Cognitive manifestations, including impairments in language and executive functions, are seen in amyotrophic lateral sclerosis (ALS), but the underlying mechanisms remain unclear. We mapped prefrontal cortex regions from ALS patients by integrating spatial and single-nucleus transcriptomics in a cognitively stratified patient cohort. We uncover that cognitive impairment in ALS is associated with distinct patterns of neuronal dysfunction and glial-vascular dysregulation that vary by region and cognitive subtype. Executive dysfunction is linked to reduced mitochondrial and synaptic activity in deep-layer dorsolateral prefrontal cortex neurons, whereas language-related deficits track with a diffuse pan-regional response involving glial and vascular abnormalities. Our analyses, validated by multiplexed imaging, further identify signatures in the prefrontal cortex that span both motor and cognitive phenotypes, including a multicellular gliosis response. The findings reveal that clinical heterogeneity in ALS is driven by phenotype-specific cellular interactions in motor and non-motor regions of the brain.

Cortical development involves rapid progenitor expansion and cell diversification supported by tightly regulated metabolic programs, yet these programs remain largely uncharacterized in human development. Here, we generated a metabolic atlas of the early human cortex using primary tissue and stem cell-derived cortical organoids. We observed dynamic changes in core metabolic functions, including an unexpected increase in glycolysis and pentose phosphate pathway (PPP) activity during late neurogenesis. Manipulation of glucose availability in cortical organoids altered cell-type composition, increasing outer radial glia (oRG) and inhibitory neuron populations. Pharmacological and genetic inhibition of PPP enzymes recapitulated these cell fate changes. Ribose was sufficient to rescue radial glia (RG) gene expression changes, revert organoid cell-type composition, and restore levels of ATP and hypotaurine. These data identify a critical role for the PPP in modulating RG cell fate specification and generate a resource for future exploration of additional metabolic pathways in human cortical development.

Bird genomes are the smallest among amniotes; however, they remain challenging to assemble due to their structural complexity. This study presents a fully phased diploid telomere-to-telomere reference genome for the zebra finch (Taeniopygia guttata), a model organism for neuroscience and evolutionary genomics. Combining sequencing strategies allowed the closing of nearly all gaps, adding ∼90 Mbp of sequence (7.8%). The assembly contains gapless sequences for all microchromosomes, including the elusive dot chromosomes with their distinctive architecture. The genome was comprehensively annotated for genes, repeats, and structural variants. Complete centromeres were identified, along with candidate DNA-binding centromere protein B (CENP-B) homologs, suggesting that birds possess a kinetochore-associated CENP system similar to that of mammals. Relative to the previous reference generated by the Vertebrate Genomes Project, 2,710 (8.13%) previously unassembled or unannotated genes were identified. This complete genome of a songbird serves as a public reference and illuminates avian genome architecture and function.

We present T2T-MMU8v2.0, a near-perfect telomere-to-telomere assembly of the rhesus macaque (Macaca mulatta), representing high base-level accuracy reported in a primate genome. Our optimized assembly strategy exposes subtelomeric satellite-rich regions as the principal bottleneck in obtaining full assemblies, owing to long-read technological biases and limitations in hybrid assembly frameworks. By resolving ∼8 Mbp of SATR satellite arrays and discovering 268 previously unannotated repeat families, we define four distinct SATR genomic architectures, each with unique satellite composition, segmental duplication organization, and epigenetic signatures, which are distinct from the subtelomeric architectures observed in hominid genomes. These regions harbor 58 actively transcribed genes, suggesting gene innovation within these repetitive regions. Functionally, T2T-MMU8v2.0 improves alignment accuracy and chromatin accessibility detection, enabling a finer resolution in population variation and regulatory elements. Together, this work establishes a benchmark for primate genomics and illustrates the functional and evolutionary importance of previously inaccessible structures of the genome.

