Gene duplication has played a critical role in the evolutionary history of proteins, enabling complex multimers to emerge from simpler precursors. Yet in protein engineering, current methods for directed evolution do not exploit gene duplication, hampering access to the vast array of diverse variants that are only enriched in the presence of a wild-type copy. We establish a directed evolution strategy for multimeric proteins that harnesses gene duplication to compensate for metabolic burden and self-assembly fitness, allowing previously inaccessible variants to be enriched. Starting from a homomeric 240-mer capsid, gene duplication enables selection of both extreme homomeric variants and obligate heteromers. This strategy significantly expands engineering access to diverse high-performing variants, while also supporting a plausible model for evolutionary diversification of higher-order multimers in nature.

Synthetic microbial ecology aims at designing communities with desired properties based on mathematical models of individual organisms. It is unclear whether simplified models harbor enough detail to predict the composition of synthetic communities in metabolically complex environments. Here, we use longitudinal exometabolite data of monocultures for 15 rhizosphere bacteria to parametrize a consumer-resource model, which we use to predict pairwise co-cultures and higher order communities. The capacity to artificially "switch off" cross-feeding interactions in the model demonstrates their importance in ecosystem structure. Leave-one-out and leave-two-out experiments demonstrate that pairwise co-cultures do not necessarily capture inter-species interactions within larger communities and broadly highlight the nonlinearity of interactions. Finally, we demonstrate that our model can be used to identify new sub-communities of three strains with high likelihood of coexistence. Our results establish hybrid mechanistic and data-driven metabolic models as a promising and extendable framework for predicting and engineering microbial communities.

Genetically encoded biosensors enable the monitoring of metabolite dynamics in living organisms. We present CoBiSe, a computational biosensor design approach using Constraint Network Analysis to identify optimal insertion sites for reporter modules in molecular recognition elements (MREs). Applied to the iron-binding protein DtxR from Corynebacterium glutamicum, CoBiSe identified a flexible connective loop (residues 138–150) for inserting the reporter module, resulting in IronSenseR, a novel ratiometric biosensor for ferrous iron (Fe2+). IronSenseR demonstrates high specificity for Fe2+ with dissociation constants of 1.78 ± 0.03 (FeSO4) and 2.90 ± 0.12 μM (FeCl2), while showing no binding to Fe3+ and other divalent cations. In vivo assessment in Escherichia coli, Pseudomonas putida, and Corynebacterium glutamicum confirmed IronSenseR’s capability to detect changes in the intracellular iron pool. The creation of IronSenseR underlines that by reducing search space and eliminating labor-intensive screening, CoBiSe streamlines biosensor development and enables precise creation of next-generation biosensors for diverse metabolites.

Integrative mobilizable elements (IMEs) are mobile genetic elements that reside stably integrated into chromosomes and rely on helper conjugative elements for horizontal transfer. Here, we identify and characterize a widespread family of IMEs, named Pseudomonadota Integrative Mobilizable Elements (PIMEs), which are distributed exclusively in the Pseudomonadota phylum. Genome and phylogenomic analyses reveal ∼1,000 putative PIMEs, comprising at least four distinct PIME subfamilies defined by distinctive genomic organizations and conserved hallmark features. Characterized PIMEs depend on helper conjugative plasmids of the incompatibility group P (IncP) and, upon induction, PIMEs excise, replicate and mobilize intra- and inter-species. Remarkably, the representative PIME and its helper conjugative plasmid engages in cross-complementation, revealing an unrecognized level of functional interplay between hijacker and helper element. We also demonstrate that PIMEs act as reservoirs of known and novel prokaryotic immune systems. Overall, our findings uncover an overlooked and disseminated family of IMEs, which likely plays an important role in bacterial ecology and evolution.

Transformer models, neural networks that learn context by identifying relationships in sequential data, underpin many recent advances in artificial intelligence. Nevertheless, their inner workings are difficult to explain. Here, we find that a transformer model within the AlphaFold architecture uses simple, sparse patterns of amino acids to select protein conformations. To identify these patterns, we developed a straightforward algorithm called Conformational Attention Analysis Tool (CAAT). CAAT identifies amino acid positions that affect AlphaFold’s predictions substantially when modified. These effects are corroborated by experiments in several cases. By contrast, modifying amino acids ignored by CAAT affects AlphaFold predictions less, regardless of experimental ground truth. Our results demonstrate that CAAT successfully identifies the positions of some amino acids important for protein structure, narrowing the search space required to make effective mutations and suggesting a framework that can be applied to other transformer-based neural networks.

