npj Antimicrobials and Resistance

Multidrug-resistant pathogens pose a major threat to One Health. Within the past decade, CRISPR-Cas systems have been explored as sequence-specific antimicrobials. While chromosomal injury has been considered the primary mechanism underlying pathogen killing by CRISPR-Cas antimicrobials, the synergistic role of gene disruption together with chromosomal injuries remains poorly understood. In this study, we characterized a new class of CRISPR-Cas antimicrobials that simultaneously cotarget essential and defensive genes to enhance potency against the clinically relevant pathogen Staphylococcus aureus. High-throughput CRISPR screening identified top-performing guide RNAs for twenty functionally diverse essential and defensive genes across the S. aureus genome. CRISPR-Cas antimicrobials were modularly formulated to target single or multiple gene loci and packaged in phage-like particles for specific delivery. By engineering an S. aureus production host with a chromosomally integrated anti-CRISPR protein, we demonstrated efficient production of CRISPR-Cas antimicrobials targeting any S. aureus chromosomal locus without self-targeting. Characterization of CRISPR-Cas antimicrobials with single guide RNA designs revealed that potency varied according to targeted gene function, achieving up to a 4-log10 reduction in viability and outperforming traditional antibiotics. Multiplexed configurations were consistently more effective than single-targeting designs, with the top-performing design demonstrating a 4.7-log10 reduction in viability. Cotargeting essential and defensive genes revealed synergies that led to improved lethality and attenuated resistance, with enhanced activity in biofilms compared to traditional antibiotics. Genes involved in signaling and stress responses were important defensive targets for developing cotargeting CRISPR-Cas antimicrobials. Overall, this study establishes design principles for synergistic CRISPR-Cas antimicrobials applicable to next-generation precision antimicrobial development. SIGNIFICANCEThe ability to effectively combat multidrug-resistant pathogens is of primary importance to One Health. This study develops a generalizable design principle for formulating potent CRISPR-Cas antimicrobials that exploit synergistic cotargeting strategies for enhanced pathogen killing. In addition to chromosomal injuries, we found that disruption of gene function plays a crucial role in determining the lethality of CRISPR-Cas antimicrobials, providing a generalizable framework for effective CRISPR-Cas antimicrobial design. The development of a CRISPR-Cas antimicrobial production host with stable, chromosomally integrated anti-CRISPR genes greatly expands the modularity, adaptability, and efficiency of formulating CRISPR-Cas antimicrobials and enables deeper insights into the molecular mechanisms involved in eliminating multidrug-resistant pathogens.

Motivation: Antimicrobial resistance (AMR) has been identified as a top global public health threat. Accurate AMR phenotype prediction from whole-genome sequencing data is an essential tool for accelerating clinical decision-making and mitigating resistance spread. Although many previous works have explored the use of tree-based machine learning (ML) models to predict resistance, the field lacks a systematic evaluation of the training pipeline across a variety of pathogenic species and antibiotics. Results: Using nine clinically relevant species-antibiotic combinations from the NCBI antimicrobial susceptibility testing database, we present a detailed analysis of the ML pipeline and identify key factors affecting model performance and evaluation. We begin by relabelling all isolates using current CLSI minimum inhibitory concentration breakpoints to resolve inconsistencies and increase available data, resulting in up to a 19% label swap and 56% data enlargement per species-antibiotic combination. We identify several key training parameters including k-mer length, which can increase classification F1 scores by over 20 points compared to commonly used k-values, feature matrix truncation, which can induce polynomial time reductions with limited performance reduction, and ML model class. By comparing 5-fold cross-validation with evaluation on an unseen clinical dataset, we show that random cross-validation splits--often criticized as overly optimistic--can act as a strong proxy for downstream clinical performance, yielding closer F1 scores than phylogeny-aware splits in all cases. We finally present an interpretability study which shows that over 95% of k-mers used by our models are associated with identifiable genomic features. Our results highlight the importance of feature design, evaluation protocol, and biological analysis in genomic AMR prediction, and support tree-based models as a robust and interpretable method.

