RMH

Biological organisms can produce magnetic nanoparticles made of magnetite to detect the Earth's magnetic field for orientation. The evolutionary mechanisms that led to the acquisition of this capacity remain unknown. The most ancient organisms known to perform magnetic orientation are magnetotactic bacteria. Under tight genetic control, they produce chains of magnetite nanoparticles in dedicated organelles to allow efficient navigation in aquatic environments containing molecular oxygen. Yet, these bacteria have been proposed to emerge before molecular oxygen became abundant on Earth, at a time when this navigation system could not have served its current function. Here, we show that gene clusters responsible of magnetite formation and organization enable resistance against hydrogen peroxide, a toxic oxygen-derived compound that was produced during the early stages of Earth's oxygenation when iron oxidation consumed molecular oxygen. Characterization of the nanoparticle's oxidation state supported the involvement of magnetite in resistance to hydrogen peroxide-induced oxidative stress. These findings are compatible with an early emergence of magnetotactic bacteria, before natural waters became oxygenated. We propose a conceptual framework to explain the origin of magnetic orientation in bacteria, and offer tools to identify the initiation of Earth's oxygenation in a time-resolved manner.

Prokaryotes form multicellular structures under both natural and experimental conditions, through developmental processes that sometimes echo those known from eukaryotes. Recent research has increasingly focused on the mechanisms underlying bacterial multicellular organization and its evolutionary origins. In this review, we first consider the modes of group formation, that is, clonal or aggregative, and show how mixed strategies in prokaryotic systems complicate this division. We then survey bacterial and, where known, archaeal multicellular structures, focusing on common patterns and differences in developmental dynamics across species, and highlighting the “soft” similarities with eukaryotic tissues and organismal organization. Next, we discuss the molecular and cellular mechanisms underlying the canonical hallmarks of multicellularity in prokaryotes. Finally, we discuss recent research on the evolution of prokaryotic multicellularity, drawing on experimental, computational, and comparative approaches. Studying the specific forms of prokaryotic multicellularity, beyond simplified descriptions (aggregative vs. clonal), is necessary for understanding its complexity, and its consequences for microbial ecology, biogeochemical cycles and the evolution of bacteria per se. The prevalence of multicellularity in bacteria and emerging evidence from archaea raise the possibility that LUCA already possessed multicellular traits, suggesting that multicellularity may be an ancestral feature of life. The emerging picture is that bacterial multicellularity only partly matches eukaryotic models, instead reflecting multiple independent evolutionary origins and a striking capacity to regenerate and diversify developmental programs.

Accurate Enzyme Commission (EC) number prediction is fundamental to protein function annotation and metabolic pathway reconstruction. Recent methods based on protein language model embeddings or 3D structure have improved EC prediction, but important limitations remain. Whole-sequence embeddings can dilute local functional signals, such as catalytic residues and functional domains, whereas structure-based methods depend on accurate protein structures and are therefore difficult to apply broadly. We present MIL-BC, a multiple instance learning framework for EC prediction that models each protein as a bag of domains rather than a single sequence-level object. This design is motivated by the observation that enzymatic function is often determined by constituent functional domains instead of the full-length sequence alone. Domain instances are generated automatically using our DCTdomain algorithm, avoiding dependence on curated domain annotations. MIL-BC aggregates domain-level evidence into protein-level EC predictions and assigns interpretable weights to domains that contribute most strongly to each prediction. We further combine the domain embedding with global embedding in MIL-BC+. MIL-BC+ achieved F1 improvements of 0.03 - 0.17 over the sequence-based baseline on all four test sets. The largest gains were observed for proteins with multiple enzymatic functions, where domain-level modeling better separates mixed functional signals. These results show that domain-centric multiple instance learning is an accurate and interpretable strategy for enzyme classification. By improving EC prediction without requiring structural data or curated domain labels, MIL-BC provides a scalable framework for large-scale protein function annotation.

Pepper fluorescent RNAs (FRs) enable spatiotemporal imaging of intracellular RNA with high resolution. Although spectral optimization of the HBC scaffold has extended coverage from cyan to red, systematic refinement of binding affinity, photostability, and brightness remains largely unexplored. We describe a hydrogen-bonding strategy that strengthens fluorophore-aptamer interactions by incorporating hydroxyl groups or polyethylene glycol (PEG) moieties into HBC derivatives. The resulting compounds, MBC and PBC620, bind Pepper over 6-fold and 10-fold more tightly than their parent dyes HBC and HBC620, respectively, and exhibit correspondingly higher signal-to-noise ratios and enhanced photostability, with 2.1-fold and 1.5-fold increases in cellular brightness. Molecular models reveal that improved affinity stems from optimally positioned hydrogen bonds with RNA nucleotides. MBC enables real-time tracking of genomic loci with enhanced contrast compared to standard HBC. This work establishes structure-guided interface engineering as a general strategy for improving fluorescent RNA systems.