High-throughput sequencing and computational protein design have created a growing gap between the discovery of new proteins and their functional characterization. In many instances, functional characterization requires one-to-one measurements—such as when detailed biochemical insights are desired or pooled selections are not possible—necessitating that individual variants be isolated and assayed. A major barrier to closing this gap is the cost to directly synthesize individual genes, which remains prohibitively expensive ($10–100 per sequence) and restricts these studies to small subsets of relevant variants, leaving many sequences without functional annotation. To address this, we developed user-defined Sorted Mutants (uSort-M), which combines pooled DNA synthesis, automated cell sorting of transformed Escherichia coli, and long-read sequencing to rapidly isolate and identify variants from diverse libraries. uSort-M can isolate, sequence, and validate individual variants from pooled libraries produced via diverse existing methods including multiplex assembly, error-prone PCR, or pooled QuickChange mutagenesis. Sorting single bacterial clones into 384-well plates is efficient: eight plates (3,072 wells) can be filled in 1–2 hours, with up to 90% of wells yielding monoclonal cultures. Commercial long-read sequencing enables accessible, fast, and cost-effective identification of individual sequences from isolated clones while tolerating wide variation in fragment length and diversity across the library. Applying this workflow to a 328-member scanning mutagenesis library of a 300-bp gene recovered 96% of desired variants at fivefold lower cost than traditional synthesis. Numerical simulations identify key parameters governing library recovery and enable accurate prediction of the sampling effort required to achieve target coverage. As library size increases, this workflow offers substantial savings over traditional gene synthesis or cloning. Due to its generalizability, efficiency, and reliance on standard instrumentation, uSort-M removes a key barrier to large-scale protein functional characterization.

The development of compact and efficient CRISPR-Cas systems is crucial for biomedical and therapeutic genome editing, particularly in vivo applications based on viral delivery. Here, we performed a comparative functional screen of seven Cas9 orthologs to systematically evaluate their genome editing activity in mammalian cells. Among these, Cme2, a 1008 amino acid nuclease recognizing a 5' NAGNGC PAM, emerged as a promising candidate based on its compact size and baseline editing activity. To overcome its limited native efficiency, we employed a dual engineering approach combining sgRNA scaffold optimization and rational protein mutagenesis. The resulting variant, enCme2, exhibits markedly improved editing efficiency across multiple loci in both mouse and human cells while maintaining extremely high specificity and minimal off-target activity. Importantly, the small size of enCme2 permits packaging of the complete system into a single rAAV vector, enabling efficient genome editing/HDR in in vivo tissues and mouse embryos, and facile generation of transgenic models. These results establish enCme2 as a compact, precise, and AAV-compatible genome editing platform with broad applicability for in vivo research and therapeutic approaches, especially where high specificity is desirable.

The rapid advancement of environmental sequencing technologies, such as metagenomics, has significantly enhanced our ability to study microbial communities. The eubiotic composition of these communities is crucial for maintaining ecological functions and host health. Species diversity is only one facet of a healthy community’s organization; together with abundance distributions and interaction structures, it shapes reproducible macroecological states, that is, joint statistical fingerprints that summarize whole-community behavior. Despite recent developments, a theoretical framework connecting empirical data with ecosystem modeling is still in its infancy, particularly in the context of disordered systems. Here, we present a novel framework that couples statistical physics tools for disordered systems with metagenomic data, explicitly linking diversity, interactions, and stability to define and compare these macroecological states. By employing the generalized Lotka–Volterra model with random interactions, we reveal two different emergent patterns of species interaction networks and species abundance distributions for healthy and diseased microbiomes. On the one hand, healthy microbiomes have similar community structures across individuals, characterized by strong species interactions and abundance diversity consistent with neutral stochastic fluctuations. On the other hand, diseased microbiomes show greater variability driven by deterministic factors, thus resulting in less ecologically stable and more divergent communities. Our findings suggest the potential of disordered system theory to characterize microbiomes and to capture the role of ecological interactions on stability and functioning.

