Heterologous expression of enzymes from higher organisms limits the construction of microbial cell factories for natural product biosynthesis. Here we develop a ProteinMPNN-based sequence redesign strategy, guided by essential structural features and evolutionary conservation, to improve the bacterial expression of heterologous enzymes. Applied to two plant glycosyltransferases, TOGT and UGT84A56, for esculetin glycosylation, this strategy generates extensively redesigned variants with markedly enhanced soluble expression and catalytic activity in E. coli. In vitro enzyme assays and in vivo whole-cell conversions show consistent improvements, outperforming conventional solubility-enhancing strategies. Molecular simulations suggest that the improved performance arises from global optimization of hydrophobic and hydrophilic residue exposure while preserving productive substrate-binding interactions. Fed-batch fermentation achieved titers of 2.11 g/L cichoriin and 4.05 g/L aesculin. This work establishes a generalizable route for converting difficult-to-express eukaryotic enzymes into efficiently expressible and functional variants, thereby expanding the enzymatic toolbox for microbial cell factory construction. Heterologous expression of plant-derived enzymes often limits microbial biosynthesis of valuable natural products. Here, the authors establish an enzyme sequence redesign strategy guided by evolutionary and structural information, achieving enhanced glycosyltransferase activity in Escherichia coli for the efficient production of coumarin glucosides.

Nanopore direct RNA sequencing holds promise for advancing our understanding of the epitranscriptome. Recently, Oxford Nanopore Technologies released basecalling models capable of detecting N6-methyladenosine (m6A), inosine (I), pseudouridine (Ψ), and 5-methylcytosine (m5C). However, their performance and cross-reactivity with other modifications remain largely unexplored. Here, we systematically benchmark four available modification-aware basecalling models by evaluating their per-read and per-site predictions across synthetic molecules and biological samples from diverse species. Models performed well on highly modified, balanced synthetic constructs (AUC = 0.93–0.97, PR-AUC = 0.84–0.91), but their performance dropped sharply on unbalanced datasets that reflect modification abundances in biological samples (PR-AUC: 0.04–0.09). Analysis of in vivo rRNA samples confirmed this limitation, with false-discovery rate ranging from 50% to 100%, even after filtering with modification-free controls. We identify two major sources of false positives: cross-reactivities with other modifications and current alterations at sites neighbouring a modified residue. Finally, we demonstrate that basecalling error- and current-based methods can accurately detect modifications, offering effective alternatives for modifications lacking dedicated models. Our results highlight the utility and limitations of modification-aware basecalling models for RNA modification detection, and underscore the importance of including control samples to mitigate false-positive predictions.

Riboswitches are useful models for revealing how some RNA molecules undergo dynamic rearrangements of their structures to perform cellular functions. A great deal is known about riboswitch aptamer domains through sequence covariation analysis, which has been difficult to apply to expression platforms given their large sequence diversity. Here, we develop an approach to generate covariation models for entire riboswitch sequences including the aptamer domain and the expression platform. The method consists of bioinformatically extending aptamer domains to include downstream sequences and filtering these sequences using either computational or high-throughput experimental approaches to identify those that include bacterial intrinsic terminators. Filtered sequences are then used to generate covariation models. We developed this approach in the context of the fluoride riboswitch, characterizing 1901 fluoride riboswitch sequences using high-throughput in vitro transcription followed by next-generation sequencing, and generating a covarion model consistent with its mechanism. We then developed covariation models of the ZTP, lysine, and TPP riboswitches and find covariation support for previously published mechanisms. Our method represents a new approach to characterizing large numbers of riboswitch sequences and to generate covariation models of complete riboswitches, which should expand our understanding of riboswitch mechanisms and the evolution of RNA structure dynamics.

Lignin-related aromatics are promising renewable feedstocks, but their microbial conversion into useful compounds is fundamentally constrained by substrate toxicity and metabolic flux imbalance. Here, we report an autonomous microbial system in Corynebacterium glutamicum, which integrates evolutionary engineering with multiplexed dynamic control to overcome these constraints. This system features a robust, evolved chassis with reinforced cellular defenses and upgraded aromatic efflux machinery. We engineer a layered, multi-input regulatory network that enables real-time, substrate-responsive and precise control of metabolic pathways. This system achieves 99.89% conversion of diverse lignin-related aromatics to protocatechuic acid and subsequently achieves a titer of 53.40 g L⁻¹. This further supports high-titer production of value-added downstream chemicals, including cis,cis-muconic acid and β-ketoadipic acid with titers of 51.90 g L⁻¹ and 38.33 g L⁻¹, respectively. This work establishes a versatile and scalable paradigm for the autonomous bioprocessing of lignin feedstocks into sustainable value-added products. Biological valorization of lignin is hindered by substrate toxicity and metabolism inefficiency. Here, the authors established a platform for lignin-derived aromatics tolerance engineering and multi-node dynamic pathway control in Corynebacterium glutamicum.

