Pruned Machine Learning Models to Predict Aqueous Solubility.

Solubility is a key metric for therapeutic compounds. Conversely, insoluble compounds cloud the accuracy of assays at all stages of chemical biology and drug discovery. Herein, we disclose naïve Bayesian classifier models to predict aqueous solubility. Publicly accessible aqueous solubility data were used to create two full, or nonpruned, training sets. These two sets were also combined to create a full fused set, and a training set comprised of a literature collation of solubility data was also considered as a reference. We tested different extents of data pruning on the training sets and constructed machine learning models that were evaluated with two independent, external test sets that contained compounds that were different from the training sets. The best pruned and fused model was significantly more accurate, in comparison to either the full model or the full fused model, with the prediction of these external test sets. By carefully removing data from the training set, less information can be used to create more accurate machine learning models for aqueous solubility. This knowledge and the curated training sets should prove useful to future machine learning approaches.

A Preclinical Candidate Targeting Mycobacterium tuberculosis KasA.

Published Mycobacterium tuberculosis ß-ketoacyl-ACP synthase KasA inhibitors lack sufficient potency and/or pharmacokinetic properties. A structure-based approach was used to optimize existing KasA inhibitor DG167. This afforded indazole JSF-3285 with a 30-fold increase in mouse plasma exposure. Biochemical, genetic, and X-ray studies confirmed JSF-3285 targets KasA. JSF-3285 offers substantial activity in an acute mouse model of infection and in the corresponding chronic infection model, with efficacious reductions in colony-forming units at doses as low as 5 mg/kg once daily orally and improvement of the efficacy of front-line drugs isoniazid or rifampicin. JSF-3285 is a promising preclinical candidate for tuberculosis.

Antitubercular Triazines: Optimization and Intrabacterial Metabolism.

The triazine antitubercular JSF-2019 was of interest due to its in vitro efficacy and the nitro group shared with the clinically relevant delamanid and pretomanid. JSF-2019 undergoes activation requiring F420H2 and one or more nitroreductases in addition to Ddn. An intrabacterial drug metabolism (IBDM) platform was leveraged to demonstrate the system kinetics, evidencing formation of NO⋅ and a des-nitro metabolite. Structure-activity relationship studies focused on improving the solubility and mouse pharmacokinetic profile of JSF-2019 and culminated in JSF-2513, relying on the key introduction of a morpholine. Mechanistic studies with JSF-2019, JSF-2513, and other triazines stressed the significance of achieving potent in vitro efficacy via release of intrabacterial NO⋅ along with inhibition of InhA and, more generally, the FAS-II pathway. This study highlights the importance of probing IBDM and its potential to clarify mechanism of action, which in this case is a combination of NO⋅ release and InhA inhibition.

Structural basis of DSF recognition by its receptor RpfR and its regulatory interaction with the DSF synthase RpfF.

The diffusible signal factors (DSFs) are a family of quorum-sensing autoinducers (AIs) produced and detected by numerous gram-negative bacteria. The DSF family AIs are fatty acids, differing in their acyl chain length, branching, and substitution but having in common a cis-2 double bond that is required for their activity. In both human and plant pathogens, DSFs regulate diverse phenotypes, including virulence factor expression, antibiotic resistance, and biofilm dispersal. Despite their widespread relevance to both human health and agriculture, the molecular basis of DSF recognition by their cellular receptors remained a mystery. Here, we report the first structure-function studies of the DSF receptor regulation of pathogenicity factor R (RpfR). We present the X-ray crystal structure of the RpfR DSF-binding domain in complex with the Burkholderia DSF (BDSF), which to our knowledge is the first structure of a DSF receptor in complex with its AI. To begin to understand the mechanistic role of the BDSF-RpfR contacts observed in the biologically important complex, we have also determined the X-ray crystal structure of the RpfR DSF-binding domain in complex with the inactive, saturated isomer of BDSF, dodecanoic acid (C12:0). In addition to these ligand-receptor complex structures, we report the discovery of a previously overlooked RpfR domain and show that it binds to and negatively regulates the DSF synthase regulation of pathogenicity factor F (RpfF). We have named this RpfR region the RpfF interaction (FI) domain, and we have determined its X-ray crystal structure alone and in complex with RpfF. These X-ray crystal structures, together with extensive complementary in vivo and in vitro functional studies, reveal the molecular basis of DSF recognition and the importance of the cis-2 double bond to DSF function. Finally, we show that throughout cellular growth, the production of BDSF by RpfF is post-translationally controlled by the RpfR N-terminal FI domain, affecting the cellular concentration of the bacterial second messenger bis-(3'-5')-cyclic dimeric guanosine monophosphate (c-di-GMP). Thus, in addition to describing the molecular basis for the binding and specificity of a DSF for its receptor, we describe a receptor-synthase interaction regulating bacterial quorum-sensing signaling and second messenger signal transduction.

Synergistic Lethality of a Binary Inhibitor of Mycobacterium tuberculosis KasA.

