Communications Biology

The rapid loss of tissue viability during hypothermic storage frequently results in organ failure and high discard rates in transplantation. In contrast, ectothermic animals can tolerate prolonged exposure to low temperatures in their natural environments. Elucidating the mechanisms underlying this cold tolerance may therefore inform improved strategies for tissue preservation. Here, we investigate proteomic and metabolomic responses to cold exposure in the livers of two frog species from distinct habitats: the cold-sensitive African clawed frog (Xenopus laevis) and the cold-tolerant Northeastern Asian brown frog (Rana dybowskii). Cold exposure induced lipid mobilization in X. laevis, whereas it promoted phospholipid mono-unsaturation in R. dybowskii. Notably, treatment with monounsaturated fatty acids (MUFAs) or overexpression of stearoyl-CoA desaturase (SCD) reduced cell death during cold storage. Moreover, MUFA infusion significantly improved cell viability in mouse liver tissue under hypothermic conditions. Together, these findings suggest that lipid mono-unsaturation, an adaptive feature of cold-tolerant frogs, can be leveraged to enhance tissue preservation during cold storage.

Cattle are broadly classified into two subspecies: Bos taurus, adapted to temperate climates, and Bos indicus, adapted to tropical environments. Indicine cattle show better heat tolerance and stronger disease resistance, whereas taurine cattle are known for higher milk yield and better meat quality. Improving milk yield while maintaining resistance to heat stress and infectious diseases is an important objective in cattle breeding. Marker-assisted selection is an effective approach for improving economically important traits, highlighting the need to understand genetic differences between taurine and indicine cattle. However, genome-wide comparisons between European taurine and Indian indicine cattle remain limited. To address this gap, whole-genome sequencing of 48 Indian indicine cattle was performed and combined with publicly available data. This enabled a comparative genomic analysis of 74 Indian indicine and 83 European taurine individuals to identify genes involved in heat tolerance and immune response. Genome-wide analyses using Fst and XP-CLR identified 4,343 and 1,457 differentiated genes, respectively, with 826 genes common to both methods. These genes were mainly associated with immune response, protein stability, and cytoskeletal structure. Strong selection signals were observed in three heat shock protein genes (DNAJC11, DNAJC5, and DNAJB11) and 229 immune-related genes. To examine the inheritance of these genes through crossbreeding, a haplotype-resolved genome assembly was generated for the Indian crossbreed Sunandini, which showed predominantly taurine ancestry (69.13-96.04%), with a smaller indicine contribution (1.22-1.87%). Several genes related to heat tolerance and immune response were inherited exclusively from indicine cattle, highlighting their importance for environmental adaptation and future breeding programs.

Motor learning alters activities in multiple brain areas. While learning-induced activity change in these areas has been investigated, how the information flow changes in the network across those areas remains to be thoroughly examined. We analysed wide-field calcium imaging data spanning from the premotor cortex to the parietal cortex of marmosets while they learned a two-target forelimb-reaching task. We applied non-negative matrix factorization (NMF) to the activity data and extracted about 30 localized activity components. Encoding model analysis indicated that learning was associated with a decrease in activity components related to hand movements, and an increase in those related to external and reward signals. Causality analysis by embedding entropy (EE) revealed increases in causal links across activity components in different areas and stabilization of the network structure with behavioural improvements. These results indicate that motor learning entails both a redistribution of task-related activity and a reorganization of large-scale cortical network interactions.

Determining physiological stress at high resolution is crucial across diverse settings to enable informed decision-making in the context of health and disease. Saliva-based targeted multi-omics testing provides a powerful, non-invasive method to quantify physiological stress and circadian dynamics at high-frequency. In a laboratory crossover trial with 24-hour sampling comprising 413 saliva samples, we demonstrate high analytical reliability, distinct molecular individuality, and robust acute and delayed responses to physical exercise across proteins, metabolites, and lipids. Moreover, we present the most comprehensive existing dataset describing 24-hour molecular kinetics across these three omics layers. Leveraging this controlled setting, we applied machine learning to single-timepoint saliva samples to accurately predict recent physical exercise both immediately after and 24 hours later. Next, we translated this analytical framework to a real-world longitudinal setting of elite football players monitored over 16 months, comprising over 12,000 saliva samples. Despite increased biological and contextual variability, the model retained robust discrimination between exercise and rest on the following day. Based on prediction probabilities, we introduce a saliva-based internal strain metric, that captures internal load and can be harnessed to monitor physical exercise and recovery. Model robustness was further supported through out-of-sample validation using previously unseen observations. Our findings demonstrate that saliva-based targeted multi-omics reliably captures physical exercise and recovery states in both laboratory and real-world environments, providing a scalable framework for monitoring physical performance. This non-invasive approach holds broad potential for physiological monitoring and can serve as a blueprint for health- and disease-related contexts.

Cryo-electron microscopy (cryo-EM) has become a central technique for determining the structures of macromolecular complexes. However, experimental cryo-EM density maps often exhibit limited interpretability due to noise and heterogeneous, anisotropic resolution, complicating downstream atomic model construction. Here we present DEMO-EMReF, a Swin Transformer-based framework specifically designed for cryo-EM density map refinement that integrates channel attention with hybrid spatial attention. By combining local window-based attention with grid-based sparse global attention, DEMO-EMReF captures fine-grained local density features while modeling long-range, cross-region spatial dependencies, enabling more robust density refinement in heterogeneous and anisotropic map regions. DEMO-EMReF was systematically evaluated on large benchmark sets of both primary maps and half-maps spanning resolutions of 3.0-8.0 [A] and was compared against multiple state-of-the-art map enhancement methods. DEMO-EMReF consistently improves resolution-related metrics, map-model correlations, and local atomic resolvability, with particularly robust gains in challenging heterogeneous or anisotropic cases. Importantly, the refined maps facilitate more accurate and efficient automated atomic model building. Together, DEMO-EMReF provides a robust approach for enhancing cryo-EM density maps and enabling more reliable downstream studies.

