AI-Accelerated Design of Targeted Covalent Inhibitors for SARS-CoV-2.

Direct-acting antivirals for the treatment of the COVID-19 pandemic caused by the SARS-CoV-2 virus are needed to complement vaccination efforts. Given the ongoing emergence of new variants, automated experimentation, and active learning based fast workflows for antiviral lead discovery remain critical to our ability to address the pandemic's evolution in a timely manner. While several such pipelines have been introduced to discover candidates with noncovalent interactions with the main protease (Mpro), here we developed a closed-loop artificial intelligence pipeline to design electrophilic warhead-based covalent candidates. This work introduces a deep learning-assisted automated computational workflow to introduce linkers and an electrophilic "warhead" to design covalent candidates and incorporates cutting-edge experimental techniques for validation. Using this process, promising candidates in the library were screened, and several potential hits were identified and tested experimentally using native mass spectrometry and fluorescence resonance energy transfer (FRET)-based screening assays. We identified four chloroacetamide-based covalent inhibitors of Mpro with micromolar affinities (KI of 5.27 µM) using our pipeline. Experimentally resolved binding modes for each compound were determined using room-temperature X-ray crystallography, which is consistent with the predicted poses. The induced conformational changes based on molecular dynamics simulations further suggest that the dynamics may be an important factor to further improve selectivity, thereby effectively lowering KI and reducing toxicity. These results demonstrate the utility of our modular and data-driven approach for potent and selective covalent inhibitor discovery and provide a platform to apply it to other emerging targets.

Integrated data-driven and experimental approaches to accelerate lead optimization targeting SARS-CoV-2 main protease.

Identification of potential therapeutic candidates can be expedited by integrating computational modeling with domain aware machine learning (ML) models followed by experimental validation in an iterative manner. Generative deep learning models can generate thousands of new candidates, however, their physiochemical and biochemical properties are typically not fully optimized. Using our recently developed deep learning models and a scaffold as a starting point, we generated tens of thousands of compounds for SARS-CoV-2 Mpro that preserve the core scaffold. We utilized and implemented several computational tools such as structural alert and toxicity analysis, high throughput virtual screening, ML-based 3D quantitative structure-activity relationships, multi-parameter optimization, and graph neural networks on generated candidates to predict biological activity and binding affinity in advance. As a result of these combined computational endeavors, eight promising candidates were singled out and put through experimental testing using Native Mass Spectrometry and FRET-based functional assays. Two of the tested compounds with quinazoline-2-thiol and acetylpiperidine core moieties showed IC[Formula: see text] values in the low micromolar range: [Formula: see text] [Formula: see text]M and 3.41±0.0015 [Formula: see text]M, respectively. Molecular dynamics simulations further highlight that binding of these compounds results in allosteric modulations within the chain B and the interface domains of the Mpro. Our integrated approach provides a platform for data driven lead optimization with rapid characterization and experimental validation in a closed loop that could be applied to other potential protein targets.

Factors underlying asymmetric pore dynamics of disaggregase and microtubule-severing AAA+ machines.

Disaggregation and microtubule-severing nanomachines from the AAA+ (ATPases associated with various cellular activities) superfamily assemble into ring-shaped hexamers that enable protein remodeling by coupling large-scale conformational changes with application of mechanical forces within a central pore by loops protruding within the pore. We probed the asymmetric pore motions and intraring interactions that support them by performing extensive molecular dynamics simulations of single-ring severing proteins and the double-ring disaggregase ClpB. Simulations reveal that dynamic stability of hexameric pores of severing proteins and of the nucleotide-binding domain 1 (NBD1) ring of ClpB, which belong to the same clade, involves a network of salt bridges that connect conserved motifs of central pore loops. Clustering analysis of ClpB highlights correlated motions of domains of neighboring protomers supporting strong interprotomer collaboration. Severing proteins have weaker interprotomer coupling and stronger intraprotomer stabilization through salt bridges involving pore loops. Distinct mechanisms are identified in the NBD2 ring of ClpB involving weaker interprotomer coupling through salt bridges formed by noncanonical loops and stronger intraprotomer coupling. Analysis of collective motions of PL1 loops indicates that the largest amplitude motions in the spiral complex of spastin and ClpB involve axial excursions of the loops, whereas for katanin they involve opening and closing of the central pore. All three motors execute primarily axial excursions in the ring complex. These results suggest distinct substrate processing mechanisms of remodeling and translocation by ClpB and spastin compared to katanin, thus providing dynamic support for the differential action of the two severing proteins. Relaxation dynamics of the distance between the PL1 loops and the center of mass of protomers reveals observation-time-dependent dynamics, leading to predicted relaxation times of tens to hundreds of microseconds on millisecond experimental timescales. For ClpB, the predicted relaxation time is in excellent agreement with the extracted time from smFRET experiments.

