Finding Relevant Retrosynthetic Disconnections for Stereocontrolled Reactions.

Machine learning-driven computer-aided synthesis planning (CASP) tools have become important tools for idea generation in the design of complex molecule synthesis but do not adequately address the stereochemical features of the target compounds. A novel approach to automated extraction of templates used in CASP that includes stereochemical information included in the US Patent and Trademark Office (USPTO) and an internal AstraZeneca database containing reactions from Reaxys, Pistachio, and AstraZeneca electronic lab notebooks is implemented in the freely available AiZynthFinder software. Three hundred sixty-seven templates covering reagent- and substrate-controlled as well as stereospecific reactions were extracted from the USPTO, while 20,724 templates were from the AstraZeneca database. The performance of these templates in multistep CASP is evaluated for 936 targets from the ChEMBL database and an in-house selection of 791 AZ designs. The potential and limitations are discussed for four case studies from ChEMBL and examples of FDA-approved drugs.

Data-Efficient, Chemistry-Aware Machine Learning Predictions of Diels-Alder Reaction Outcomes.

The application of machine learning models to the prediction of reaction outcomes currently needs large and/or highly featurized data sets. We show that a chemistry-aware model, NERF, which mimics the bonding changes that occur during reactions, allows for highly accurate predictions of the outcomes of Diels-Alder reactions using a relatively small training set, with no pretraining and no additional features. We establish a diverse data set of 9537 intramolecular, hetero-, aromatic, and inverse electron demand Diels-Alder reactions. This data set is used to train a NERF model, and the performance is compared against state-of-the-art classification and generative machine learning models across low- and high-data regimes, with and without pretraining. The predictive accuracy (regio- and site selectivity in the major product) achieved by NERF exceeds 90% when as little as 40% of the data set is used for training. Another high-performing model, Chemformer, requires a larger training data set (>45%) and pretraining to reach 90% Top-1 accuracy. Accurate predictions of less-represented reaction subclasses, such as those involving heteroatomic or aromatic substrates, require higher percentages of training data. We also show how NERF can use small amounts of additional training data to quickly learn new systems and improve its overall understanding of reactivity. Synthetic chemists stand to benefit as this model can be rapidly expanded and tailored to areas of chemistry corresponding to the low-data regime.

Development of phenyl-urea-based small molecules that target penicillin-binding protein 4.

Staphylococcus aureus has the ability to invade cortical bone osteocyte lacuno-canalicular networks (OLCNs) and cause osteomyelitis. It was recently established that the cell wall transpeptidase, penicillin-binding protein 4 (PBP4), is crucial for this function, with pbp4 deletion strains unable to invade OLCNs and cause bone pathogenesis in a murine model of S. aureus osteomyelitis. Moreover, PBP4 has recently been found to modulate S. aureus resistance to ß-lactam antibiotics. As such, small molecule inhibitors of S. aureus PBP4 may represent dual functional antimicrobial agents that limit osteomyelitis and/or reverse antibiotic resistance. A high throughput screen recently revealed that the phenyl-urea 1 targets PBP4. Herein, we describe a structure-activity relationship (SAR) study on 1. Leveraging in silico docking and modeling, a set of analogs was synthesized and assessed for PBP4 inhibitory activities. Results revealed a preliminary SAR and identified lead compounds with enhanced binding to PBP4, more potent antibiotic resistance reversal, and diminished PBP4 cell wall transpeptidase activity in comparison to 1.

Computational and experimental identification of keystone interactions in Ebola virus matrix protein VP40 dimer formation.

