Discovery of a structural class of antibiotics with explainable deep learning.

The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1-9. Deep learning approaches have aided in exploring chemical spaces1,10-15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of 39,312 compounds and applied ensembles of graph neural networks to predict antibiotic activity and cytotoxicity for 12,076,365 compounds. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.

Virtual screening for small-molecule pathway regulators by image-profile matching.

Identifying the chemical regulators of biological pathways is a time-consuming bottleneck in developing therapeutics and research compounds. Typically, thousands to millions of candidate small molecules are tested in target-based biochemical screens or phenotypic cell-based screens, both expensive experiments customized to each disease. Here, our uncustomized, virtual, profile-based screening approach instead identifies compounds that match to pathways based on the phenotypic information in public cell image data, created using the Cell Painting assay. Our straightforward correlation-based computational strategy retrospectively uncovered the expected, known small-molecule regulators for 32% of positive-control gene queries. In prospective, discovery mode, we efficiently identified new compounds related to three query genes and validated them in subsequent gene-relevant assays, including compounds that phenocopy or pheno-oppose YAP1 overexpression and kill a Yap1-dependent sarcoma cell line. This image-profile-based approach could replace many customized labor- and resource-intensive screens and accelerate the discovery of biologically and therapeutically useful compounds.

Identification of potent inhibitors of SARS-CoV-2 infection by combined pharmacological evaluation and cellular network prioritization.

Pharmacologically active compounds with known biological targets were evaluated for inhibition of SARS-CoV-2 infection in cell and tissue models to help identify potent classes of active small molecules and to better understand host-virus interactions. We evaluated 6,710 clinical and preclinical compounds targeting 2,183 host proteins by immunocytofluorescence-based screening to identify SARS-CoV-2 infection inhibitors. Computationally integrating relationships between small molecule structure, dose-response antiviral activity, host target, and cell interactome produced cellular networks important for infection. This analysis revealed 389 small molecules with micromolar to low nanomolar activities, representing >12 scaffold classes and 813 host targets. Representatives were evaluated for mechanism of action in stable and primary human cell models with SARS-CoV-2 variants and MERS-CoV. One promising candidate, obatoclax, significantly reduced SARS-CoV-2 viral lung load in mice. Ultimately, this work establishes a rigorous approach for future pharmacological and computational identification of host factor dependencies and treatments for viral diseases.

Novel Fluorescence-Based High-Throughput FLIPR Assay Utilizing Membrane-Tethered Genetic Calcium Sensors to Identify T-Type Calcium Channel Modulators.

T-type voltage-gated Ca2+ channels have been implicated in many human disorders, and there has been increasing interest in developing highly selective and potent T-type Ca2+ channel modulators for potential clinical use. However, the unique biophysical properties of T-type Ca2+ channels are not conducive for developing high-throughput screening (HTS) assays to identify modulators, particularly potentiators. To illustrate, T-type Ca2+ channels are largely inactivated and unable to open to allow Ca2+ influx at -25 mV, the typical resting membrane potential of the cell lines commonly used in cellular screening assays. To address this issue, we developed cell lines that express Kir2.3 channels to hyperpolarize the membrane potential to -70 mV, thus allowing T-type channels to return to their resting state where they can be subsequently activated by membrane depolarization in the presence of extracellular KCl. Furthermore, to simplify the HTS assay and to reduce reagent cost, we stably expressed a membrane-tethered genetic calcium sensor, GCaMP6s-CAAX, that displays superior signal to the background compared to the untethered GCaMP6s or the synthetic Ca2+ sensor Fluo-4AM. Here, we describe a novel GCaMP6s-CAAX-based calcium assay utilizing a high-throughput fluorometric imaging plate reader (Molecular Devices, Sunnyvale, CA) format that can identify both activators and inhibitors of T-type Ca2+ channels. Lastly, we demonstrate the utility of this novel fluorescence-based assay to evaluate the activities of two distinct G-protein-coupled receptors, thus expanding the use of GCaMP6s-CAAX to a wide range of applications relevant for developing cellular assays in drug discovery.