Triculamin is a potent antibiotic lasso peptide first isolated in 1967. Previous studies have demonstrated that its biosynthesis follows a non-canonical logic unlike any other lasso peptide. In this study, we investigate the role of the unusual follower peptide and demonstrate that it is essential for efficient biosynthesis. Using structural prediction and targeted mutations of key conserved residues, we hypothesize that the interactions between the follower peptide and the macrocyclase create an enzyme-substrate complex that ensures delivery of the core peptide to the enzyme active site. Moreover, we demonstrate that analogs of the lasso peptide can be produced by modifying the core peptide, highlighting the substrate promiscuity of the lasso macrocyclase and identifying lysine-3 in the lasso peptide ring as the site of acetylation. Lastly, we achieve successful heterologous expression in Burkholderia sp. FERM 3421, which proves to be a superior heterologous host.

Directed evolution consisting of iterative rounds of diversification, selection, and counter-selection, underlies modern protein and antibody engineering, yet small-molecule drug design still advances largely through high-throughput screening and medicinal-chemistry intuition. Transformer softmax attention is mathematically identical to the Boltzmann distribution that governs molecular binding at thermal equilibrium1, an isomorphism that prescribes a sequence-native Specificity Foundation Model (SFM)2. This framework was recently applied across seven molecular recognition domains3,4 and scaled into the drug-target SFM (dtSFM), the first to pair a full-scale encoder with a generative decoder5. Whether such a model can be driven, iteratively and under selection, to optimize leads rather than sample them once has not been shown. Here we present GenLoop, a closed generative drug design loop that turns single-pass generation into directed evolution of chemistry. dtSFM generates target-conditioned molecules and reranks them by their thermodynamic compatibility score. An orthogonal structural verifier, AlphaFold 3, is used that shares no architecture or training data with dtSFM. Cheminformatics filters enforce developability, and generative evolution is performed on the structurally verified candidates, selecting for predicted binders and counter-selecting against off-target chemistry. Applied across twelve drug targets spanning pharmacologically distinct mechanism classes, GenLoop produced AlphaFold 3-verified designs that reached the structural confidence of the approved drug for five of the twelve targets, with the best designs at interface iPTM 0.93-0.98 and PAE 0.8-2.0 [A], as well as resolving paralog selectivity across nine targets. Two full disease campaigns followed. For the cystic-fibrosis transmembrane conductance regulator, GenLoop designed nine developability-filtered and structurally novel lead candidates (iPTM up to 0.93, interface PAE 2.3 [A]) targeting all three orthogonal sites of the approved drug Trikafta. For the GLP-1 receptor family, dtSFM engineered tunable single-, dual-, and triple-receptor incretin designs, yielding 23 central-pocket candidates that are structurally novel at median iPTM 0.89 and interface PAE 1.95 [A]. GenLoop with dtSFM brings directed evolution to small molecules through computational-thermodynamic selection; wet-lab validation is the immediate next step.

Background: Antimicrobial resistance poses a major threat to global public health. Large language models (LLMs) offer new possibilities for optimizing antibiotic prescribing decisions, but the capabilities of general-purpose versus domain-specific medical LLMs under different prompting strategies remain to be clarified. Methods: This double-blind, randomized-sequence evaluation used a 2X2 factorial design comparing four AI conditions-the domain-specific model MedGo and the general-purpose model DeepSeek V3.5, each under standard direct prompting and chain-of-thought (CoT) prompting-alongside real physician prescriptions across 59 complex inpatient infection cases. Five parallel regimens were generated per case and independently evaluated by three senior clinicians (1-5 comprehensive score and five domain sub-scores). ChatGPT 5.2 was additionally assessed as an automated evaluation tool. Results: Score ranking: real physicians > MedGo-CoT > DeepSeek-CoT > MedGo> DeepSeek (Friedman test, p<0.001). In base mode, MedGo significantly outperformed DeepSeek (Holm-adjusted p=0.040). CoT improved both models (Holm-adjusted p<0.001 for DeepSeek; p=0.024 for MedGo) and reduced score dispersion. MedGo-CoT significantly outperformed DeepSeek-CoT in individualized adjustment (adjusted p<0.001) and dosing precision (adjusted p=0.005). ChatGPT-expert correlation was negligible (overall Kendall {tau}=0.153, p=0.003; subgroup {tau}=0.06-0.20, all p>0.05). Conclusions: Domain-specific medical LLMs enhanced by CoT approach the antibiotic decision-making level of real physicians, with advantages in individualization and dosing precision. However, notable deficiencies persist in antimicrobial stewardship ecological awareness and automated evaluation reliability, underscoring the continued indispensability of senior clinical expertise.

