Machine learning (ML) has accelerated molecular discovery, yet training models to generalize to out-of-distribution (OOD) chemical spaces remains fundamentally constrained by the high cost of experimental validation. In antibiotic discovery, where whole-cell phenotypic high throughput screening (HTS) is resource-intensive, iterative ML-guided compound selection - or Active Learning (AL) - offers a pathway to efficiently navigate available chemical spaces. However, the algorithmic tradeoffs between prioritizing compound novelty (exploration), predicted bioactivity (exploitation), and their impact on OOD generalizability remain unresolved for noisy, whole-cell biological systems. In this work, we systematically evaluate three AL strategies for whole-cell bacterial bioactivity and benchmark their effects on model accuracy, hit rate, and OOD performance. Using retrospective simulations on Mycobacterium tuberculosis HTS data, we identify an optimal AL strategy that balances predicted hit/non-hit novelty with overall hit rate. We then integrate the strategy in a closed-loop Borrelia burgdorferi antibiotic discovery HTS campaign. The AL-guided approach successfully increased the experimental screening hit rate five-fold (from a 0.2% rate within investigator-selected plates to 1.0%). Further, when the trained model was applied in prospective in silico selection of highly diverse compounds across multiple bacterial species, the AL-trained whole-cell inhibition predictor demonstrates 53-fold enrichment over investigator-directed screening (11.0% experimental validation of predicted hits). Of these, 100% demonstrated the intended narrow spectrum activity for Borrelia burgdorferi. These results demonstrate that calibrated AL strategies can overcome data acquisition bottlenecks and train generalizable property predictors able to extrapolate to OOD molecules.

The essential mycobacterial enzyme phosphopantetheinyl transferase (PptT) is a potential target for antituberculosis drugs. We here report the identification of (pyridin-3-ylmethyl)-substituted thioquinazolinones as hits in a screen against PptT and a structure-activity relationship (SAR) study of the chemotype. The resulting thioquinazolinones exert on-target whole-cell activity against Mtb in axenic culture and in macrophages with efficacy comparable to that of clinically used antituberculosis drugs. Increased susceptibility of a pptT hypomorphic strain of Mtb, resistance of PptT Trp170 mutant Mtb strains, and analysis of relevant mutations corroborated that the thioquinazolinones are on target, do not inhibit growth of other bacteria, are not cytotoxic to mammalian cells, did not bind to most mammalian receptors and ion channels tested, and did not react with glutathione. Mtb rapidly took up the thioquinazolinones and methylated them, affording products inactive against recombinant PptT.

Mycobacterium tuberculosis (Mtb) remains the world's deadliest bacterial pathogen1. There is an urgent medical need to develop new drugs that shorten the treatment duration to combat widespread multi-drug-resistant and extensive-drug-resistant Mtb. Here, we present a preclinical covalent compound, CMX410, that contains an aryl fluorosulfate (SuFEx)2 warhead and uniquely targets the acyltransferase domain of Pks13, an essential enzyme in cell-wall biosynthesis. CMX410 is equipotent against drug-sensitive and drug-resistant strains of Mtb and efficacious in multiple mouse models of infection. Inhibition by CMX410 is irreversible through a previously undescribed mechanism: CMX410 reacts with the catalytic serine of the AT domain of Pks13, rapidly and irreversibly disabling the active site by forming a β-lactam. CMX410 is highly selective for its target and thus demonstrates excellent pharmacological and safety profiles, including no adverse effects in a 14-day rat toxicity study up to 1,000 mg kg-1 per day. The distinctive mode of action from current drugs, high potency across all tested clinical isolates, oral bioavailability, favourable performance in drug combination testing and superior pharmacological and safety characteristics make CMX410 a promising first-in-class candidate to replace outdated cell-wall biosynthesis inhibitors, such as isoniazid and ethambutol, in tuberculosis regimens.