Identification of druggable host targets needed for SARS-CoV-2 infection by combined pharmacological evaluation and cellular network directed prioritization both in vitro and in vivo.

Identification of host factors contributing to replication of viruses and resulting disease progression remains a promising approach for development of new therapeutics. Here, we evaluated 6710 clinical and preclinical compounds targeting 2183 host proteins by immunocytofluorescence-based screening to identify SARS-CoV-2 infection inhibitors. Computationally integrating relationships between small molecule structure, dose-response antiviral activity, host target and cell interactome networking produced cellular networks important for infection. This analysis revealed 389 small molecules, >12 scaffold classes and 813 host targets with micromolar to low nanomolar activities. From these classes, representatives were extensively evaluated for mechanism of action in stable and primary human cell models, and additionally against Beta and Delta SARS-CoV-2 variants and MERS-CoV. One promising candidate, obatoclax, significantly reduced SARS-CoV-2 viral lung load in mice. Ultimately, this work establishes a rigorous approach for future pharmacological and computational identification of novel host factor dependencies and treatments for viral diseases.

Discovery of a First-in-Class Inhibitor of the PRMT5-Substrate Adaptor Interaction.

PRMT5 and its substrate adaptor proteins (SAPs), pICln and Riok1, are synthetic lethal dependencies in MTAP-deleted cancer cells. SAPs share a conserved PRMT5 binding motif (PBM) which mediates binding to a surface of PRMT5 distal to the catalytic site. This interaction is required for methylation of several PRMT5 substrates, including histone and spliceosome complexes. We screened for small molecule inhibitors of the PRMT5-PBM interaction and validated a compound series which binds to the PRMT5-PBM interface and directly inhibits binding of SAPs. Mode of action studies revealed the formation of a covalent bond between a halogenated pyridazinone group and cysteine 278 of PRMT5. Optimization of the starting hit produced a lead compound, BRD0639, which engages the target in cells, disrupts PRMT5-RIOK1 complexes, and reduces substrate methylation. BRD0639 is a first-in-class PBM-competitive inhibitor that can support studies of PBM-dependent PRMT5 activities and the development of novel PRMT5 inhibitors that selectively target these functions.

Inhibition of Histone Deacetylases 1, 2, and 3 Enhances Clearance of Cholesterol Accumulation in Niemann-Pick C1 Fibroblasts.

Niemann-Pick disease type C1 (NPC1) is a rare genetic cholesterol storage disorder caused by mutations in the NPC1 gene. Mutations in this transmembrane late endosome protein lead to loss of normal cholesterol efflux from late endosomes and lysosomes. It has been shown that broad spectrum histone deacetylase inhibitors (HDACi's) such as Vorinostat correct the cholesterol accumulation phenotype in the majority of NPC1 mutants tested in cultured cells. In order to determine the optimal specificity for HDACi correction of the mutant NPC1s, we screened 76 HDACi's of varying specificity. We tested the ability of these HDACi's to correct the excess accumulation of cholesterol in patient fibroblast cells that homozygously express NPC1 I1061T , the most common mutation. We determined that inhibition of HDACs 1, 2, and 3 is important for correcting the defect, and combined inhibition of all three is needed to achieve the greatest effect, suggesting a need for multiple effects of the HDACi treatments. Identifying the specific HDACs involved in the process of regulating cholesterol trafficking in NPC1 will help to focus the search for more specific druggable targets.

The Use of Informer Sets in Screening: Perspectives on an Efficient Strategy to Identify New Probes.