Toxicity Originating from Thiophene Containing Drugs: Exploring the Mechanism using Quantum Chemical Methods.

Drug metabolism of thiophene containing substrates by cytochrome P450s (CYP450) leads to toxic side effects, for example, nephrotoxicity (suprofen, ticlopidine), hepatotoxicity (tienilic acid), thrombotic thrombocytopenic purpura (clopidogrel), and aplastic anemia (ticlopidine). The origin of toxicity in these cases has been attributed to two different CYP450 mediated metabolic reactions: S-oxidation and epoxidation. In this work, the molecular level details of the bioinorganic chemistry associated with the generation of these competitive reactions are reported. Density functional theory was utilized (i) to explore the molecular mechanism for S-oxidation and epoxidation using the radical cationic center Cpd I [(iron(IV)-oxo-heme porphine system with SH(-) as the axial ligand, to mimic CYP450s] as the model oxidant, (ii) to establish the 3D structures of the reactants, transition states, and products on both the metabolic pathways, and (iii) to examine the potential energy (PE) profile for both the pathways to determine the energetically preferred toxic metabolite formation. The energy barrier required for S-oxidation was observed to be 14.75 kcal/mol as compared to that of the epoxidation reaction (13.23 kcal/mol) on the doublet PE surface of Cpd I. The formation of the epoxide metabolite was found to be highly exothermic (-23.24 kcal/mol), as compared to S-oxidation (-8.08 kcal/mol). Hence, on a relative scale the epoxidation process was observed to be thermodynamically and kinetically more favorable. The energy profiles associated with the reactions of the S-oxide and epoxide toxic metabolites were also explored. This study helps in understanding the CYP450-catalyzed toxic reactions of drugs containing the thiophene ring at the atomic level.

Native Mass Spectrometry Dissects the Structural Dynamics of an Allosteric Heterodimer of SARS-CoV-2 Nonstructural Proteins.

Structure-based drug design, which relies on precise understanding of the target protein and its interaction with the drug candidate, is dramatically expedited by advances in computational methods for candidate prediction. Yet, the accuracy needs to be improved with more structural data from high throughput experiments, which are challenging to generate, especially for dynamic and weak associations. Herein, we applied native mass spectrometry (native MS) to rapidly characterize ligand binding of an allosteric heterodimeric complex of SARS-CoV-2 nonstructural proteins (nsp) nsp10 and nsp16 (nsp10/16), a complex essential for virus survival in the host and thus a desirable drug target. Native MS showed that the dimer is in equilibrium with monomeric states in solution. Consistent with the literature, well characterized small cosubstrate, RNA substrate, and product bind with high specificity and affinity to the dimer but not the free monomers. Unsuccessfully designed ligands bind indiscriminately to all forms. Using neutral gas collision, the nsp16 monomer with bound cosubstrate can be released from the holo dimer complex, confirming the binding to nsp16 as revealed by the crystal structure. However, we observed an unusual migration of the endogenous zinc ions bound to nsp10 to nsp16 after collisional dissociation. The metal migration can be suppressed by using surface collision with reduced precursor charge states, which presumably resulted in minimal gas-phase structural rearrangement and highlighted the importance of complementary techniques. With minimal sample input (∼µg), native MS can rapidly detect ligand binding affinities and locations in dynamic multisubunit protein complexes, demonstrating the potential of an "all-in-one" native MS assay for rapid structural profiling of protein-to-AI-based compound systems to expedite drug discovery.

Microtubule Severing Enzymes Oligomerization and Allostery: A Tale of Two Domains.

Severing proteins are nanomachines from the AAA+ (ATPases associated with various cellular activities) superfamily whose function is to remodel the largest cellular filaments, microtubules. The standard AAA+ machines adopt hexameric ring structures for functional reasons, while being primarily monomeric in the absence of the nucleotide. Both major severing proteins, katanin and spastin, are believed to follow this trend. However, studies proposed that they populate lower-order oligomers in the presence of cofactors, which are functionally relevant. Our simulations show that the preferred oligomeric assembly is dependent on the binding partners and on the type of severing protein. Essential dynamics analysis predicts that the stability of an oligomer is dependent on the strength of the interface between the helical bundle domain (HBD) of a monomer and the convex face of the nucleotide binding domain (NBD) of a neighboring monomer. Hot spots analysis found that the region consisting of the HBD tip and the C-terminal (CT) helix is the only common element between the allosteric networks responding to nucleotide, substrate, and intermonomer binding. Clustering analysis indicates the existence of multiple pathways for the transition between the secondary structure of the HBD tip in monomers and the structure(s) it adopts in oligomers.

Exploring the Effect of Mechanical Anisotropy of Protein Structures in the Unfoldase Mechanism of AAA+ Molecular Machines.

