Research Briefing

We show that dinoflagellates accumulate and crystallize nitrogen-rich molecules from aqueous solutions and use these crystals for growth. This behavior might have implications for understanding marine nitrogen cycling. Moreover, the crystallization behavior can be manipulated to biosynthesize difficult-to-crystallize reflective materials with tailored morphologies and optical properties.

A hemodynamics-driven magnetoelastic vascular graft (MVG) has been developed, and its biosafety and functionalities were successfully validated in vivo in rat and pig. The MVG restores blood flow and enables wireless, real-time and continuous stenosis diagnosis, and has clinical potential for improving the management of human vascular disease.

GRAPE-LM (generator of RNA aptamers powered by activity-guided evolution and language model) is a generative AI framework that enables the one-round generation of short RNA binders. When guided by CRISPR–Cas-based intracellular screening, GRAPE-LM outperforms traditional multi-round methods for aptamer evolution.

Given current challenges in gene-replacement and gene-editing therapies, tRNA-based approaches show promise for treating diseases caused by nonsense mutations. We engineered a suppressor tRNA gene for AAV delivery in vivo that targets UGA — the stop codon that is most commonly introduced by pathogenic nonsense mutations — in a disease-agnostic manner.

We developed Protein2PAM, an evolution-informed protein language model that predicts protospacer-adjacent motif (PAM) specificity of CRISPR–Cas enzymes. Using in silico mutagenesis and model-guided evolution, we generated Cas9 variants with broadened PAM compatibility and higher activity, enabling editing of genomic sites that were previously inaccessible owing to PAM recognition constraints.

We engineered bacteria to function as therapeutic ‘living glues’ that autonomously detect gastrointestinal bleeding and, in response, secrete adhesive and therapeutic proteins to enable targeted mucosal adhesion, promote healing and reduce inflammation in mouse models of inflammatory bowel disease.

Comprehensively resolving the cell state landscape requires integrating single-cell omics data from diverse studies. We developed CONCORD, a contrastive learning framework that leverages principled mini-batch sampling to learn denoised, batch-integrated and high-resolution representations of cells, capturing intricate structures such as differentiation trajectories and cell-cycle loops across numerous biological contexts.

Although base editing has potential as a gene therapy tool, bystander edits limit its clinical use for many pathogenic mutations. This work uses directed evolution to optimize adenine base editors and 3′-extended guide RNAs to enhance targeting, producing more precise editors with reduced bystander effects.

Antibody–lectin chimeras (AbLecs) are an immunotherapy modality designed to bind and block glycans that contribute to immune suppression in cancer. AbLecs enhance antitumor immune responses both in vitro and in vivo, outperforming conventional antibody therapies, including approved cancer drugs.

Scientists have now sequenced more than two million bacterial and archaeal genomes, outstripping the ability to perform gold-standard searches for genes. LexicMap is a new tool that can scan millions of genomes for a gene in minutes, precisely locating mutations. This approach will enable many applications in epidemiology, ecology and evolution.

Achieving uniform, isotropic sub-micrometer-resolution imaging at high speed across large, cleared tissues has been a major challenge for light-sheet microscopy. To address this, we developed a cost-effective system that enables tenfold faster volumetric imaging with isotropic resolution of complex biological organs such as the mouse brain and cochlea.

We engineered chimeric variants of the Arabidopsis thaliana pattern recognition receptor RLP23 by replacing the C-terminal domain from orthologous proteins in crop species. Expression of these chimeric receptors in tomato markedly enhanced broad-spectrum resistance without compromising yield. Similar results in rice and poplar demonstrate the broad applicability of this engineering strategy.

We applied emerging label-free biophysical techniques to map the size, composition and shape of RNA lipid nanoparticles. By linking these physical measurements to gene expression in human T cells and mice, we uncovered structure–activity relationships that guide RNA delivery.

Challenges in somatic variant calling include a lack of long-read variant callers and of publicly available benchmarking datasets. We developed DeepSomatic, a somatic variant caller for short- and long-read technologies, and created seven somatic variant benchmarks derived from cancer cell lines, which we make available as a public database: CASTLE-panel.

Precise control of transgene expression remains a challenge in engineering primary cells for diverse applications. We developed DIAL, a promoter editing framework that transmits transient inputs into stable setpoints of expression in primary cells and human induced pluripotent stem cells, paving the way for predictable programming of gene circuits in therapeutically relevant cells.

Most bacteria cannot be grown in the laboratory, which means that their genetic diversity is hidden to traditional culture-based studies. We combined a new DNA extraction method, long-read sequencing, bioinformatics and chemical synthesis to access the genetic diversity of uncultured soil bacteria and abiotically decode it to discover bioactive small molecules.

A new screen platform named DEFUSE (DEath FUSion Escape) enables high-throughput discovery of small molecule protein degraders. DEFUSE identified SKPer1, a molecule that works by inducing proximity between an oncogenic driver and an E3 ligase, opening new avenues for targeted protein degradation.

Antibody–drug conjugates (ADCs) are promising targeted cancer therapies but have a limited payload scope. Antibody–bottlebrush prodrug conjugates offer modular synthesis and high drug-to-antibody ratios, enabling the use of a broad range of payloads, including lower potency drugs, while performing favorably compared to traditional ADCs in preclinical models.

We developed a universal control system and multitarget optimization framework that enables dynamic and synchronized expression of multiple genes and facilitates the overproduction of specific secondary metabolites in Streptomyces species.

Precision CRISPR–Cas9-mediated genome engineering remains challenging, particularly gene integration and editing in non-dividing cells. We present Pythia, a deep learning solution that forecasts optimal repair templates and enables predictable and accurate genome editing in diverse cellular contexts, both in vivo and in vitro.