Small-molecule discovery typically involves large-scale screening campaigns, spanning multiple compound collections. However, such activities can be cost- or time-prohibitive, especially when using complex assay systems, limiting the number of compounds tested. Further, low hit rates can make the process inefficient. Sparse coverage of chemical structure or biological activity space can lead to limited success in a primary screen and represents a missed opportunity by virtue of selecting the "wrong" compounds to test. Thus, the choice of screening collections becomes of paramount importance. In this perspective, we discuss the utility of generating "informer sets" for small-molecule discovery, and how this strategy can be leveraged to prioritize probe candidates. While many researchers may assume that informer sets are focused on particular targets (e.g., kinases) or processes (e.g., autophagy), efforts to assemble informer sets based on historical bioactivity or successful human exposure (e.g., repurposing collections) have shown promise as well. Another method for generating informer sets is based on chemical structure, particularly when the compounds have unknown activities and targets. We describe our efforts to screen an informer set representing a collection of 100,000 small molecules synthesized through diversity-oriented synthesis (DOS). This process enables researchers to identify activity early and more extensively screen only a few chemical scaffolds, rather than the entire collection. This elegant and economic outcome is a goal of the informer set approach. Here, we aim not only to shed light on this process, but also to promote the use of informer sets more widely in small-molecule discovery projects.

HDAC2 targeting stabilizes the CoREST complex in renal tubular cells and protects against renal ischemia/reperfusion injury.

Histone/protein deacetylases (HDAC) 1 and 2 are typically viewed as structurally and functionally similar enzymes present within various co-regulatory complexes. We tested differential effects of these isoforms in renal ischemia reperfusion injury (IRI) using inducible knockout mice and found no significant change in ischemic tolerance with HDAC1 deletion, but mitigation of ischemic injury with HDAC2 deletion. Restriction of HDAC2 deletion to the kidney via transplantation or PAX8-controlled proximal renal tubule-specific Cre resulted in renal IRI protection. Pharmacologic inhibition of HDAC2 increased histone acetylation in the kidney but did not extend renal protection. Protein analysis demonstrated increased HDAC1-associated CoREST protein in HDAC2-/- versus WT cells, suggesting that in the absence of HDAC2, increased CoREST complex occupancy of HDAC1 can stabilize this complex. In vivo administration of a CoREST inhibitor exacerbated renal injury in WT mice and eliminated the benefit of HDAC2 deletion. Gene expression analysis of endothelin showed decreased endothelin levels in HDAC2 deletion. These data demonstrate that contrasting effects of HDAC1 and 2 on CoREST complex stability within renal tubules can affect outcomes of renal IRI and implicate endothelin as a potential downstream mediator.

Multigram Preparation of BRD4780 Enantiomers and Assignment of Absolute Stereochemistry.

The development of a multigram synthesis of 3-exo-isopropylbicyclo[2.2.1]heptan-2-endo-amine hydrochloride (1) (also known as BRD4780 and AGN-192403) is described. The process involves protection of the amine as 4-nitrobenzyl carbamate, pNZ, which enables chiral SFC chromatography. The absolute configuration (AC) of the individual enantiomers has been determined by Mosher's amide method, VCD spectroscopy, and X-ray crystallography. We highlight the VCD approach as a rapid and effective means of AC determination that can be deployed directly on the target compounds.

Regulation of purine metabolism connects KCTD13 to a metabolic disorder with autistic features.

Genetic variation of the 16p11.2 deletion locus containing the KCTD13 gene and of CUL3 is linked with autism. This genetic connection suggested that substrates of a CUL3-KCTD13 ubiquitin ligase may be involved in disease pathogenesis. Comparison of Kctd13 mutant (Kctd13 -/- ) and wild-type neuronal ubiquitylomes identified adenylosuccinate synthetase (ADSS), an enzyme that catalyzes the first step in adenosine monophosphate (AMP) synthesis, as a KCTD13 ligase substrate. In Kctd13 -/- neurons, there were increased levels of succinyl-adenosine (S-Ado), a metabolite downstream of ADSS. Notably, S-Ado levels are elevated in adenylosuccinate lyase deficiency, a metabolic disorder with autism and epilepsy phenotypes. The increased S-Ado levels in Kctd13 -/- neurons were decreased by treatment with an ADSS inhibitor. Lastly, functional analysis of human KCTD13 variants suggests that KCTD13 variation may alter ubiquitination of ADSS. These data suggest that succinyl-AMP metabolites accumulate in Kctd13 -/- neurons, and this observation may have implications for our understanding of 16p11.2 deletion syndrome.

GSK3α, not GSK3ß, drives hippocampal NMDAR-dependent LTD via tau-mediated spine anchoring.