We report GSK3011724A (DG167) as a binary inhibitor of ß-ketoacyl-ACP synthase (KasA) in Mycobacterium tuberculosis Genetic and biochemical studies established KasA as the primary target. The X-ray crystal structure of the KasA-DG167 complex refined to 2.0-Å resolution revealed two interacting DG167 molecules occupying nonidentical sites in the substrate-binding channel of KasA. The binding affinities of KasA to DG167 and its analog, 5g, which binds only once in the substrate-binding channel, were determined, along with the KasA-5g X-ray crystal structure. DG167 strongly augmented the in vitro activity of isoniazid (INH), leading to synergistic lethality, and also synergized in an acute mouse model of M. tuberculosis infection. Synergistic lethality correlated with a unique transcriptional signature, including upregulation of oxidoreductases and downregulation of molecular chaperones. The lead structure-activity relationships (SAR), pharmacokinetic profile, and detailed interactions with the KasA protein that we describe may be applied to evolve a next-generation therapeutic strategy for tuberculosis (TB).IMPORTANCE Cell wall biosynthesis inhibitors have proven highly effective for treating tuberculosis (TB). We discovered and validated members of the indazole sulfonamide class of small molecules as inhibitors of Mycobacterium tuberculosis KasA-a key component for biosynthesis of the mycolic acid layer of the bacterium's cell wall and the same pathway as that inhibited by the first-line antitubercular drug isoniazid (INH). One lead compound, DG167, demonstrated synergistic lethality in combination with INH and a transcriptional pattern consistent with bactericidality and loss of persisters. Our results also detail a novel dual-binding mechanism for this compound as well as substantial structure-activity relationships (SAR) that may help in lead optimization activities. Together, these results suggest that KasA inhibition, specifically, that shown by the DG167 series, may be developed into a potent therapy that can synergize with existing antituberculars.

Novel Pyrimidines as Antitubercular Agents.

Mycobacterium tuberculosis infection is responsible for a global pandemic. New drugs are needed that do not show cross-resistance with the existing front-line therapeutics. A triazine antitubercular hit led to the design of a related pyrimidine family. The synthesis of a focused series of these analogs facilitated exploration of their in vitro activity, in vitro cytotoxicity, and physiochemical and absorption-distribution-metabolism-excretion properties. Select pyrimidines were then evaluated for their pharmacokinetic profiles in mice. The findings suggest a rationale for the further evolution of this promising series of antitubercular small molecules, which appear to share some similarities with the clinical compound PA-824 in terms of activation, while highlighting more general guidelines for the optimization of small-molecule antitubercular agents.

Human Microbiome Inspired Antibiotics with Improved ß-Lactam Synergy against MDR Staphylococcus aureus.

The flippase MurJ is responsible for transporting the cell wall intermediate lipid II from the cytoplasm to the outside of the cell. While essential for the survival of bacteria, it remains an underexploited target for antibacterial therapy. The humimycin antibiotics are lipid II flippase (MurJ) inhibitors that were synthesized on the basis of bioinformatic predictions derived from secondary metabolite gene clusters found in the human microbiome. Here, we describe an SAR campaign around humimycin A that produced humimycin 17S. Compared to humimycin A, 17S is a more potent ß-lactam potentiator, has a broader spectrum of activity, which now includes both methicillin resistant Staphylococcus aureus (MRSA) and vancomycin resistant Enterococcus faecalis (VRE), and did not lead to any detectable resistance when used in combination with a ß-lactam. Combinations of ß-lactam and humimycin 17S provide a potentially useful long-term MRSA regimen.

Antimicrobials Inspired by Nonribosomal Peptide Synthetase Gene Clusters.

Bacterial culture broth extracts have been the starting point for the development of numerous therapeutics. However, only a small fraction of bacterial biosynthetic diversity is accessible using this strategy. Here, we apply a discovery approach that bypasses the culturing step entirely by bioinformatically predicting small molecule structures from the primary sequences of the biosynthetic gene clusters. These structures are then chemically synthesized to give synthetic-bioinformatic natural products (syn-BNPs). Using this approach, we screened syn-BNPs inspired by nonribosomal peptide synthetases against microbial pathogens, and discovered an antibiotic for which no resistance could be identified and an antifungal agent with activity against diverse fungal pathogens.

Discovery of MRSA active antibiotics using primary sequence from the human microbiome.

Here we present a natural product discovery approach, whereby structures are bioinformatically predicted from primary sequence and produced by chemical synthesis (synthetic-bioinformatic natural products, syn-BNPs), circumventing the need for bacterial culture and gene expression. When we applied the approach to nonribosomal peptide synthetase gene clusters from human-associated bacteria, we identified the humimycins. These antibiotics inhibit lipid II flippase and potentiate ß-lactam activity against methicillin-resistant Staphylococcus aureus in mice, potentially providing a new treatment regimen.

Discovery of a small-molecule inhibitor and cellular probe of Keap1-Nrf2 protein-protein interaction.