Abstract Climate change is altering environmental conditions that influence foodborne disease transmission, yet traditional systematic reviews cannot keep pace with expanding evidence. We assessed whether an LLM-assisted workflow could generate a rapid, repeatable, and policy-relevant living evidence base for climate-sensitive foodborne disease. We combined structured PubMed searches (2010-2023), gold-standard human labelling, and iterative refinement of a GPT?4?Turbo?based auto-labeller within the SysRev platform. Pathogens of public-health importance in England were selected a priori. Model performance was evaluated against human reviewers using recall, precision, specificity, accuracy, and balanced accuracy. The refined inclusion model achieved 89{middle dot}2% recall, 59{middle dot}2% precision, 84{middle dot}5% specificity, and 85{middle dot}4% accuracy across 1,044 screened abstracts, identifying 436 studies for inclusion. Post-hoc re-evaluation of discordant abstracts showed that records excluded by the model but included during initial human screening did not meet the refined inclusion criteria. Frequently identified climate exposures included rainfall, temperature, seasonality, and humidity; norovirus, Salmonella, Campylobacter, and Cryptosporidium were the most common pathogens. An LLM-assisted workflow can generate living evidence for climate-sensitive foodborne disease with high recall and improved screening consistency. The approach is scalable, auditable, and suitable for secure institutional environments, supporting horizon scanning and climate-health risk assessment.

Macrolides are key agents for treating infections caused by non-tuberculous mycobacteria (NTM). Nevertheless, chromosomal erm genes conferring inducible macrolide resistance are described in some NTM species, such as Mycobacterium abscessus and M. fortuitum, whereas M. chelonae had long been considered as lacking a functional erm. Recent descriptions from the USA and Japan of a new plasmid-borne erm(55) (erm(55)P) in M. chelonae and other rapidly growing mycobacteria (RGM) have challenged this assumption. We investigated erm(55)P occurrence in clinical RGM referred to the French National Reference Centre for Mycobacteria between 2012 and 2026 by genome screening and erm(55)P specific real-time PCR. Positive isolates underwent long-read whole genome sequencing (GridIon, Oxford Nanopore Technologies). Clarithromycin (CLR) minimum inhibitory concentration (MIC) was determined by broth microdilution (RAPMYCO and FRATMYC, Thermo Fisher) and read up to 14 days. Five clinical isolates showing inducible CLR resistance (MIC range <0.25-64 mg/L on day 3-4 and 128 - >128 mg/L on day 14) were positive for erm(55)P: one M. chelonae, three M. neoaurum, and one M. parafortuitum. erm(55)P-positive M. chelonae genomes from this and previous descriptions did not cluster together in the phylogenetic analysis of 263 genomes. The assembled plasmids showed high similarity to previously reported erm(55)-carrying plasmids, especially within the erm(55)P region. The upstream sequence of erm(55)P showed a secondary structure compatible with a possible translation attenuation mechanism. These findings document the first report of a plasmid-borne erm(55) in Europe in M. chelonae and other RGM and raise concern about the emergence of plasmid macrolide resistance in NTM.