Tuberculosis (TB) remains a critical public health challenge, causing over 1.5 million deaths annually. Rising resistance to existing TB antibiotics threatens to reverse decades of progress in combating the disease. Developing new antibiotics is challenging due to high development costs and failure rates. Artificial intelligence offers a promising solution to streamline early drug discovery by predicting inhibitory properties of novel compounds, generating molecules with desired properties, and automating workflows through large language model agents. In this work, we introduce Fleming, an AI agent designed for TB antibiotic discovery. We curated and generated the largest dataset of TB inhibitors to date (n=114,933), and used these data to train both discriminative and generative models for TB inhibitor identification. Fleming orchestrates four specialized agents—a bacterial inhibition prediction agent, a molecular generation agent, a molecular optimization agent, and an ADMET agent—to perform key tasks in early TB antibiotic discovery. These agents incorporate literature search capabilities and chemical optimization tools, creating an integrated platform for TB antibiotic discovery. Using experimental data and human medicinal chemist review, we demonstrate how Fleming enhances the antibiotic discovery process, and mirrors the decision-making of medicinal chemists through an intuitive natural language interface. ### Competing Interest Statement The authors have declared no competing interest.

Bacterial adaptation to stress involves changes in transcription and mRNA degradation rates. In Escherichia coli , the Nudix hydrolase RppH initiates mRNA degradation by removing pyrophosphate from mRNA 5’-ends, converting 5’-triphosphates to 5’-monophosphates. We aimed to identify the RppH homolog in the globally important pathogen Mycobacterium tuberculosis (Mtb). We deleted each non-essential Nudix gene from Mtb to determine their impacts on mRNA phosphorylation states. Deletion of mutT4 (Rv3908) increased the relative abundance of 5’-triphosphates on myriad mRNAs across the transcriptome. Purified MutT4 converted mRNA 5’-triphosphates into monophosphates, and stimulated degradation by RNase E and RNase J. MutT4 has intrinsically disordered regions (IDRs), a common domain for biomolecular condensate formation. Microscopy showed that MutT4 forms condensates that dissociate upon addition of rifampicin, and that the N-terminal IDR is sufficient for condensate formation. These MutT4 condensates localize with RNase E and RNase J. Deletion of mutT4 in Mtb leads to a higher outer membrane permeability and resistance to oxidative stress. We conclude that MutT4 is the RppH homolog of Mtb, assembling in condensates that may act as degradation hubs. Our data indicate that MutT4 is unlikely to participate in DNA repair or nucleotide pool cleansing, and as such would more accurately be named RppH. ### Competing Interest Statement The authors have declared no competing interest. National Science Foundation, https://ror.org/021nxhr62, 1652756 National Institute of Allergy and Infectious Diseases, AI143575, AI156415, AI179080 Welch Foundation, A-0015 University of Massachusetts Amherst Institute for Applied Life Sciences, Midigrant and Core Facilities Incentive Funds