The common marmoset is a New World monkey widely used to study primate evolution and human disease. We present a telomere-to-telomere (T2T) reference assembly for the species, plus three near-T2T haplotypes. These resolve previously inaccessible regions, including the centromeres, sex chromosomes, subterminal satellites, acrocentric chromosomes, and the major histocompatibility complex (MHC). We find marmoset centromeres carry dimeric alpha satellites with chromosomal specificity, flanked by inactive layers interpreted as ancestral centromere remnants. We assemble gene-poor, satellite-rich short arms of the acrocentrics and find that most can harbor rDNA and all share pseudo-homolog regions (PHRs). PHR-sharing chromosomes also share closely related centromeric satellites, consistent with a model of ongoing rDNA-facilitated recombinational exchange between heterologous chromosomes. We further identify over 500 marmoset-lineage-specific transcribed genes with previously unknown transcript models or expansions. These resources, along with a preliminary pangenome, improve the utility of the marmoset as a model organism and address gaps in primate genome evolution.

Highly repetitive short arms of human acrocentric chromosomes have remained largely inaccessible to studies of meiotic recombination and de novo mutation. Integrating long-read and complementary sequencing approaches, we created 156 phased short arms and assessed 107 transmissions from 23 samples in a four-generation human pedigree. We observed a significant depletion of p-arm allelic recombination, although one ectopic recombination was identified between chromosomes 13 and 21, mediated by a large segmental duplication near the SST1 array. In contrast, 18 maternal-biased q-arm allelic recombination events were significantly enriched near the centromere. Relative to autosomal euchromatin, acrocentric short arms showed a 10-fold higher single-nucleotide variant rate, with a distinct mutation spectrum marked by reduced C>T but increased C>G and A>C mutations. These findings suggest that acrocentric sequence composition biases and limited allelic recombination contribute to an elevated mutation rate and promote distinct mutational processes linked to mismatch repair defects and oxidative DNA damage.

Human genome sequencing typically relies on mapping reads to a reference genome to call variants, but this approach introduces technical biases, excluding duplicated and structurally polymorphic regions of the genome. To overcome this, we present a telomere-to-telomere genome benchmark with near-perfect accuracy across 99.4% of the diploid HG002 genome. This benchmark adds 701.4 Mb of autosomal sequence and both sex chromosomes (216.8 Mb), which were absent from prior benchmarks. We annotated genes and repeats on both haplotypes, including 19,956 protein-coding genes on the maternal haplotype and 19,190 on the paternal haplotype, and developed new methods to measure the accuracy of reads, phased variant call sets, and assemblies against a diploid reference. Genome-wide analyses show that de novo assembly resolves 2%-7% more sequence and outperforms variant calling accuracy by an order of magnitude, expanding the reach of genomic medicine to the entire genome and enabling a new era of personalized genomics.

Activation-induced cytidine deaminase (AID) accomplishes somatic hypermutation (SHM) of VH(D)JH genes in germinal center B cells for antibody diversification and affinity maturation. How AID specifically targets VH(D)JH remains unclear. We report the discovery of LINE-1 (L1) retrotransposons upstream to many VH genes in the immunoglobulin locus. These L1s are evolutionarily old, truncated, and retrotransposition dead. Recombined VH promoters generate long, strong antisense RNAs encoding upstream L1s, triggering the human silencing hub (HUSH) complex and AID recruitment, which we term L1-driven SHM. We show that L1-driven SHM occurs in vivo using HUSH conditional knockout mice and engineered mice with VH genes devoid of upstream L1s. Insertion of transcriptionally active L1s at non-immunoglobulin loci endogenously lacking upstream L1s promotes off-target SHM. We demonstrate that old retrotransposons serve physiological roles, and our findings reveal how B cells co-opted an anti-retrotransposon silencing mechanism to promote antibody diversity. In doing so, we established a new link between cell-intrinsic innate and adaptive immunity.