Despite extensive protein engineering efforts to significantly increase the catalytic efficiency of plastic-degrading enzymes, inefficient plastic-enzyme interactions during the interfacial biocatalytic process continue to impede their widespread applications. Here, we engineered the complete csgBACEFG curli operon through multicopy chromosomal integration to generate a csgBACEFG-optimized (CSGO) biofilm with enhanced curli assembly and plastic-surface adhesion. Unlike previou curli-based enzyme immobilization studies using individual CsgA curli components or pre-existing curli scaffolds, our strategy uses the complete curli assembly machinery to construct a dense, surface-associated catalytic biofilm. FAST-PETase was then embedded in a 2-day PET-surface CSGO biofilm through the SpyTag/SpyCatcher system, producing 2.5-fold more hydrolysis products from PET film depolymerization than soluble FAST-PETase with the same initial enzyme mass. The carrier-free biofilm-supported FAST-PETase also retained more than 50% of its first-cycle product formation after five 24-h PET nanoplastic (NanoPET) degradation cycles. These findings support biofilm-mediated enhancement of PET depolymerization and provide a proof-of-concept strategy that may inform future biofilm-assisted plastic treatment.

Enzyme function prediction is a long-standing problem in biology, and deep learning methods have recently become popular to improve performance on this task. While most of these models focus on classifying enzymes or mapping them to known reactions, predicting unknown biochemical reactions remains underexplored. Here we thoroughly examine an approach that generates reactions absent from the training data from the enzyme alone. Existing methods such as BLASTp or CLIPZyme cannot, by design, output reactions that are absent from their training data or were not supplied at inference. The closest prior model, RxnPred, reports high fingerprint-similarity scores on held-out enzymes, a less stringent criterion than exact recovery, but fails to recover the correct answer on unseen reactions. We present a structure-conditioned generative model that outputs a complete reaction, reactants and products, and evaluate it on EnzymeMap_x, a reaction-disjoint benchmark in which no test reaction appears in training, under exact, stereo-respecting scoring. On this benchmark, within 400 sampled candidates, the model generates the complete, correct reaction for 25.0% of the subset of test enzyme—reaction pairs whose reaction is a new combination of metabolites seen in training. Surprisingly, it also correctly generates a few reactions composed of metabolites it has not seen in training (1.6%), though their embeddings are present in the vocabulary. With a lower budget, re-ranking places the correct reaction first for 5.6% of pairs and in the top ten for 12.4%, when the reaction is assumed not to be already known (0.2% and 8.1% without that assumption) along with confidence scores. Assessed more leniently, the model produces at least one correct reaction known for the enzyme (usually memorised from training) for 71.2% of test pairs. Going further, without any retraining, a sequence-only model given the reactants as well as the enzyme produces the correct product set for 9.5% of pairs within five candidates, without re-ranking. We also present head-to-head comparisons on these tasks with BLASTp, DIAMOND, RxnPred and Horizyn. Structure conditioning improves fingerprint similarity on homology-stratified splits, but a matched sequence-only ablation at the best configuration finds no measurable effect on exact recovery. Data coverage and chemical initialisation of the vocabulary seem to matter more than encoder geometry.

Comparative genomics and synthetic biology require models that can interpret, annotate, and design DNA sequences across species, yet genomic models have not combined these capabilities beyond a few well-characterized organisms. Here we introduce Nucleotide Transformer v3 (NTv3), a long-context cross-kingdom foundation model that provides a common backbone for functional prediction, gene annotation and controllable sequence generation across phylogenetically diverse species. NTv3 is pretrained on more than 128,000 genomes using sequences up to 1 Mb and post-trained on 16,262 functional tracks across 15 animal and plant genomes. NTv3 performs comparably to AlphaGenome on matched human assays and matches or exceeds evaluated published models for functional-track prediction across 19 species spanning mammals, birds, fish, plants, fungi and protists. Ablation experiments show that both genome pre-training and expanding functional supervision from human and mouse to 15 species improve track prediction. Leveraging these learned functional representations, NTv3 outperforms leading ab initio gene predictors across mammals, plants and insects, with fewer spurious gene models and improved recovery and reduced fragmentation of long genes. Finally, the same representations can be used with discrete diffusion to design regulatory DNA sequences with specific activity levels. We demonstrate this in plant and fly promoters and enhancers, and experimentally validate the activity and promoter selectivity of the generated fly enhancers. Together, NTv3 unifies genome interpretation, gene annotation and regulatory sequence design across the animal and plant kingdoms, with direct relevance to agriculture, biodiversity research and biotechnology.