ABSTRACT Urinary tract infections (UTIs) are a major global health problem and are caused predominantly by uropathogenic Escherichia coli (UPEC). UTIs are a leading cause of prescription antimicrobial use. Incessant increase in antimicrobial resistance in UPEC and other uropathogens poses a serious threat to the current treatment practices. Copper is an effector of nutritional immunity that impedes the growth of pathogens during infection. We hypothesized that copper would augment the toxicity of select small molecules against bacterial pathogens. We conducted a small molecule screening campaign with a library of 51,098 molecules to detect hits that inhibit a UPEC Δ tolC mutant in a copper-dependent manner. A molecule, denoted as E. coli inhibitor or ECIN, was identified as a copper-responsive inhibitor of wild-type UPEC strains. Our gene expression and metal content analysis results demonstrate that ECIN works in concert with copper to exacerbate Cu toxicity in UPEC. ECIN has a broad spectrum of activity against pathogens of medical and veterinary significance including Acinetobacter baumannii, Pseudomonas aeruginosa , and methicillin-resistant Staphylococcus aureus . Subinhibitory levels of ECIN eliminate UPEC biofilm formation. Transcriptome analysis of UPEC treated with ECIN reveals induction of multiple stress response systems. Furthermore, we demonstrate that L-cysteine rescues the growth of UPEC exposed to ECIN. In summary, we report the identification and characterization of a novel copper-responsive small molecule inhibitor of UPEC. IMPORTANCE Urinary tract infection (UTI) is a ubiquitous infectious condition affecting millions of people annually. Uropathogenic Escherichia coli (UPEC) is the predominant etiological agent of UTI. However, UTIs are becoming increasingly difficult to resolve with antimicrobials due to increased antimicrobial resistance in UPEC and other uropathogens. Here, we report the identification and characterization of a novel copper-responsive small molecule inhibitor of UPEC. In addition to E. coli , this small molecule also inhibits pathogens of medical and veterinary significance including Acinetobacter baumannii, Pseudomonas aeruginosa, and methicillin-resistant Staphylococcus aureus .

BACKGROUND: A GPCR (G protein-coupled receptor) target-based approach was applied to identify antagonists of the arthropod-specific tick kinin receptor. These small molecules were expected to reproduce the detrimental phenotypic effects that had been observed in Rhipicephalus microplus females when the kinin receptor was silenced by RNA interference. Rhipicephalus microplus, the southern cattle tick, cattle fever tick, or Asian blue tick, is the vector of pathogenic microorganisms causing the deadly bovine babesiosis and anaplasmosis. The widespread resistance to acaricides in tick populations worldwide emphasizes that exploring novel targets for effective tick control is imperative. RESULTS: Fifty-three structural analogs of previously identified tick kinin antagonists were screened in a 'dual-addition' calcium fluorescence assay using a CHO-K1 cell line expressing the tick kinin receptor. Seven molecules were validated as non-cytotoxic antagonists, four of which were partial (SACC-0428764, SACC-0428780, SACC-0428800, and SACC-0428803), and three were full antagonists (SACC-0428799, SACC-0428801, and SACC-0428815). Four of these antagonists (SACC-0428764, SACC-0428780, SACC-0428799, and SACC-0428815) also inhibited the tick midgut contractions induced by the myotropic kinin agonist analog 1728, verifying their antagonistic bioactivity. The small molecules were tested on recombinant human neurokinin (NK) receptors, the one most similar to the invertebrate kinin receptors. Most molecules were inhibitors of the NK1 receptor, except SACC-0412066, a previously identified tick kinin receptor antagonist, which inhibited the NK1 receptor only at the highest concentration tested (25 mu m). None of the molecules inhibited the NK3 human receptor. CONCLUSION: Molecules identified through this approach could be useful probes for studying the tick kinin signaling system and midgut physiology. (c) 2024 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

There is a compelling need to find drugs active against Mycobacterium tuberculosis ( Mtb ). 4′-Phosphopantetheinyl transferase (PptT) is an essential enzyme in Mtb that has attracted interest as a potential drug target. We optimized a PptT assay, used it to screen 422,740 compounds, and identified raltitrexed, an antineoplastic antimetabolite, as the most potent PptT inhibitor yet reported. While trying unsuccessfully to improve raltitrexed’s ability to kill Mtb and remove its ability to kill human cells, we learned three lessons that may help others developing antibiotics. First, binding of raltitrexed substantially changed the configuration of the PptT active site, complicating molecular modeling of analogs based on the unliganded crystal structure or the structure of cocrystals with inhibitors of another class. Second, minor changes in the raltitrexed molecule changed its target in Mtb from PptT to dihydrofolate reductase (DHFR). Third, the structure-activity relationship for over 800 raltitrexed analogs only became interpretable when we quantified and characterized the compounds’ intrabacterial accumulation and transformation.