Breastfeeding reduces the risk of severe lower respiratory infections (sLRIs), a leading cause of infant mortality; however, the protective mechanisms remain elusive. Here, we demonstrated that the absence of milk-derived osteopontin (OPN), highly expressed in colostrum, predisposes neonatal mice to viral and bacterial sLRI, consequent to disrupted dendritic cell (DC) hematopoiesis in the developing liver and lung. Amelioration of disease severity by oral OPN supplementation was associated with increased enteric abundance of Lactobacillaceae and elevated levels of serum 3-phenyllactic acid (PLA), a peroxisome proliferator-activated receptor gamma (PPARγ) agonist. Supplementation with PLA or the PPARγ agonist rosiglitazone restored lung DC hematopoiesis via airway epithelium-derived chemokine ligand 25 (CCL25)-mediated recruitment of lymphoid-myeloid primed progenitors and induction of a supportive lung niche. PLA-induced DC hematopoiesis and disease tolerance were attenuated by plasmacytoid DC depletion, immunoneutralization of stem cell factor, or genetic deletion of airway epithelial Flt3L. Our findings elucidate a microbiome-host interaction by which milk OPN confers protection against sLRI.

The brain vasculature comprises diverse specialized cells that are essential for brain function, yet their spatial organization remains poorly understood. Here, we construct a comprehensive cerebrovascular cell atlas encompassing 314,535 transcriptomes that captures the arteriovenous axis and defines consensus cell states. We then perform spatial transcriptomics to map 1,529,740 cells across the human temporal cortex and hippocampus, uncovering stereotyped micro-communities termed vascular cell ensembles. These ensembles comprise specialized subsets of endothelial cells, mural cells, fibroblasts, and perivascular macrophages that align with the arteriovenous architecture to coordinate segment-specific functions, such as neurovascular coupling, blood-brain barrier transport, and immune surveillance. By overlaying genetic risk and pharmacologic reactivity, we identify ensemble-specific susceptibilities and candidate therapeutic targets across neurological diseases, including small vessel disease and stroke. This study provides a resource to dissect the spatial and functional logic underlying human cerebrovascular biology and establishes a blueprint for decoding neurological disease susceptibility and therapeutic response.

The regulation of 3D cell shape is a fundamental problem of life. In multicellular tissues, cell shape emerges through the balance of forces inside and outside the cell. In epithelia, the basement membrane (BM) is the first extracellular barrier that cells sense biochemically and mechanically. Despite this, little is known about how BM mechanical properties are regulated and how they impact cell shape. Through mathematical modeling, we show that the stress relaxation time of the BM can regulate cell shape. Using molecular dynamics simulations, we show that the stress relaxation time of a collagen IV network can be inferred from the lifetime of collagen IV molecules. To measure collagen IV lifetime in vivo, we develop a fluorescent timer reporter for collagen IV and show that perlecan modifies collagen IV lifetime. This cross-disciplinary approach establishes a multiscale framework to probe matrix turnover, and its regulation and function in cell shape control.

Defense-associated reverse transcriptase (DRT) systems mediate antiviral immunity through distinct modes of cDNA synthesis: class 1 DRTs catalyze untemplated synthesis, whereas class 2 DRTs polymerize non-coding RNA-templated products. However, how these distinct modes drive defense remains unclear. Here, we report that DRT3 immunity arises when class 1 and class 2 RT activities cooperate to produce self-complementary double-stranded DNA (dsDNA). DRT3a uses a 5'-ACACAC-3' RNA template to synthesize poly-(dTdG) repeats, whereas DRT3b synthesizes poly-(dCdA) repeats without any nucleic acid template. Cryo-electron microscopy reveals that DRT3b forms a hexamer and uses active-site-adjacent residues as deoxyadenosine and deoxycytidine gates to enforce alternating nucleotide addition, representing a unique example of amino-acid-templated DNA polymerization. DRT3 is toxic in cells lacking RecBCD, implicating host recombination machinery in limiting dsDNA accumulation, and the phage-encoded RecBCD inhibitor Gam triggers DRT3-mediated abortive infection. These findings reveal how two polymerases with distinct templating strategies generate complementary DNA for antiviral defense.