The Ebola virus (EBOV) is a lipid-enveloped virus with a negative sense RNA genome that can cause severe and often fatal viral hemorrhagic fever. The assembly and budding of EBOV is regulated by the matrix protein, VP40, which is a peripheral protein that associates with anionic lipids at the inner leaflet of the plasma membrane. VP40 is sufficient to form virus-like particles (VLPs) from cells, which are nearly indistinguishable from authentic virions. Due to the restrictions of studying EBOV in BSL-4 facilities, VP40 has served as a surrogate in cellular studies to examine the EBOV assembly and budding process from the host cell plasma membrane. VP40 is a dimer where inhibition of dimer formation halts budding and formation of new VLPs as well as VP40 localization to the plasma membrane inner leaflet. To better understand VP40 dimer stability and critical amino acids to VP40 dimer formation, we integrated computational approaches with experimental validation. Site saturation/alanine scanning calculation, combined with molecular mechanics-based generalized Born with Poisson-Boltzmann surface area (MM-GB/PBSA) method and molecular dynamics simulations were used to predict the energetic contribution of amino acids to VP40 dimer stability and the hydrogen bonding network across the dimer interface. These studies revealed several previously unknown interactions and critical residues predicted to impact VP40 dimer formation. In vitro and cellular studies were then pursued for a subset of VP40 mutations demonstrating reduction in dimer formation (in vitro) or plasma membrane localization (in cells). Together, the computational and experimental approaches revealed critical residues for VP40 dimer stability in an alpha-helical interface (between residues 106-117) as well as in a loop region (between residues 52-61) below this alpha-helical region. This study sheds light on the structural origins of VP40 dimer formation and may inform the design of a small molecule that can disrupt VP40 dimer stability.

pH-dependent reaction triggering in PmHMGR crystals for time-resolved crystallography.

Serial crystallography and time-resolved data collection can readily be employed to investigate the catalytic mechanism of Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl (HMG)-coenzyme-A (CoA) reductase (PmHMGR) by changing the environmental conditions in the crystal and so manipulating the reaction rate. This enzyme uses a complex mechanism to convert mevalonate to HMG-CoA using the co-substrate CoA and cofactor NAD+. The multi-step reaction mechanism involves an exchange of bound NAD+ and large conformational changes by a 50-residue subdomain. The enzymatic reaction can be run in both forward and reverse directions in solution and is catalytically active in the crystal for multiple reaction steps. Initially, the enzyme was found to be inactive in the crystal starting with bound mevalonate, CoA, and NAD+. To observe the reaction from this direction, we examined the effects of crystallization buffer constituents and pH on enzyme turnover, discovering a strong inhibition in the crystallization buffer and a controllable increase in enzyme turnover as a function of pH. The inhibition is dependent on ionic concentration of the crystallization precipitant ammonium sulfate but independent of its ionic composition. Crystallographic studies show that the observed inhibition only affects the oxidation of mevalonate but not the subsequent reactions of the intermediate mevaldehyde. Calculations of the pKa values for the enzyme active site residues suggest that the effect of pH on turnover is due to the changing protonation state of His381. We have now exploited the changes in ionic inhibition in combination with the pH-dependent increase in turnover as a novel approach for triggering the PmHMGR reaction in crystals and capturing information about its intermediate states along the reaction pathway.

A Proline-Squaraine Ligand Framework (Pro-SqEB) for Stereoselective Rhodium(II)-Catalyzed Cyclopropanations.

A proline-squaraine ligand (Pro-SqEB) that demonstrates high levels of stereoselectivity in olefin cyclopropanations when anchored to a Rh2II scaffold is introduced. High yields and enantioselectivities were achieved in the cyclopropanation of alkenes with diazo compounds in the presence of Rh2(Pro-SqEB)4. Notably, the unique electronic and steric design of this catalyst enabled the use of polar solvents that are otherwise incompatible with most RhII complexes.

Elucidating Residue-Level Determinants Affecting Dimerization of Ebola Virus Matrix Protein Using High-Throughput Site Saturation Mutagenesis and Biophysical Approaches.

The Ebola virus (EBOV) is a filamentous virus that acquires its lipid envelope from the plasma membrane of the host cell it infects. EBOV assembly and budding from the host cell plasma membrane are mediated by a peripheral protein, known as the matrix protein VP40. VP40 is a 326 amino acid protein with two domains that are loosely linked. The VP40 N-terminal domain (NTD) contains a hydrophobic α-helix, which mediates VP40 dimerization. The VP40 C-terminal domain has a cationic patch, which mediates interactions with anionic lipids and a hydrophobic region that mediates VP40 dimer-dimer interactions. The VP40 dimer is necessary for trafficking to the plasma membrane inner leaflet and interactions with anionic lipids to mediate the VP40 assembly and oligomerization. Despite significant structural information available on the VP40 dimer structure, little is known on how the VP40 dimer is stabilized and how residues outside the NTD hydrophobic portion of the α-helical dimer interface contribute to dimer stability. To better understand how VP40 dimer stability is maintained, we performed computational studies using per-residue energy decomposition and site saturation mutagenesis. These studies revealed a number of novel keystone residues for VP40 dimer stability just adjacent to the α-helical dimer interface as well as distant residues in the VP40 CTD that can stabilize the VP40 dimer form. Experimental studies with representative VP40 mutants in vitro and in cells were performed to test computational predictions that reveal residues that alter VP40 dimer stability. Taken together, these studies provide important biophysical insights into VP40 dimerization and may be useful in strategies to weaken or alter the VP40 dimer structure as a means of inhibiting the EBOV assembly.