Safe, effective, and low-cost oral antiviral therapies are needed to treat those at high risk for developing severe COVID-19. To that end, we performed a high-throughput screen to identify non-peptidic, non-covalent inhibitors of the SARS-CoV-2 main protease (Mpro), an essential enzyme in viral replication. NZ-804 was developed from a screening hit through iterative rounds of structure-guided medicinal chemistry. NZ-804 potently inhibits SARS-CoV-2 Mpro (0.009 μM IC50) as well as SARS-CoV-2 replication in human lung cell lines (0.008 μM EC50) and primary human airway epithelial cell cultures. Antiviral activity is maintained against distantly related sarbecoviruses and endemic human CoV OC43. In SARS-CoV-2 mouse and hamster disease models, NZ-804 therapy given once or twice daily significantly diminished SARS-CoV-2 replication and pathogenesis. NZ-804 synthesis is low cost and uncomplicated, simplifying global production and access. These data support the exploration of NZ-804 as a therapy for COVID-19 and future emerging sarbecovirus infections.

DNA-encoded chemical library (DEL) technology provides a time- and cost-efficient method to simultaneously screen billions of compounds for their affinity to a protein target of interest. Here we report its use to identify a novel chemical series of inhibitors of the thioesterase activity of polyketide synthase 13 (Pks13) from Mycobacterium tuberculosis (Mtb). We present three chemically distinct series of inhibitors along with their enzymatic and Mtb whole cell potency, the measure of on-target activity in cells, and the crystal structures of inhibitor-enzyme complexes illuminating their interactions with the active site of the enzyme. One of these inhibitors showed a favorable pharmacokinetic profile and demonstrated efficacy in an acute mouse model of tuberculosis (TB) infection. These findings and assay developments will aid in the advancement of TB drug discovery.

Lipid biosynthesis in the pathogen Mycobacterium tuberculosis depends on biotin for posttranslational modification of key enzymes. However, the mycobacterial biotin synthetic pathway is not fully understood. Here, we show that rv1590, a gene of previously unknown function, is required by M. tuberculosis to synthesize biotin. Chemical-generic interaction experiments mapped the function of rv1590 to the conversion of dethiobiotin to biotin, which is catalyzed by biotin synthases (BioB). Biochemical studies confirmed that in contrast to BioB of Escherichia coli, BioB of M. tuberculosis requires Rv1590 (which we named "biotin synthase auxiliary protein" or BsaP), for activity. We found homologs of bsaP associated with bioB in many actinobacterial genomes, and confirmed that BioB of Mycobacteriumsmegmatis also requires BsaP. Structural comparisons of BsaP-associated biotin synthases with BsaP-independent biotin synthases suggest that the need for BsaP is determined by the [2Fe-2S] cluster that inserts sulfur into dethiobiotin. Our findings open new opportunities to seek BioB inhibitors to treat infections with M. tuberculosis and other pathogens.

ABSTRACT YisK is an uncharacterized protein in Bacillus subtilis previously shown to interact genetically with the elongasome protein Mbl. YisK overexpression leads to cell widening and lysis, phenotypes that are dependent on mbl and suppressed by mbl mutations. In the present work, we characterize YisK’s localization, structure, and enzymatic activity. We show that YisK localizes as puncta that depend on Mbl. YisK belongs to the fumarylacetoacetate hydrolase (FAH) superfamily, and crystal structures revealed close structural similarity to two oxaloacetate (OAA) decarboxylases: human mitochondrial FAHD1 and Corynebacterium glutamicum Cg1458. We demonstrate that YisK can also catalyze the decarboxylation of OAA ( K m = 134 µM, K cat = 31 min −1 ). A catalytic dead variant (YisK E148A, E150A) retains wild-type localization and still widens cells following overexpression, indicating these activities are not dependent on YisK catalysis. Conversely, a non-localizing variant (YisK E30A) retains wild-type enzymatic activity in vitro but localizes diffusely and no longer widens cells following overexpression. Together, these results suggest that YisK may be subject to spatial regulation that depends on the cell envelope synthesis machinery. IMPORTANCE The elongasome is a multiprotein complex that guides lengthwise growth in some bacteria. We previously showed that, in B. subtilis , overexpression of an uncharacterized putative enzyme (YisK) perturbed function of the actin-like elongasome protein Mbl. Here, we show that YisK exhibits Mbl-dependent localization. Through biochemical and structural characterization, we demonstrate that, like its mitochondrial homolog FAHD1, YisK can catalyze the decarboxylation of the oxaloacetate to pyruvate and CO 2 . YisK is the first example of an enzyme implicated in central carbon metabolism with subcellular localization that depends on Mbl.