Achromobacter species are emerging multidrug-resistant (MDR) pathogens in people with cystic fibrosis. Their increasing resistance has grown an interest in phage therapy as an alternative treatment strategy. However, the factors governing phage susceptibility remain poorly understood, thereby limiting the rational selection of phage candidates. Using 15 strictly lytic Achromobacter phages and 7 clinical cystic fibrosis isolates representing Achromobacter insolitus and Achromobacter xylosoxidans, we demonstrate substantial variation in infection efficiency across all 105 phage-host combinations, variation that could not be discerned from qualitative plaque assays alone. We integrated complete bacterial and phage genomes with quantitative efficiency-of-plating (EOP) assays and lineage-aware Bayesian mixed-effects modelling to show that phage infectivity in Achromobacter is governed predominantly by bacterial lineage and strain identity, accounting for 90% of total variance in log-normalised EOP, with individual strains varying substantially in permissiveness irrespective of species membership. After accounting for this lineage structure, no individual defence system, antimicrobial resistance gene class, or phage tail cluster retained a statistically significant independent or interaction association with infectivity. Together, these findings demonstrate that bacterial strain identity is the primary driver of infection outcome. Host defence systems and phage tail-associated genes remain biologically plausible contributors; their independent effect could not be resolved after accounting for lineage structure, indicating that infection outcomes are largely strain-dependent. This work shifts the question from which individual traits predict infection to how strain lineage and specific host-phage combinations jointly determine infectivity, and argues that quantitative phenotyping of individual phage-host pairs is essential for guiding phage candidate selection and supporting rational cocktail design against multidrug-resistant Achromobacter infections in cystic fibrosis. Impact statementChronic Achromobacter infections in cystic fibrosis are increasingly difficult to treat due to multidrug resistance and biofilm formation. Although phage therapy is a promising alternative, its development is limited by poorly understood and highly variable infectivity. Here, we show that infectivity within a phage host range spans a broad quantitative continuum spanning several orders of magnitude that cannot be captured by qualitative plaque assays. These infection efficiencies are primarily structured by bacterial lineage and strain identity, while the contributions of individual genomic features remain unresolved, given the current sample size. This work provides a framework for predicting phage-host compatibility and supports a shift from empirical screening toward rational, evidence-based phage selection for MDR Achromobacter infections.

The minimum inhibitory concentration (MIC) is a standard measure for describing the lowest effective dose concentration of an antimicrobial compound in clinical practice; yet, conventional assays often require a substantial amount of antimicrobial compound, limiting their use with scarce, purified agents. Here, we describe a simple and reproducible technique to evaluate the MIC for purified compounds with a limited sample size. The protocol describes the MIC steps against a bacterial strain while minimizing the use of reagents and materials. It is helpful for screening purified natural products as antimicrobial agents and in early-stage drug discovery. The protocol adapts standard microplate-based assays for two-fold dilution of the compound, ensuring their applicability in microbiological studies. The MIC value of the standard antibiotic kanamycin against Staphylococcus aureus, Vibrio fischeri, Klebsiella pneumoniae, and Escherichia coli was determined using our method, and was found to be consistent with the conventional broth microdilution method, validating its reliability. Therefore, this method offers a practical and viable solution for antimicrobial drug discovery, addressing the disparity between limited compound availability and comprehensive microbiological assessment of MIC.

Recent benchmarks show that deep-learning models for perturbation prediction do not outperform simple baselines operating in principal-component (PCA) space. We explain this with an information-theoretic ceiling: for any orthonormal projection basis Phi, the squared correlation between prediction and truth is bounded by the variance the basis explains (r2 [&le;] VE), so no model complexity can recover signal discarded at projection. On chemical perturbations (sciPlex3, LINCS L1000), the eigenbasis of a gene association network captures only 10-12% of drug-response variance and yields chance-level predictions, while PCA captures 90-99%. Graph wavelets built on the same network recover approx. 88%, localising the drug signal in high-frequency modes that the standard low-pass eigenbasis discards. On CRISPRa genetic perturbations the ranking inverts: the network basis outperforms PCA across all dimensions tested. Controls on topology, null networks and data leakage confirm the effect is structural. The right basis depends on the perturbation modality: PCA captures the variance that drives chemical responses, the network basis captures the cascade structure that drives genetic ones, and bases that access the network's full graph spectrum (such as graph wavelets) recover both from the same topology.

By utilizing a targeted genetic assay within a Fox Insight cohort (N = 1,987), this research establishes a hybrid, transparent, and interpretable predictive framework. Initial modeling via Firth penalized logistic regression discovered enrichment regarding the GBA N370S locus (OR = 0.01, FDR < .001), highlighting the critical role of epidemiological evaluation in enriched, human study populations. Advanced ensemble learning methods, refined through a meta-learner gradient boosting machine, attained an out-of-sample AUC of 0.929 on 15% of the analysis dataset partitioned via random sampling and strictly held-out from model training. Both global, visual machine learning explanations and local-Shapley interpretations provide transparency into the models and individual predictions representative of practical, collaborative human-artificial intelligence efforts, offering a solution that supports classification while remaining accessible and economical.