Computational Study of Base-Catalyzed Thiohemiacetal Decomposition in Pseudomonas mevalonii HMG-CoA Reductase.

Thiohemiacetals are key intermediates in the active sites of many enzymes catalyzing a variety of reactions. In the case of Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl coenzyme A reductase (PmHMGR), this intermediate connects the two hydride transfer steps where a thiohemiacetal is the product of the first hydride transfer and its breakdown forms the substrate of the second one, serving as the intermediate during cofactor exchange. Despite the many examples of thiohemiacetals in a variety of enzymatic reactions, there are few studies that detail their reactivity. Here, we present computational studies on the decomposition of the thiohemiacetal intermediate in PmHMGR using both QM-cluster and QM/MM models. This reaction mechanism involves a proton transfer from the substrate hydroxyl to an anionic Glu83 followed by a C-S bond elongation stabilized by a cationic His381. The reaction provides insight into the varying roles of the residues in the active site that favor this multistep mechanism.

On the use of real-world datasets for reaction yield prediction.

The lack of publicly available, large, and unbiased datasets is a key bottleneck for the application of machine learning (ML) methods in synthetic chemistry. Data from electronic laboratory notebooks (ELNs) could provide less biased, large datasets, but no such datasets have been made publicly available. The first real-world dataset from the ELNs of a large pharmaceutical company is disclosed and its relationship to high-throughput experimentation (HTE) datasets is described. For chemical yield predictions, a key task in chemical synthesis, an attributed graph neural network (AGNN) performs as well as or better than the best previous models on two HTE datasets for the Suzuki-Miyaura and Buchwald-Hartwig reactions. However, training the AGNN on an ELN dataset does not lead to a predictive model. The implications of using ELN data for training ML-based models are discussed in the context of yield predictions.

Targeting the chromatin effector Pygo2 promotes cytotoxic T cell responses and overcomes immunotherapy resistance in prostate cancer.

The noninflamed microenvironment in prostate cancer represents a barrier to immunotherapy. Genetic alterations underlying cancer cell-intrinsic oncogenic signaling are increasingly appreciated for their role in shaping the immune landscape. Recently, we identified Pygopus 2 (PYGO2) as the driver oncogene for the amplicon at 1q21.3 in prostate cancer. Here, using transgenic mouse models of metastatic prostate adenocarcinoma, we found that Pygo2 deletion decelerated tumor progression, diminished metastases, and extended survival. Pygo2 loss augmented the activation and infiltration of cytotoxic T lymphocytes (CTLs) and sensitized tumor cells to T cell killing. Mechanistically, Pygo2 orchestrated a p53/Sp1/Kit/Ido1 signaling network to foster a microenvironment hostile to CTLs. Genetic or pharmacological inhibition of Pygo2 enhanced the antitumor efficacy of immunotherapies using immune checkpoint blockade (ICB), adoptive cell transfer, or agents inhibiting myeloid-derived suppressor cells. In human prostate cancer samples, Pygo2 expression was inversely correlated with the infiltration of CD8+ T cells. Analysis of the ICB clinical data showed association between high PYGO2 level and worse outcome. Together, our results highlight a potential path to improve immunotherapy using Pygo2-targeted therapy for advanced prostate cancer.

A Quantum-Guided Molecular Mechanics Force Field for the Ferrocene Scaffold.