Glycogen synthase kinase-3 (GSK3) is an important signalling protein in the brain and modulates different forms of synaptic plasticity. Neuronal functions of GSK3 are typically attributed to one of its two isoforms, GSK3ß, simply because of its prevalent expression in the brain. Consequently, the importance of isoform-specific functions of GSK3 in synaptic plasticity has not been fully explored. We now directly address this question for NMDA receptor-dependent long-term depression (LTD) in the hippocampus. Here, we specifically target the GSK3 isoforms with shRNA knock-down in mouse hippocampus and with novel isoform-selective drugs to dissect their roles in LTD. Using electrophysiological and live imaging approaches, we find that GSK3α, but not GSK3ß, is required for LTD. The specific engagement of GSK3α occurs via its transient anchoring in dendritic spines during LTD induction. We find that the major GSK3 substrate, the microtubule-binding protein tau, is required for this spine anchoring of GSK3α and mediates GSK3α-induced LTD. These results link GSK3α and tau in a common mechanism for synaptic depression and rule out a major role for GSK3ß in this process.

A High-Content Screen for Mucin-1-Reducing Compounds Identifies Fostamatinib as a Candidate for Rapid Repurposing for Acute Lung Injury.

Drug repurposing has the advantage of identifying potential treatments on a shortened timescale. In response to the pandemic spread of SARS-CoV-2, we took advantage of a high-content screen of 3,713 compounds at different stages of clinical development to identify FDA-approved compounds that reduce mucin-1 (MUC1) protein abundance. Elevated MUC1 levels predict the development of acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) and correlate with poor clinical outcomes. Our screen identifies fostamatinib (R788), an inhibitor of spleen tyrosine kinase (SYK) approved for the treatment of chronic immune thrombocytopenia, as a repurposing candidate for the treatment of ALI. In vivo, fostamatinib reduces MUC1 abundance in lung epithelial cells in a mouse model of ALI. In vitro, SYK inhibition by the active metabolite R406 promotes MUC1 removal from the cell surface. Our work suggests fostamatinib as a repurposing drug candidate for ALI.

Comprehensive characterization of amino acid positions in protein structures reveals molecular effect of missense variants.

Interpretation of the colossal number of genetic variants identified from sequencing applications is one of the major bottlenecks in clinical genetics, with the inference of the effect of amino acid-substituting missense variations on protein structure and function being especially challenging. Here we characterize the three-dimensional (3D) amino acid positions affected in pathogenic and population variants from 1,330 disease-associated genes using over 14,000 experimentally solved human protein structures. By measuring the statistical burden of variations (i.e., point mutations) from all genes on 40 3D protein features, accounting for the structural, chemical, and functional context of the variations' positions, we identify features that are generally associated with pathogenic and population missense variants. We then perform the same amino acid-level analysis individually for 24 protein functional classes, which reveals unique characteristics of the positions of the altered amino acids: We observe up to 46% divergence of the class-specific features from the general characteristics obtained by the analysis on all genes, which is consistent with the structural diversity of essential regions across different protein classes. We demonstrate that the function-specific 3D features of the variants match the readouts of mutagenesis experiments for BRCA1 and PTEN, and positively correlate with an independent set of clinically interpreted pathogenic and benign missense variants. Finally, we make our results available through a web server to foster accessibility and downstream research. Our findings represent a crucial step toward translational genetics, from highlighting the impact of mutations on protein structure to rationalizing the variants' pathogenicity in terms of the perturbed molecular mechanisms.

A High Content Screen for Mucin-1-Reducing Compounds Identifies Fostamatinib as a Candidate for Rapid Repurposing for Acute Lung Injury during the COVID-19 pandemic.