Cancer management remains fragmented across its continuum, from late-stage diagnosis and salvage therapies to non-personalized surveillance. Here, we present Oncoformer, a unified multimodal transformer model trained on the China Oncology Multimodal Prediction and Surveillance Study (COMPASS) cohort (3.67 million individuals, 17.7 million clinical visits) and validated on independent external cohorts, including the UK Biobank. Oncoformer integrates longitudinal electronic health records with chest X-ray imaging to address multiple clinical tasks: pan-cancer diagnosis (area under the receiver operating characteristic curve [AUROC] = 0.956), future cancer prediction up to 1 year before diagnosis (AUROC = 0.869), tumor stage inference (mean AUROC > 0.90), patient-specific treatment-response forecasting, and recurrence-free survival stratification across ten cancer types (all p < 0.01). Staging predictions were independently validated against postoperative pathological endpoints and shown to converge on core cancer genomic pathways. By translating routine clinical data into a dynamic view of cancer evolution, Oncoformer provides a framework for risk-informed cancer prediction and treatment stratification using routine clinical data.

Systematically simulating and predicting phenotypic effects of diverse interventions across heterogeneous cellular environments is central to the vision of artificial intelligence virtual cell (AIVC). However, disparities in perturbagen modalities, assay formats, and data production efficiency hinder unified modeling and analysis of genetic and chemical screens. This study presents UniPert-G2CP, a two-stage deep learning framework that bridges genetic and chemical screens by unifying multimodal molecular perturbagen (cause) representations and enabling genetic-to-chemical perturbation phenotype (effect) transfer learning. Validated on large-scale genetic and chemical screening datasets, UniPert-G2CP enables more efficient, accurate, robust, interpretable, and generalizable simulation of multicellular and multidomain perturbation cause-effect spaces. Further joint analysis of these spaces reveals cellular heterogeneity underlying drug responses, providing mechanistic insights into drug action and resistance. Collectively, UniPert-G2CP advances universal biological causal modeling, accelerates AIVC realization, and expands the potential of AI-powered precision medicine.

Mg2+ is essential for all living organisms, yet its transport across mammalian membranes remains poorly understood. Here, we present cryoelectron microscopy (cryo-EM) structures of a full-length mammalian Mg2+ transporter on the plasma membrane, human CNNM4, in outward-facing and occluded states, revealing an unexpected tetrameric assembly organized as a dimer of asymmetric dimers-distinct from the symmetric dimers in prokaryotic homologs and long assumed for eukaryotic CNNMs. We show that Mg2+/ATP binding stabilizes the dynamic intracellular domains and promotes tetramerization, while an acidic patch binds additional Mg2+, potentially acting as a sensor to couple cytoplasmic Mg2+ levels to transport activity. Within the transmembrane domain, a key glutamate flips upon Na+ binding and destabilizes the Mg2+-binding site in the outward-facing state, thereby promoting Mg2+/Na+ exchange. Together, these findings establish a mechanistic framework for CNNM transport and regulation that diverges from prokaryotic models and links CNNM function to human physiology and disease.

Bacteriophage genomes are densely packed with coding sequences and frequently encode genes of unknown function. Unbiased phage functional genomics approaches are therefore needed, particularly for large lytic phages. Here, we harness the mariner transposase to develop phage transposon mutagenesis and sequencing (phage TnSeq), which enables pooled sequencing to identify both fitness-conferring and dispensable genes. Using the Pseudomonas aeruginosa-infecting nucleus-forming jumbo phage ΦKZ (280,334 bp; 371 predicted genes), we show that ∼110 genes are fitness-conferring via phage TnSeq, identifying many known and previously unknown essential genes. Moreover, this phage harbors ∼261 non-essential genes, including some capsid and tail proteins, many of which are important for fitness across different clinical isolates or conditions. Phage TnSeq was also extended to a base-modified phage. Together, phage TnSeq is a scalable technology that can identify essential phage genes, generate knockouts in all non-essential genes, and sensitively assign the quantitative fitness contributions of every gene in parallel.