Ferrocene derivatives have a wide range of applications, including as ligands in asymmetric catalysis, due to their chemical stability, rigid backbone, steric bulk, and ability to encode stereochemical information via planar chirality. Unfortunately, few of the available molecular mechanics force fields incorporate parameters for the accurate study of this important building block. Here, we present a MM3* force field for ferrocenyl ligands, which was generated using the quantum-guided molecular mechanics (Q2MM) method. Detailed validation by comparison to DFT calculations and crystal structures demonstrates the accuracy of the parameters and uncovers the physical origin of deviations through excess energy analysis. Combining the ferrocene force field with a force field for Pd-allyl complexes and comparing the crystal structures shows the compatibility with previously developed MM3* force fields. Finally, the ferrocene force field was combined with a previously published transition-state force field to predict the stereochemical outcomes of the aminations of Pd-allyl complexes with different amines and different chiral ferrocenyl ligands, with an R2 of ∼0.91 over 10 examples.

Automated fitting of transition state force fields for biomolecular simulations.

The generation of surrogate potential energy functions (PEF) that are orders of magnitude faster to compute but as accurate as the underlying training data from high-level electronic structure methods is one of the most promising applications of fitting procedures in chemistry. In previous work, we have shown that transition state force fields (TSFFs), fitted to the functional form of MM3* force fields using the quantum guided molecular mechanics (Q2MM) method, provide an accurate description of transition states that can be used for stereoselectivity predictions of small molecule reactions. Here, we demonstrate the applicability of the method for fit TSFFs to the well-established Amber force field, which could be used for molecular dynamics studies of enzyme reaction. As a case study, the fitting of a TSFF to the second hydride transfer in Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl coenzyme A reductase (PmHMGR) is used. The differences and similarities to fitting of small molecule TSFFs are discussed.

Rational Design and Identification of Harmine-Inspired, N-Heterocyclic DYRK1A Inhibitors Employing a Functional Genomic In Vivo Drosophila Model System.

Deregulation of dual-specificity tyrosine phosphorylation-regulated kinase 1A (DYRK1A) plays a significant role in developmental brain defects, early-onset neurodegeneration, neuronal cell loss, dementia, and several types of cancer. Herein, we report the discovery of three new classes of N-heterocyclic DYRK1A inhibitors based on the potent, yet toxic kinase inhibitors, harmine and harmol. An initial in vitro evaluation of the small molecule library assembled revealed that the core heterocyclic motifs benzofuranones, oxindoles, and pyrrolones, showed statistically significant DYRK1A inhibition. Further, the utilization of a low cost, high-throughput functional genomic in vivo model system to identify small molecule inhibitors that normalize DYRK1A overexpression phenotypes is described. This in vivo assay substantiated the in vitro results, and the resulting correspondence validates generated classes as architectural motifs that serve as potential DYRK1A inhibitors. Further expansion and analysis of these core compound structures will allow discovery of safe, more effective chemical inhibitors of DYRK1A to ameliorate phenotypes caused by DYRK1A overexpression.

EP300 Selectively Controls the Enhancer Landscape of MYCN-Amplified Neuroblastoma.

Gene expression is regulated by promoters and enhancers marked by histone H3 lysine 27 acetylation (H3K27ac), which is established by the paralogous histone acetyltransferases (HAT) EP300 and CBP. These enzymes display overlapping regulatory roles in untransformed cells, but less characterized roles in cancer cells. We demonstrate that the majority of high-risk pediatric neuroblastoma (NB) depends on EP300, whereas CBP has a limited role. EP300 controls enhancer acetylation by interacting with TFAP2ß, a transcription factor member of the lineage-defining transcriptional core regulatory circuitry (CRC) in NB. To disrupt EP300, we developed a proteolysis-targeting chimera (PROTAC) compound termed "JQAD1" that selectively targets EP300 for degradation. JQAD1 treatment causes loss of H3K27ac at CRC enhancers and rapid NB apoptosis, with limited toxicity to untransformed cells where CBP may compensate. Furthermore, JQAD1 activity is critically determined by cereblon (CRBN) expression across NB cells. SIGNIFICANCE: EP300, but not CBP, controls oncogenic CRC-driven transcription in high-risk NB by binding TFAP2ß. We developed JQAD1, a CRBN-dependent PROTAC degrader with preferential activity against EP300 and demonstrated its activity in NB. JQAD1 has limited toxicity to untransformed cells and is effective in vivo in a CRBN-dependent manner. This article is highlighted in the In This Issue feature, p. 587.