Drug repurposing is the only method capable of delivering treatments on the shortened time-scale required for patients afflicted with lung disease arising from SARS-CoV-2 infection. Mucin-1 (MUC1), a membrane-bound molecule expressed on the apical surfaces of most mucosal epithelial cells, is a biochemical marker whose elevated levels predict the development of acute lung injury (ALI) and respiratory distress syndrome (ARDS), and correlate with poor clinical outcomes. In response to the pandemic spread of SARS-CoV-2, we took advantage of a high content screen of 3,713 compounds at different stages of clinical development to identify FDA-approved compounds that reduce MUC1 protein abundance. Our screen identified Fostamatinib (R788), an inhibitor of spleen tyrosine kinase (SYK) approved for the treatment of chronic immune thrombocytopenia, as a repurposing candidate for the treatment of ALI. In vivo , Fostamatinib reduced MUC1 abundance in lung epithelial cells in a mouse model of ALI. In vitro , SYK inhibition by Fostamatinib promoted MUC1 removal from the cell surface. Our work reveals Fostamatinib as a repurposing drug candidate for ALI and provides the rationale for rapidly standing up clinical trials to test Fostamatinib efficacy in patients with COVID-19 lung injury.

Multimodal small-molecule screening for human prion protein binders.

Prion disease is a rapidly progressive neurodegenerative disorder caused by misfolding and aggregation of the prion protein (PrP), and there are currently no therapeutic options. PrP ligands could theoretically antagonize prion formation by protecting the native protein from misfolding or by targeting it for degradation, but no validated small-molecule binders have been discovered to date. We deployed a variety of screening methods in an effort to discover binders of PrP, including 19F-observed and saturation transfer difference (STD) NMR spectroscopy, differential scanning fluorimetry (DSF), DNA-encoded library selection, and in silico screening. A single benzimidazole compound was confirmed in concentration-response, but affinity was very weak (Kd > 1 mm), and it could not be advanced further. The exceptionally low hit rate observed here suggests that PrP is a difficult target for small-molecule binders. Whereas orthogonal binder discovery methods could yield high-affinity compounds, non-small-molecule modalities may offer independent paths forward against prion disease.

Exploiting the Therapeutic Interaction of WNT Pathway Activation and Asparaginase for Colorectal Cancer Therapy.

Colorectal cancer is driven by mutations that activate canonical WNT/ß-catenin signaling, but inhibiting WNT has significant on-target toxicity, and there are no approved therapies targeting dominant oncogenic drivers. We recently found that activating a ß-catenin-independent branch of WNT signaling that inhibits GSK3-dependent protein degradation induces asparaginase sensitivity in drug-resistant leukemias. To test predictions from our model, we turned to colorectal cancer because these cancers can have WNT-activating mutations that function either upstream (i.e., R-spondin fusions) or downstream (APC or ß-catenin mutations) of GSK3, thus allowing WNT/ß-catenin and WNT-induced asparaginase sensitivity to be unlinked genetically. We found that asparaginase had little efficacy in APC or ß-catenin-mutant colorectal cancer, but was profoundly toxic in the setting of R-spondin fusions. Pharmacologic GSK3α inhibition was sufficient for asparaginase sensitization in APC or ß-catenin-mutant colorectal cancer, but not in normal intestinal progenitors. Our findings demonstrate that WNT-induced therapeutic vulnerabilities can be exploited for colorectal cancer therapy. SIGNIFICANCE: Solid tumors are thought to be asparaginase-resistant via de novo asparagine synthesis. In leukemia, GSK3α-dependent protein degradation, a catabolic amino acid source, mediates asparaginase resistance. We found that asparaginase is profoundly toxic to colorectal cancers with WNT-activating mutations that inhibit GSK3. Aberrant WNT activation can provide a therapeutic vulnerability in colorectal cancer.See related commentary by Davidsen and Sullivan, p. 1632.This article is highlighted in the In This Issue feature, p. 1611.

Selective inhibition of glycogen synthase kinase 3α corrects pathophysiology in a mouse model of fragile X syndrome.