Understanding protein architecture and predicting its structural tolerance to profound remodeling is pivotal for engineering functional proteins. We present SplitSeek-Pro, a deep learning model that evaluates amino acid-level splittability in folded proteins, a property critical for protein engineering tasks such as circular permutation and split reconstitution. By integrating primary sequences with 3D structural features, SplitSeek-Pro achieves residue-resolution predictions through a two-stage training process: large-scale pre-training followed by high-quality fine-tuning. Experimental validation on three distinct proteins confirms its superior predictive power over existing methods. Notably, SplitSeek-Pro identifies characteristic segments that function as cohesive, integral fragments analogous to super-secondary structural motifs. These results establish SplitSeek-Pro as a robust tool for rational protein engineering and offer insights into the fundamental structural building blocks of protein folding. To facilitate community access, we provide an automated web server at http://splitseek.topo.bio. Predicting protein splitability is pivotal for engineering functional variants. Here the authors present SplitSeek-Pro, a deep learning model integrating sequence and 3D features to achieve accurate residue-resolution split site prediction for protein design.

Accurate image annotation is essential for training artificial intelligence (AI) systems in biomedical image analysis, enabling tasks such as cell detection, tissue quantification, and disease characterization. However, creating pixel-level annotations is a time-consuming and labor-intensive process that requires expert input, limiting the development and adoption of AI methods. Recent advances in foundation models, such as the Segment Anything Model (SAM), enable interactive object segmentation from simple user prompts, but their integration into widely used bioimage analysis platforms remains limited and often requires technical expertise. Here we show that SAMJ, a user-friendly plugin for ImageJ/Fiji, enables fast, interactive, and accurate image annotation on standard computers without requiring programming skills or specialized hardware. SAMJ integrates efficient SAM variants into a familiar graphical interface, allowing users to delineate objects in large scientific images in real time using simple clicks or bounding boxes. This approach significantly reduces annotation effort, accelerates dataset creation, and broadens access to advanced AI-assisted annotation tools for the biomedical research community. García-López-de-Haro and colleagues present SAMJ, a plugin that brings the Segment Anything AI to ImageJ/Fiji, enabling fast, click-based image annotation on standard computers, and accelerating creation of biomedical training datasets.

Post-translational modifications are important for regulating cellular functions. Although traditional experimental methods accurately identify PTM sites, they are time-consuming. In this study, we propose a novel model capable of predicting 17 types of PTMs through multi-modal integration and AlphaFold predictions. Our model employs an enhanced CNN-transformer architecture to capture local dependencies within the sequence, while incorporating structural features and evolutionary patterns to effectively capture complex spatial relationships and global contextual dependencies. Through rigorous cross-validation and testing, our model demonstrates exceptional performance, achieving area under the curve scores of 96.5%, 91.6%, 91.0%, and 89.5% for the prediction of hydroxylation, malonylation, O-linked glycosylation, and phosphorylation, respectively. Notably, our model accurately identified known phosphorylation sites on tau and two recently identified residues linked to pre-tangle stages and early Alzheimer’s disease pathology. This work not only deepens the understanding of PTMs but also holds promise for advancing future research in the prediction of PTM sites and functional annotation.

Pseudomonas aeruginosa is a highly adaptive Gram-negative pathogen that is commonly managed with fluoroquinolone (FQ) antibiotics. Even without heritable resistance mutations, FQ-susceptible cells of P. aeruginosa can survive treatment in a phenomenon termed antibiotic persistence. Current models of stationary-phase FQ persistence established with other bacterial species posit that persister survival is dependent on DNA double-stranded break (DSB) repair via RecA-mediated homologous recombination (HR). In P. aeruginosa, however, the effects of RecA loss on FQ persistence vary depending on the culture conditions. In this current work, we find that stationary-phase P. aeruginosa does not depend on either of its canonical DSB repair pathways—HR or non-homologous end joining (NHEJ)—for FQ persistence. By creating cross-pathway deletion strains, we found that RecA and Ku (but not LigD) are conditionally essential, which suggests that P. aeruginosa can conduct LigD-independent NHEJ. Using a chromosomally integrated DSB reporter assay, we further show that P. aeruginosa can repair DSBs using two non-canonical pathways: LigD-independent NHEJ and microhomology-mediated repair. Together, these findings demonstrate that P. aeruginosa does not adhere to the conventional requirement for DSB repair after FQ treatment and can engage in alternative DSB repair pathways than those previously described.

Antimicrobial resistance is a major global public health threat, yet no new antibiotic class has been introduced in over 30 years. Fluoroquinolones (FQs), a type of bacterial DNA gyrase poison, are potent antibiotics which stabilizes gyrase–DNA cleavage complexes and therefore cause lethal DNA double-stranded breaks. However, rising FQ resistance combined with adverse effects highlight the need to identify new chemical scaffolds for gyrase poisons. Discovery efforts in this area have been limited by the lack of robust high-throughput screening (HTS) assays. Here, we describe a fluorescence-based, T5 exonuclease (T5E)-assisted DNA cleavage assay that exploits the formation of multiple gyrase–DNA cleavage complexes on a single plasmid. When stabilized by a gyrase poison, these complexes are denatured with sarkosyl and processed by proteinase K to generate small linear DNA fragments, which are selectively degraded by T5E, thereby decreasing fluorescence upon SYBR Green staining. This assay was miniaturized into an automated 1536-well ultra-HTS format and used to screen 7415 compounds, identifying numerous potent gyrase poisons, predominantly FQs. We also discovered a new gyrase poison, Pyr-AMC. Biochemical assays and molecular dynamic simulation demonstrate that a single Pyr-AMC molecule binds the hydrophobic gepotidacin-binding pocket of gyrase, stabilizing cleavage complexes, and trapping gyrase-mediated DNA cleavages.