Essential cellular processes of microtubule disassembly and protein degradation, which span lengths from tens of µm to nm, are mediated by specialized molecular machines with similar hexameric structure and function. Our molecular simulations at atomistic and coarse-grained scales show that both the microtubule-severing protein spastin and the caseinolytic protease ClpY, accomplish spectacular unfolding of their diverse substrates, a microtubule lattice and dihydrofolate reductase (DHFR), by taking advantage of mechanical anisotropy in these proteins. Unfolding of wild-type DHFR requires disruption of mechanically strong ß-sheet interfaces near each terminal, which yields branched pathways associated with unzipping along soft directions and shearing along strong directions. By contrast, unfolding of circular permutant DHFR variants involves single pathways due to softer mechanical interfaces near terminals, but translocation hindrance can arise from mechanical resistance of partially unfolded intermediates stabilized by ß-sheets. For spastin, optimal severing action initiated by pulling on a tubulin subunit is achieved through specific orientation of the machine versus the substrate (microtubule lattice). Moreover, changes in the strength of the interactions between spastin and a microtubule filament, which can be driven by the tubulin code, lead to drastically different outcomes for the integrity of the hexameric structure of the machine.

Modeling the Mechanical Response of Microtubule Lattices to Pressure.

Microtubules, the largest and stiffest filaments of the cytoskeleton, have to be well adapted to the high levels of crowdedness in cells to perform their multitude of functions. Furthermore, fundamental processes that involve microtubules, such as the maintenance of the cellular shape and cellular motion, are known to be highly dependent on external pressure. In light of the importance of pressure for the functioning of microtubules, numerous studies interrogated the response of these cytoskeletal filaments to osmotic pressure, resulting from crowding by osmolytes, such as poly(ethylene glycol)/poly(ethylene oxide) (PEG/PEO) molecules, or to direct applied pressure. The interpretation of experiments is usually based on the assumptions that PEG molecules have unfavorable interactions with the microtubule lattices and that the behavior of microtubules under pressure can be described by using continuous models. We probed directly these two assumptions. First, we characterized the interaction between the main interfaces in a microtubule filament and PEG molecules of various sizes using a combination of docking and molecular dynamics simulations. Second, we studied the response of a microtubule filament to compression using a coarse-grained model that allows for the breaking of lattice interfaces. Our results show that medium length PEG molecules do not alter the energetics of the lateral interfaces in microtubules but rather target and can penetrate into the voids between tubulin monomers at these interfaces, which can lead to a rapid loss of lateral interfaces under pressure. Compression of a microtubule under conditions corresponding to high osmotic pressure results in the formation of the deformed phase found in experiments. Our simulations show that the breaking of lateral interfaces, rather than the buckling of the filament inferred from the continuous models, accounts for the deformation.

Molecular investigations into the unfoldase action of severing enzymes on microtubules.

Microtubule (MT)-associated proteins regulate the dynamic behavior of MTs during cellular processes. MT severing enzymes are the associated proteins which destabilize MTs by removing subunits from the lattice. One model for how severing enzymes remove tubulin dimers from the MT lattice is by unfolding its subunits through pulling on the carboxy-terminal tails of tubulin dimers. This model stems from the fact that severing enzymes are AAA+ unfoldases. To test this mechanism, we apply pulling forces on the carboxy-terminal regions of MT subunits using coarse grained molecular simulations. In our simulations, we used different MT lattices and concentrations of severing enzymes. We compare our simulation results with data from in vitro severing assays and find that the experimental data is best fit by a model of cooperative removal of protofilament fragments by severing enzymes, which depends on the severing enzyme concentration and placement on the MT lattice.

Structure based virtual screening to identify selective phosphodiesterase 4B inhibitors.

Phosphodiesterase 4 (PDE4), is a hydrolytic enzyme, is proposed as a promising target in asthma and chronic obstructive pulmonary disease. PDE4B selective inhibitors are desirable to reduce the dose limiting adverse effect associated with non-selective PDE4B inhibitors. To achieve this goal, ligand based pharmacophore modeling and molecular docking approach is employed. Pharmacophore hypotheses for PDE4B and PDE4D are generated using HypoGen algorithm. The best PDE4B pharmacophore hypothesis (Hypo1_PDE4B) consist of one hydrogen-bond acceptor and two ring aromatic features, whereas PDE4D pharmacophore hypothesis (Hypo1_PDE4D) consist of one hydrogen-bond acceptor, one hydrophobic aliphatic, and two ring aromatic features. The validated pharmacophore hypotheses are used in virtual screening to identify selective PDE4B inhibitors. The hits were screened for their estimated activity, FitValue, and quantitative estimation of drug likeness. After molecular docking analysis, ten hits were purchased for in vitro analysis. Out of these, six hits have shown potent and selective inhibitory activity against PDE4B with IC50 values ranging from 2 to 378nM.