Clinical informatics pipelines increasingly compute validated clinical endpoints from upstream NLP outputs. Even when the endpoint is defined by an established rubric, translating that rubric across representations - natural language instructions, program logic, and reference implementations - can introduce specification drift, where ostensibly equivalent calculators yield meaningfully different scores. We study this phenomenon for the Expanded Disability Status Scale (EDSS), a standard measure of disability in multiple sclerosis. Holding constant a shared set of functional system (FS) subscores extracted by a large language model (LLM), we compare EDSS values computed across three representations of the same scoring rubric: prompt-executed natural language, LLM-generated code, and a canonical reference implementation. We characterize disagreement structure, distributional shifts, and clinically salient boundary flips, and we propose an audit workflow that treats endpoint computation as a first-class verification target in clinical NLP systems.

Accurate prediction of blood-brain barrier permeability (BBBP) is essential for central nervous system drug discovery, yet existing models are often limited by their reliance on predefined physicochemical descriptors, small-molecule-centered training sets, or conformation-dependent representations, which restricts their transferability across chemically diverse modalities especially peptides. In addition, publicly available BBBP datasets remain fragmented, inconsistently standardized, and weakly controlled for molecular redundancy, increasing the risk of data leakage and overestimated model performance. In this study, we propose BBBP-Atlas, a structure-aware BBB permeability prediction model designed for unified modeling of small molecules and peptides with the first cross-modal dataset OmniBBBP. Designed to bypass descriptor and conformation dependencies, our model represents standardized molecular structures as atom-level graphs to capture local atom-bond environments and long-range topological dependencies associated with BBB transport. This design enables direct learning of structure-permeability relationships from molecular topology. For model training and evaluation, we curated a cross-modal, redundancy-filtered database OmniBBBP that seamlessly unifies small molecules and complex peptides, containing 10,218 unique compounds with 9,316 small molecules and 902 peptides. BBBP-Atlas achieved an accuracy of 0.8914 and an MCC of 0.7678 on the independent test set. On a balanced external benchmark of 200 compounds, our model reached an AUC of 0.9108, an accuracy of 0.8500, and an MCC of 0.7000, outperforming LightBBB by an absolute MCC gain of 6%. Case studies further showed that BBBP-Atlas captured clinically meaningful BBB permeability patterns, correctly identifying lorlatinib as BBB-permeable and vancomycin as BBB-impermeable with high confidence. The OmniBBBP-backed BBBP-Atlas offers a versatile and cross-modal approach for single-compound prediction, batch screening, and dataset exploration for CNS drug discovery. BBBP-Atlas is available at https://cadd.drugflow.com/bbbp/.

Small molecule inhibitors of cell death have wide-ranging potential applications, both as tool compounds in the laboratory and as clinical modulators of pathologic cell death. Previous screening efforts have identified candidate compounds targeting the pro-apoptotic, pore-forming BCL-2 family proteins BAX and BAK, but the complex interactions of these proteins at the mitochondrial outer membrane (with other proteins and the membrane itself) present challenges for compound screening. Although no inhibitors of BAX or BAK have advanced to clinical testing to date, candidate inhibitors have thus far been identified via screening of membrane-containing systems such as liposomes and isolated mitochondria. To address some of the challenges of chemical screening for apoptosis inhibitors, we conducted a small molecule screen utilizing BH3 profiling, a method that quantifies mitochondrial outer membrane permeabilization (MOMP) upon treatment with pro-apoptotic peptides derived from BCL-2 family proteins. Of over 40,000 compounds screened, we identified a series of compounds that prevent MOMP in response to pro-apoptotic peptides. The most potent of these, CDL36, binds to BAX and prevents MOMP at early timepoints. In longer term viability assays, the cytoprotective effect of CDL36 is most potent against death induced by doxorubicin, a widely used chemotherapeutic agent that causes dose-limiting cardiovascular toxicity. Our results elucidate the mechanism of action of new and existing cell death inhibitors, providing a foundation for further development of these inhibitors and potential insights into the mechanisms mediating doxorubicin toxicity in patients.