A high-throughput screen (HTS) of the MLPCN library using a homogenous fluorescence polarization assay identified a small molecule as a first-in-class direct inhibitor of Keap1-Nrf2 protein-protein interaction. The HTS hit has three chiral centers; a combination of flash and chiral chromatographic separation demonstrated that Keap1-binding activity resides predominantly in one stereoisomer (SRS)-5 designated as ML334 (LH601A), which is at least 100× more potent than the other stereoisomers. The stereochemistry of the four cis isomers was assigned using X-ray crystallography and confirmed using stereospecific synthesis. (SRS)-5 is functionally active in both an ARE gene reporter assay and an Nrf2 nuclear translocation assay. The stereospecific nature of binding between (SRS)-5 and Keap1 as well as the preliminary but tractable structure-activity relationships support its use as a lead for our ongoing optimization.

Optimization of fluorescently labeled Nrf2 peptide probes and the development of a fluorescence polarization assay for the discovery of inhibitors of Keap1-Nrf2 interaction.

Activation of the antioxidant response element (ARE) upregulates enzymes involved in detoxification of electrophiles and reactive oxygen species. The induction of ARE genes is regulated by the interaction between redox sensor protein Keap1 and the transcription factor Nrf2. Fluorescently labeled Nrf2 peptides containing the ETGE motif were synthesized and optimized as tracers in the development of a fluorescence polarization (FP) assay to identify small-molecule inhibitors of the Keap1-Nrf2 interaction. The tracers were optimized to increase the dynamic range of the assay and their binding affinities to the Keap1 Kelch domain. The binding affinities of Nrf2 peptide inhibitors obtained in our FP assay using FITC-9mer Nrf2 peptide amide as the probe were in good agreement with those obtained previously by a surface plasmon resonance assay. The FP assay exhibits considerable tolerance toward DMSO and produced a Z' factor greater than 0.6 in a 384-well format. Further optimization of the probe led to cyanine-labeled 9mer Nrf2 peptide amide, which can be used along with the FITC-9mer Nrf2 peptide amide in a high-throughput screening assay to discover small-molecule inhibitors of Keap1-Nrf2 interaction.

Kinetic analyses of Keap1-Nrf2 interaction and determination of the minimal Nrf2 peptide sequence required for Keap1 binding using surface plasmon resonance.

The Keap1-Nrf2 interaction plays important roles in regulation of Nrf2 activity and induction of chemopreventive enzymes. To better understand the interaction and to determine the minimal Nrf2 sequence required for Keap1 binding, we synthesized a series of Nrf2 peptides containing ETGE motif and determined their binding affinities to the Kelch domain of Keap1 in solution using a surface plasmon resonance-based competition assay. The equilibrium dissociation constant for the interaction between 16mer Nrf2 peptide and Keap1 Kelch domain in solution (K(solution)(D)) was found to be 23.9 nM, which is 10× lower than the surface binding constant (K(surface)(D)) of 252 nM obtained for the direct binding of Keap1 Kelch domain to the immobilized 16mer Nrf2 peptide on a surface plasmon resonance sensor chip surface. The binding affinity of Nrf2 peptides to Keap1 Kelch domain was not lost until after deletion of eight residues from the N-terminus of the 16mer Nrf2 peptide. The 9mer Nrf2 peptide has a moderate binding affinity with a (K(solution)(D)) of 352 nM and the affinity was increased 15× upon removal of the positive charge at the peptide N-terminus by acetylation. These results suggest that the minimal Nrf2 peptide sequence required for Keap1 binding is the 9mer sequence of LDEETGEFL.

The intrinsic conformational propensities of the 20 naturally occurring amino acids and reflection of these propensities in proteins.

Here, we compare the distributions of main chain (Phi,Psi) angles (i.e., Ramachandran maps) of the 20 naturally occurring amino acids in three contexts: (i) molecular dynamics (MD) simulations of Gly-Gly-X-Gly-Gly pentapeptides in water at 298 K with exhaustive sampling, where X = the amino acid in question; (ii) 188 independent protein simulations in water at 298 K from our Dynameomics Project; and (iii) static crystal and NMR structures from the Protein Data Bank. The GGXGG peptide series is often used as a model of the unstructured denatured state of proteins. The sampling in the peptide MD simulations is neither random nor uniform. Instead, individual amino acids show preferences for particular conformations, but the peptide is dynamic, and interconversion between conformers is facile. For a given amino acid, the (Phi,Psi) distributions in the protein simulations and the Protein Data Bank are very similar and often distinct from those in the peptide simulations. Comparison between the peptide and protein simulations shows that packing constraints, solvation, and the tendency for particular amino acids to be used for specific structural motifs can overwhelm the "intrinsic propensities" of amino acids for particular (Phi,Psi) conformations. We also compare our helical propensities with experimental consensus values using the host-guest method, which appear to be determined largely by context and not necessarily the intrinsic conformational propensities of the guest residues. These simulations represent an improved coil library free from contextual effects to better model intrinsic conformational propensities and provide a detailed view of conformations making up the "random coil" state.