Fragile X syndrome is caused by FMR1 gene silencing and loss of the encoded fragile X mental retardation protein (FMRP), which binds to mRNA and regulates translation. Studies in the Fmr1-/y mouse model of fragile X syndrome indicate that aberrant cerebral protein synthesis downstream of metabotropic glutamate receptor 5 (mGluR5) signaling contributes to disease pathogenesis, but clinical trials using mGluR5 inhibitors were not successful. Animal studies suggested that treatment with lithium might be an alternative approach. Targets of lithium include paralogs of glycogen synthase kinase 3 (GSK3), and nonselective small-molecule inhibitors of these enzymes improved disease phenotypes in a fragile X syndrome mouse model. However, the potential therapeutic use of GSK3 inhibitors has been hampered by toxicity arising from inhibition of both α and ß paralogs. Recently, we developed GSK3 inhibitors with sufficient paralog selectivity to avoid a known toxic consequence of dual inhibition, that is, increased ß-catenin stabilization. We show here that inhibition of GSK3α, but not GSK3ß, corrected aberrant protein synthesis, audiogenic seizures, and sensory cortex hyperexcitability in Fmr1-/y mice. Although inhibiting either paralog prevented induction of NMDA receptor-dependent long-term depression (LTD) in the hippocampus, only inhibition of GSK3α impaired mGluR5-dependent and protein synthesis-dependent LTD. Inhibition of GSK3α additionally corrected deficits in learning and memory in Fmr1-/y mice; unlike mGluR5 inhibitors, there was no evidence of tachyphylaxis or enhanced psychotomimetic-induced hyperlocomotion. GSK3α selective inhibitors may have potential as a therapeutic approach for treating fragile X syndrome.

MISCAST: MIssense variant to protein StruCture Analysis web SuiTe.

Human genome sequencing efforts have greatly expanded, and a plethora of missense variants identified both in patients and in the general population is now publicly accessible. Interpretation of the molecular-level effect of missense variants, however, remains challenging and requires a particular investigation of amino acid substitutions in the context of protein structure and function. Answers to questions like 'Is a variant perturbing a site involved in key macromolecular interactions and/or cellular signaling?', or 'Is a variant changing an amino acid located at the protein core or part of a cluster of known pathogenic mutations in 3D?' are crucial. Motivated by these needs, we developed MISCAST (missense variant to protein structure analysis web suite; http://miscast.broadinstitute.org/). MISCAST is an interactive and user-friendly web server to visualize and analyze missense variants in protein sequence and structure space. Additionally, a comprehensive set of protein structural and functional features have been aggregated in MISCAST from multiple databases, and displayed on structures alongside the variants to provide users with the biological context of the variant location in an integrated platform. We further made the annotated data and protein structures readily downloadable from MISCAST to foster advanced offline analysis of missense variants by a wide biological community.

Translation of HDAC6 PET Imaging Using [18F]EKZ-001-cGMP Production and Measurement of HDAC6 Target Occupancy in Nonhuman Primates.

Histone deacetylase 6 (HDAC6) is a multifunctional cytoplasmic enzyme involved in diverse cellular processes such as intracellular transport and protein quality control. Inhibition of HDAC6 can alleviate defects in cell and rodent models of certain diseases, particularly neurodegenerative disorders, including Alzheimer's disease and amyotrophic lateral sclerosis. However, while HDAC6 represents a potentially powerful therapeutic target, development of effective brain-penetrant HDAC6 inhibitors remains challenging. Recently, [18F]EKZ-001 ([18F]Bavarostat), a brain-penetrant positron emission tomography (PET) radioligand with high affinity and selectivity toward HDAC6, was developed and evaluated preclinically for its ability to bind HDAC6. Herein, we describe the efficient and robust fully automated current Good Manufacturing Practices (cGMP) compliant production method. [18F]EKZ-001 quantification methods were validated in nonhuman primates (NHP) using full kinetic modeling, and [18F]EKZ-001 PET was applied to compare dose-occupancy relationships between two HDAC6 inhibitors, EKZ-317 and ACY-775. [18F]EKZ-001 is cGMP produced with an average decay-corrected radiochemical yield of 14% and an average molar activity of 204 GBq/µmol. We demonstrate that a two-tissue compartmental model and Logan graphical analysis are appropriate for [18F]EKZ-001 PET quantification in NHP brain. Blocking studies show that the novel compound EKZ-317 achieves higher target occupancy than ACY-775. This work supports the translation of [18F]EKZ-001 PET for first-in-human studies.