Bacteria produce the alarmone nucleotides (p)ppGpp during stress to affect replication, transcription, translation, and metabolism. Recently, pGpp was identified as a third alarmone that is produced from the hydrolysis of (p)ppGpp. Although pGpp is a major component of bacterial stress responses, its precise role in mediating these responses is poorly understood. ppGpp and pppGpp bind translation GTPases and therefore directly affect translation to conserve resources during periods of stress. Here, we show that while pGpp is a weaker inhibitor of protein synthesis than ppGpp and pppGpp in vitro, pGpp production in the model Gram-positive bacterium Bacillus subtilis leads to faster translation inhibition in vivo. Faster translation inhibition is accompanied by greater levels of disengaged ribosomal subunits and hibernating ribosome dimers, suggesting that translation initiation is strongly inhibited. We show that alarmone production in vivo causes a severe depletion of GTP, which is sufficient for translation inhibition. Finally, we find that pGpp production also causes more robust transcriptome remodeling than (p)ppGpp production. This work supports a model that implicates all three alarmones in translation inhibition via GTP depletion.

Soil bacteria are critical to agricultural productivity and play a central role in several biogeochemical cycles. These organisms frequently experience desiccation, which deprives them of access to both water and nutrients. Desiccation eliminates the aqueous connections between soil pores, removing pathways for nutrient diffusion. Well-studied yet resource intensive water stress responses like osmolyte synthesis may thus be impractical in unsaturated environments. Accordingly, we observed how the rhizobacterium Pseudomonas synxantha 2-79 responds to co-occurring water and nutrient limitation at the single-cell level to understand what role osmolyte synthesis may play in its desiccation response. We constructed a transcriptional reporter to track P. synxanthas osmolyte response and collected extensive morphological and reporter expression data through experiments designed to mimic different rates and extents of soil drying. Only actively growing cells responded to an osmotic shock by synthesizing osmolytes: this response was not observed when we pre-starved bacteria. Soluble nutrient diffusivity in soil is restricted even when there is sufficient water to keep bacterial cells hydrated, so this pre-starved condition reflects gradual drying in which starvation precedes water stress. Despite the lack of osmolyte synthesis, prior starvation enhanced P. synxanthas ability to recover from osmotic stress once water and nutrients were restored. These results suggest that cellular changes associated with the response to starvation that go beyond osmolyte synthesis play an important role in microbial desiccation tolerance.

Hyperlysinemia is a life-threatening metabolic disorder that requires the continuous clearance of lysine. Engineered probiotics capable of degrading lysine in the gut represent a promising therapeutic strategy. However, the introduction of heterologous metabolic pathways can impose a substantial fitness burden on the bacterial host, potentially compromising the therapeutic efficacy. Current screening methods fail to adequately assess this pathway-induced stress. Therefore, optimizing methods to evaluate bacterial fitness after pathway modification is essential for developing effective bacterial therapies. Here, we present a label-free phenotypic screening approach using Fourier transform infrared (FTIR) spectroscopy to evaluate the physiological burden imposed by two distinct lysine catabolism pathways engineered E. coli Nissle 1917 (EcN): the plant-derived bifunctional enzyme LKR-SDR and the yeast-derived two-enzyme cascade Lys2-Lys5. Employing FTIR under lysine stress mimicking pathological concentrations, decoded pathway-specific stress signatures, and molecular resilience. Probiotics expressing LKR-SDR exhibited severe multisystem damage, including proteotoxicity, lipid peroxidation, and significant nucleic acid stress. In contrast, the Lys2-Lys5 strain demonstrated superior resilience, maintained structural integrity, and exhibited adaptive metabolic changes, primarily through lipid membrane remodeling. This study establishes FTIR spectroscopy as a rapid screening platform that identifies the Lys2-Lys5 pathway as optimal for probiotic therapies. By directly linking spectroscopic signatures to cellular fitness, FTIR spectroscopy accelerates the rational development of durable microbial therapeutics for inborn metabolic disorders.