Primitive data organization practices struggle to deliver at the scale and consistency required to meet multidisciplinary collaborations in drug discovery. For effective data sharing and coordination, a unified platform that can collect and analyze scientific information is essential. We present DAIKON, an open-source framework that integrates targets, screens, hits, and manages projects within a target-based drug discovery portfolio. Its knowledge capture components enable teams to record subsequent molecules as their properties improve, facilitate team collaboration through discussion threads, and include modules that visually illustrate the progress of each target as it advances through the pipeline. It serves as a repository for scientists sourcing data from Mycobrowser, UniProt, PDB. The goal is to globalize several variations of the drug-discovery program without compromising local aspects of specific workflows. DAIKON is modularized by abstracting the database and creating separate layers for entities, business logic, infrastructure, APIs, and frontend, with each tier allowing for extensions. Using Docker, the framework is packaged into two solutions: daikon-server-core and daikon-client. Organizations may deploy the project to on-premises servers or VPC. Active-Directory/SSO is supported for user administration. End users can access the application with a web browser. Currently, DAIKON is implemented in the TB Drug Accelerator program (TBDA).

4 & PRIME;-Phosphopantetheinyl transferase (PptT) is anessentialenzyme for Mycobacterium tuberculosis (Mtb) survival and virulence and therefore an attractive target for atuberculosis therapeutic. In this work, two modeling-informed approachestoward the isosteric replacement of the amidinourea moiety presentin the previously reported PptT inhibitor AU 8918 are reported. Althougha designed 3,5-diamino imidazole unexpectedly adopted an undesiredtautomeric form and was inactive, replacement of the amidinourea moietyafforded a series of active PptT inhibitors containing 2,6-diaminopyridinescaffolds.

Alternative mode-of-inhibition of clinically validated targets is an effective strategy for circumventing existing clinical drug resistance. Herein, we report 1,3-diarylpyrazolyl-acylsulfonamides as potent inhibitors of HadAB/BC, a 3-hydroxyl-ACP dehydratase complex required to iteratively elongate the meromycolate chain of mycolic acids in Mycobacterium tuberculosis (Mtb). Mutations in compound 1-resistant Mtb mutants mapped to HadC (Rv0637; K157R), while chemoproteomics confirmed the compound's binding to HadA (Rv0635), HadB (Rv0636), and HadC. The compounds effectively inhibited the HadAB and HadBC enzyme activities and affected mycolic acid biosynthesis in Mtb, in a concentration-dependent manner. Unlike known 3-hydroxyl-ACP dehydratase complex inhibitors of clinical significance, isoxyl and thioacetazone, 1,3-diarylpyrazolyl-acylsulfonamides did not require activation by EthA and thus are not liable to EthA-mediated resistance. Further, the crystal structure of a key compound in a complex with Mtb HadAB revealed unique binding interactions within the active site of HadAB, providing a useful tool for further structure-based optimization of the series.

