an average improvement of 5.13 Å in root-mean-square deviation (RMSD) is observed for cases of inaccurate Rosetta predictions. Ultimately, seven out of 10 predictions are accurate without including HDX data, and nine out of 10 are accurate when using our PF-based HDX score.