Carbonic anhydrases are among the fastest known biocatalysts, reversibly facilitating the hydration of CO2 to HCO3- at rates up to 107 s-1, which warrants their investigation for industrial carbon capture technologies. However, engineering carbonic anhydrases to maintain stability under harsh industrial process conditions remains a key challenge, and sequence-to-function datasets compatible with machine learning to inform forward engineering are lacking. Here, we developed a high-throughput platform that couples cell-free gene expression with a gaseous CO2 colorimetric assay to map the fitness landscapes of carbonic anhydrases. From 96 diverse natural homologs, we identified a robust variant from the Aquificota phylum and conducted an exhaustive mutational scan and functional assessment of this enzyme at 70C and 90C, covering >99% of all single-amino acid substitutions (totaling 4,365 mutations assayed in 39,285 reactions). This biochemical landscape was used to benchmark 22 zero-shot protein fitness models and identify critical mutations that improved enzyme stability at 90C by more than three-fold. We then used both zero-shot protein language models and supervised learning to filter 419 model-generated variants from a ProteinMPNN library of 100,000 sequences, leading to a best-in-class enzyme that retained activity after incubation at 95C. This work demonstrates that integrating cell-free enzyme engineering with machine learning enables opportunities for high-throughput experimental measurements to benchmark and improve protein language models, accelerate design loops, and expand functional exploration within protein families where experimental information is limited.

Prime editing enables the precise modification of genomes, thereby holding great potential for the treatment of genetic diseases. Despite substantial advancements in prime editing technology and the initiation of the first clinical trial for treating chronic granulomatous disease, further enhancement of editing efficiency across edit types is still urgently needed. Here, we developed a compact prime editor, PE2{Delta}R, by deleting the RNase H domain of the MMLV reverse transcriptase (MMLV-RT). We then conducted a saturated mutagenesis screen targeting two DNA interacting regions within the PE2{Delta}R-RT Fingers domain. By integrating three highly effective mutations (I61R, V101R, S67W) into PEmax lacking RNase H domain (termed PEmax{Delta}RM3), we achieved up to a 90% increase in editing efficiency across editing types compared to PEmax. Structural modelling using AlphaFold 3 suggests that these mutations enhance primer-template stabilization and guide the RNA/DNA hybrid into a catalytically favourable trajectory, providing a mechanistic explanation for the enhanced activity. Taken together, our study demonstrates proof-of-concept for the application of unbiased mutagenesis screen to identify novel mutations that enhance prime editor performance. Furthermore, we discovered that RT variants (I61R, V101R, S67W) synergize with PEmax and epegRNA to improve prime editing efficiency across edit types, with the strongest improvement observed in introducing small deletions.

Introduction: Standardizing fluid and vasopressor resuscitation in sep- tic shock is challenging due to patient heterogeneity. We trained a causal model to identify optimal dosing during the first six hours of intensive care unit (ICU) admission. Methods: Graphical causal inference models were applied to estimate het- erogeneous treatment effects. Grounding models in expert clinical knowl- edge minimizes bias from spurious correlations to generate robust, contextu- ally meaningful recommendations. Our model was trained on 1,702 MIMIC database admissions and externally validated on 1,434 eICU admissions. Pri- mary outcomes were in-hospital survival and 24-hour clinical improvement (SOFA score reduction of two points or more). Findings: The cohort comprised 3,136 participants (median age 65 years [IQR 53-75]; 42.7% female). Deviation from vasopressor recommendations was associated with increased in-hospital mortality (median OR 5.61, 95% CI 5.44-5.78) and failed clinical improvement (median OR 6.33, 95% CI 6.17-6.50). Fluid deviations yielded corresponding median ORs of 1.02 (95% CI 1.02-1.02) and 1.14 (95% CI 1.14-1.14). In external validation, the model achieved a median survival AUROC of 0.73 (95% CI 0.69-0.77) and clini- cal improvement AUROC of 0.69 (95% CI 0.66-0.72), matching predictive baselines. Treatment effects were heterogeneous: optimal fluids increased survival by up to 4% in low-severity subgroups, while vasopressor responses varied from 0.5% to 17% across acute severity levels. Sensitivity analyses across 36 scenarios confirmed primary associations in 33 cases (91.7%). Interpretation: Recommendations from expert-grounded causal models correlate with improved septic shock outcomes in external validation, cap- turing significant heterogeneity in patient response.