The ability of Mycobacterium tuberculosis ( Mtb ) to resist and tolerate antibiotics complicates the development of improved tuberculosis (TB) chemotherapies. Here we define the Mtb protein CinA as a major determinant of drug tolerance and as a potential target to shorten TB chemotherapy. By reducing the fraction of drug-tolerant persisters, genetic inactivation of cinA accelerated killing of Mtb by four antibiotics in clinical use: isoniazid, ethionamide, delamanid and pretomanid. Mtb Δ cinA was killed rapidly in conditions known to impede the efficacy of isoniazid, such as during nutrient starvation, during persistence in a caseum mimetic, in activated macrophages and during chronic mouse infection. Deletion of CinA also increased in vivo killing of Mtb by BPaL, a combination of pretomanid, bedaquiline and linezolid that is used to treat highly drug-resistant TB. Genetic and drug metabolism studies suggest that CinA mediates drug tolerance via cleavage of NAD-drug adducts.

Diacylglycerol (DAG) is a versatile lipid whose 1,2-sn-stereoisomer serves both as second messenger in signal transduction pathways that control vital cellular processes, and as metabolic precursor for downstream signaling lipids such as phosphatidic acid. Effector proteins translocate to available DAG pools in the membranes by using conserved homology 1 (C1) domains as DAG-sensing modules. Yet, how C1 domains recognize and capture DAG in the complex environment of a biological membrane has remained unresolved for the 40 years since the discovery of Protein Kinase C (PKC) as the first member of the DAG effector cohort. Herein, we report the high-resolution crystal structures of a C1 domain (C1B from PKCδ) complexed to DAG and to each of four potent PKC agonists that produce different biological readouts and that command intense therapeutic interest. This structural information details the mechanisms of stereospecific recognition of DAG by the C1 domains, the functional properties of the lipid-binding site, and the identities of the key residues required for the recognition and capture of DAG and exogenous agonists. Moreover, the structures of the five C1 domain complexes provide the high-resolution guides for the design of agents that modulate the activities of DAG effector proteins.

Diacylglycerol (DAG) is a versatile lipid whose 1,2-sn-stereoisomer serves both as second messenger in signal transduction pathways that control vital cellular processes, and as metabolic precursor for downstream signaling lipids such as phosphatidic acid. DAG effector proteins compete for available lipid using conserved homology 1 (C1) domains as DAG-sensing modules. Yet, how C1 domains recognize and capture DAG in the complex environment of a biological membrane has remained unresolved for the 40 years since the discovery of Protein Kinase C (PKC) as the first member of the DAG effector cohort.

Diacylglycerol (DAG)-sensing C1 domains of Protein kinase C (PKC) isoforms are targets for therapeutic intervention in cancers, neurodegenerative disorders, and HIV latency reversal. High-resolution structural information on the C1 domains complexed to DAG or exogenous PKC agonists remained elusive for nearly 3 decades. This is attributed to the extreme hydrophobicity of the C1 complexes and the need to create an optimal membrane-mimicking environment to facilitate the complex formation and crystallization. We overcame these challenges and obtained high-resolution crystal structures of the C1B domain from PKCδ (C1Bδ) complexed to DAG and four potent PKC agonists of pharmacological relevance. The structures reveal precise stereospecific binding modes and hydrophobic interactions formed by the lipophilic groups of DAG and DAG lactone (AJH-836) with C1Bδ. Comparison of the C1Bδ complexes with PDBu (a tumor-promoting phorbol ester) and Prostratin (a candidate for the HIV latency reversal) provides a structural rationale for their distinct PKC-mediated cellular responses. The complex with Ingenol mebutate, a drug clinically approved for the treatment of actinic keratosis, shows how its angelyl group completes the hydrophobic rim of C1Bδ. The analysis of all five complexes enabled us to assign the functional roles to residues that form the consensus sequence of DAG-sensitive C1 domains. We conclude that the ligand promiscuity of C1 domains stems from a subset of conserved residues that can "cage" chemically diverse ligands while interfacing with the membrane environment. These structures serve as valuable guides for pushing the frontiers of PKC pharmacology and suggest a new strategy for the crystallization of peripheral membrane proteins.