Following its discovery nearly a century ago, the signaling molecule auxin has captured the interest of plant scientists. Mostly through genetics, key components of the auxin response were identified, and these were rapidly connected in a topologically simple nuclear auxin pathway (NAP). While the identification of the NAP established a coherent framework for transcriptional auxin response, recent years have seen a remarkable series of advances in the evolution, mechanisms, and regulation of NAP components that shed new light on auxin response. Furthermore, it has long been known that auxin triggers both rapid and slow responses, but success in NAP characterization has eclipsed insights into rapid responses. Recently, a paradigm for rapid responses has emerged. This involves widespread protein phosphorylation, new roles for NAP components, and a set of contentious, yet resurrected, extracellular auxin-binding proteins. Here, we present recent progress in auxin responses, reflect on their historical context, and formulate a set of pertinent questions for the field.

The intrinsic cardiac nervous system (ICNS) is a key node in heart-brain communication and an emerging target for cardiac therapy, yet its physiological importance and functional organization remain poorly understood. Here, we show that the ICNS is essential for cardiac performance and survival across conditions. Using integrated genetic and imaging approaches in mice, we identify two molecularly distinct intrinsic cardiac neuron (ICN) subtypes that differ in extrinsic inputs, projection architectures, and physiological roles. Npy⁺ ICNs preferentially receive vagal input and mediate parasympathetic control of heart rate and coronary perfusion, and their ablation leads to fatal cardiac failure. By contrast, Ddah1⁺ ICNs receive sympathetic input and are required to preserve electrical stability and prevent sudden cardiac arrest under extreme physiological or psychological stress, with their activation providing cardioprotection. Together, these findings establish the ICNS as a critical regulator of cardiac function, providing a framework for precise, cell-type-targeted neuromodulatory therapies.

Genes that define cell identity are often highly transcribed, but how their activity is regulated to prevent pathological dysregulation remains elusive. Using a two-tiered genetic screen, we identify a network of genes regulating hypertranscribed loci, which is notably enriched for genes mutated in developmental disorders. Among these, ANKRD11, a chromatin regulator haploinsufficient in KBG syndrome, exerts progressively stronger repression on more highly transcribed genes. ANKRD11 enriches around hypertranscribed loci and forms biomolecular condensates via charge-block-patterned intrinsically disordered regions. These condensates spatially sequester elongation factors away from RNA polymerase II, thereby restricting transcription elongation. In mice, Ankrd11 loss abrogates this restriction, causing aberrant developmental gene activation, disrupted organogenesis, and embryonic lethality. Critically, KBG patient cells with ANKRD11 mutations show defective condensate formation and consequent overactivation of hypertranscribed genes. These results uncover a condensate-mediated mechanism that restricts hypertranscribed genes and suggest its disruption underlies developmental disorders such as KBG syndrome.

Protein-protein interactions underlie biological complexity, and modeling their coevolution is essential for characterizing and engineering molecular assemblies. While protein and genomic language models have excelled at modeling individual proteins, extending these capabilities to protein complexes remains challenging. We present multiple sequence alignment (MSA) Pairformer, a protein language model that builds on AlphaFold2/3's bidirectional refinement between sequence and pairwise residue representations to accurately model the evolution of protein-protein interactions, despite training exclusively on individual chains. MSA Pairformer achieves nearly 3-fold improvement over existing methods in predicting protein-protein interface contacts and better distinguishes binding from non-binding sequences. A learned attention mechanism selectively weights sequences by their inferred evolutionary relevance, enabling discovery of subfamily-specific contacts. On single-protein benchmarks, it achieves state-of-the-art contact prediction and strong variant effect prediction using only 111 million parameters, over two orders of magnitude smaller than frontier models. These results offer an evolutionarily grounded, computationally efficient alternative to the scaling paradigm.