Strategic elements in computer-assisted retrosynthesis: A case study of the pupukeanane natural products.

Computer-assisted synthesis planning represents a growing area of research, especially for complex molecule synthesis. Here, we present a case study involving the pupukeanane natural products, which are complex, marine-derived, natural products with unique tricyclic scaffolds. Proposed routes to members of each skeletal class informed by pathways generated using the program Synthia™ are compared to previous syntheses of these molecules. In addition, novel synthesis routes are proposed to pupukeanane congeners that have not been prepared previously.

Proofreading experimentally assigned stereochemistry through Q2MM predictions in Pd-catalyzed allylic aminations.

The palladium-catalyzed enantioselective allylic substitution by carbon or nitrogen nucleophiles is a key transformation that is particularly useful for the synthesis of bioactive compounds. Unfortunately, the selection of a suitable ligand/substrate combination often requires significant screening effort. Here, we show that a transition state force field (TSFF) derived by the quantum-guided molecular mechanics (Q2MM) method can be used to rapidly screen ligand/substrate combinations. Testing of this method on 77 literature reactions revealed several cases where the computationally predicted major enantiomer differed from the one reported. Interestingly, experimental follow-up led to a reassignment of the experimentally observed configuration. This result demonstrates the power of mechanistically based methods to predict and, where necessary, correct the stereochemical outcome.

HSP90 inhibitors reduce cholesterol storage in Niemann-Pick type C1 mutant fibroblasts.

Niemann-Pick type C1 (NPC1) disease is a lysosomal lipid storage disorder caused by mutations of the NPC1 gene. More than 300 disease-associated mutations are reported in patients, resulting in abnormal accumulation of unesterified cholesterol, glycosphingolipids, and other lipids in late endosomes and lysosomes (LE/Ly) of many cell types. Previously, we showed that treatment of many different NPC1 mutant fibroblasts with histone deacetylase inhibitors resulted in reduction of cholesterol storage, and we found that this was associated with enhanced exit of the NPC1 protein from the endoplasmic reticulum and delivery to LE/Ly. This suggested that histone deacetylase inhibitors may work through changes in protein chaperones to enhance the folding of NPC1 mutants, allowing them to be delivered to LE/Ly. In this study, we evaluated the effect of several HSP90 inhibitors on NPC1I1061T skin fibroblasts. We found that HSP90 inhibition resulted in clearance of cholesterol from LE/Ly, and this was associated with enhanced delivery of the mutant NPC1I1061T protein to LE/Ly. We also observed that inhibition of HSP90 increased the expression of HSP70, and overexpression of HSP70 also reduced cholesterol storage in NPC1I1061T fibroblasts. However, we did not see correction of cholesterol storage by arimoclomol, a drug that is reported to increase HSP70 expression, at doses up to 0.5 mM. The increase in other chaperones as a consequence of HSP90 improves folding of NPC1 protein and relieves cholesterol accumulation in NPC1 mutant fibroblasts.

Production of Proteins of the SARS-CoV-2 Proteome for Drug Discovery.

The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) is the causative agent of the coronavirus disease of 2019 (COVID-19). Its genome encodes two open reading frames for two large proteins, PP1a and PP1ab. Within the two polypeptide stretches, there are two proteases that process the large proteins into 15 discrete proteins essential for the assembly of the virion during its replication. We describe herein the cloning of the genes for these discrete proteins optimized for expression in Escherichia coli, production of the proteins, and their purification to homogeneity. These included all but six: NSP6, which possesses eight transmembrane regions, and five that are small proteins/peptides (E, ORF3b, ORF6, ORF7b, and ORF10). These proteins are intended for experimental validation of small-molecule binders as molecular template hits. The proof of concept was established with the ADP-ribosylhydrolase (ARH) domain of NSP3 in discovery of small-molecule templates that could serve as the basis for further optimization. The hit molecules include one submicromolar and a few low-micromolar binders to the ARH domain. Availability of these proteins in soluble forms opens up the opportunity for discoveries of novel templates with the potential for anti-COVID-19 pharmaceuticals.