Thiopeptides are ribosomally synthesized and post-translationally modified peptides (RiPPs) that form complex bioactive scaffolds through extensive enzymatic tailoring. The polyglycosylated thiopeptides persiathiacins, exhibit potent activity against multidrug-resistant Mycobacterium tuberculosis (Mtb) and methicillin-resistant Staphylococcus aureus (MRSA). The persiathiacin biosynthetic gene cluster encodes six cytochrome P450 (CYP) enzymes, but the logic of their oxidative modifications was unknown. Here, we establish a protoplast-based genetic system for Actinokineospora and systematically assign functions to all P450s. We demonstrate that PerX hydroxylates the central thiazole, PerV installs the third indole-core crosslink required for macrocyclization, and PerT, not PerU, catalyses indole N-hydroxylation. Combined gene inactivation and metabolite profiling reveal a hierarchical enzymatic sequence leading to the mature scaffold prior to sugar installation. Notably, the intermediate accumulating in the {Omega}perX mutant exhibits enhanced anti-M. tuberculosis potency compared to persiathiacin A (IC50 = 0.07 vs 1.5 g mL1). These results define the enzymatic logic and temporal organization of persiathiacin biosynthesis, providing a conceptual framework for rational diversification of complex thiopeptide natural products.

Natural products are central mediators of microbial interactions. However, once released into the environment, they also become available for neighboring microorganisms capable of degrading and modifying them through biotransformation. These biotransformations may fundamentally reshape metabolomes and influence community behavior, yet our understanding of these processes remains limited. Ribosomally synthesized peptides are particularly compelling in this context because their structural complexity and potent antimicrobial activity coexist with the potential to yield essential nutrients and reduced bioactivity through biotransformation. Identifying the pathways underlying these biotransformations is essential for understanding mechanisms that support microbial coexistence and nutrient recycling in soil microbiomes. Here, we used nisin as a model peptide to investigate biotransformation by soil bacteria. Selective isolation under nisin-rich, carbon-limited conditions yielded two Gram-negative isolates, Burkholderia stabilis and Pseudomonas fragi. Using growth assays and liquid chromatography-mass spectrometry, we found that both isolates grow in the presence of nisin while biotransforming and depleting the peptide. Burkholderia stabilis completely converted nisin through sequential cleavage of the C-terminus, hinge region and lanthionine ring C, whereas Pseudomonas fragi showed more limited processing restricted to the C-terminal region. Although these biotransformations dismantled structural features required for nisins antimicrobial activity, the intrinsic resistance of both isolates suggests a role beyond detoxification. We further detected nisin biosynthetic genes in the source environment, supporting nisins ecological relevance and suggesting that these bacteria may participate in its turnover in soil. Together, these findings reveal extensive microbial processing of nisin and support a role for antimicrobial peptide recycling in soil microbiomes.

Environmental resistome comprises diverse antibiotic resistance genes (ARGs) that can play critical roles in "One Health", facilitating the evolution, persistence, and dissemination of microbial resistances. Yet, knowledge gaps exist in resistome structures and ecological connectivity across various ecosystems at a national scale. Here, we combined nationwide extensive short- and long-read sequencing efforts for soils, sediments, waters and wastewater treatment plants across Denmark to resolve resistome composition, habitat specificity and connectivity. From over 7,000 sequenced environmental samples (24 Tb of metagenomic data) that were classified into 21 distinct habitat classifications, resistomes exhibited habitat-specific patterns. We identified core ARGs for establishing environmental baseline of ARGs, and habitat-associated indicator ARGs facilitating source tracking. Using 110 deep long-read metagenomes (9 Tb data), we showed that only a subset of cross-habitat commonly-abundant ARGs showed elevated associations with MGEs and broad host range, suggesting unequal resistome connectivity across ecosystems among environmental ARGs. Additionally, although natural habitats had much lower resistome relative abundance and transferability than human-associated habitats, some mobile environmental ARGs exhibited links to those in human pathogens. These findings establish an ecological framework for interpreting environmental resistomes and prioritizing ARGs for environmental surveillance in the One Health framework.