We used an autonomous lab, comprising a large language model (LLM) and a fully automated cloud laboratory, to optimize the cost efficiency of cell-free protein synthesis (CFPS). By conducting iterative optimization, the LLM-driven autonomous lab was able to achieve a 40% reduction in the specific cost ($/g protein) of CFPS relative to the state of the art (SOTA). This cost reduction was accompanied by a 27% increase in protein production titer (g/L). Iterative experimental design, experiment execution, data capture and analysis, data interpretation, and new hypothesis generation were all handled by the LLM-driven autonomous lab. The interface between OpenAIs GPT-5 LLM and Ginkgo Bioworks cloud laboratory incorporated built-in validation checks via a Pydantic schema to ensure that AI-designed experiments were properly specified. Experimental designs were translated into programmatic specification of multi-instrument biological workflows by Ginkgos Catalyst software and executed on Ginkgos Reconfigurable Automation Cart (RAC) laboratory automation platform, with human intervention largely limited to reagent and consumables preparation, loading and unloading. By integrating LLMs with programmatic control of a cloud lab, we demonstrate that an LLM-driven autonomous lab can successfully perform a real-world scientific task, highlighting the potential of AI-driven autonomous labs for scientific advancement.

Quantification of intracellular Trypanosoma cruzi infection remains a central, yet methodologically challenging step in Chagas disease research and early-stage drug discovery. Current approaches largely rely on manual microscopy-based counting or on genetically modified parasites, both of which present limitations in scalability, reproducibility, or accessibility. Here, we developed and validated a neural network (NN)-based pipeline for the automated quantification of infection rates and parasite burden in mammalian cells using images stained exclusively with DNA-binding fluorescent dyes. Two independently refined deep-learning models were trained to segment host cell nuclei and intracellular amastigotes, respectively, and subsequently integrated into a unified algorithm that assigns each parasite to its nearest host cell. The pipeline was evaluated using confocal images from six mammalian cell lines infected with two T. cruzi strains and compared against blinded manual quantification. Automated detection of both, host nuclei and parasites, showed high concordance with manual counts, with median deviations around 5% and similar distributions of parasite burden per cell. In contrast to morphology-based image analysis methods, our NN-based approach demonstrated improved robustness across diverse cell types and staining conditions, reduced parameter dependency, and independent segmentation of host and parasite objects, minimizing error propagation. Although minor biases in parasite-to-cell assignment were observed in densely clustered cultures, overall infection indexes and burden estimates closely matched manual analysis. This accessible and scalable AI-assisted workflow provides a reproducible alternative to manual quantification and represents a methodological advance for standardized phenotypic screening of intracellular T. cruzi, supporting more robust and harmonized drug discovery efforts in Chagas disease.

Resolution estimation by Fourier shell correlation (FSC) using half data sets is the standard method for map quality assessment in cryo-EM. Currently, the FSC method is largely used for refined cryo-EM maps in the context of single particle cryo-EM or subtomogram averaging. Here, we extended resolution estimation to assess the quality of electron micrographs, tilt-series and tomograms. We developed a robust statistics-based framework, capable of determining local quality estimates in the above cryo-EM data types. We show that the determined quality values on a micrograph and tomogram level can be used as a particle quality criterion to improve averaged 3D reconstructions. Using local quality assessments of tomograms, we were able to characterize tomogram quality dependence on distance inferred by radiation damage of FIB-milled lamella. This robust resolution-based quality assessment approach suitable for multiple cryo-EM data types opens new possibilities for automated quality control and method development in cryo-EM maps as well as tomograms and micrographs.

Sudden cardiac death risk is 2-3-fold higher in athletes than in non-athletes. We classify sports-related cardiac arrhythmias using a novel explainability framework comprising data analysis, model interpretability, post-hoc visualisation, and systematic assessment. Two neural networks--one with interpretable sinc convolution and one with standard convolution--were trained on general-population ECGs (PhysioNet, n=88,253, 30 arrhythmias, three continents) and tested on professional footballers (PF12RED, n=161) via domain adaptation for normal sinus rhythm (NSR), sinus bradycardia (SB), incomplete right bundle branch block (IRBBB), and T-wave inversion (TWI). Sinc convolution achieved superior NSR detection (AUROC 0.75 vs 0.70), whilst standard convolution excelled at SB (0.74 vs 0.73), IRBBB (0.66 vs 0.58), and TWI (0.59 vs 0.54). Gradient-weighted Class Activation Mapping revealed that sinc models focus on physiologically relevant ECG segments (the PR interval for NSR/SB and the T wave for TWI). We hypothesise that sinc convolution better captures periodic rhythms but struggles with complex morphological patterns, suggesting architectural choice should align with underlying cardiac pathophysiology. Graphical abstractAbbreviations: AI, artificial intelligence; AUPRC, area under the precision-recall curve; AUROC, area under the receiver operating characteristic curve; Conv, convolution; ECG, electrocardiogram; Grad-CAM, gradient-weighted class activation mapping; IAVB, first-degree atrioventricular block; IRBBB, incomplete right bundle branch block; LAD, left axis deviation; LBBB, left bundle branch block; LVH, left ventricular hypertrophy; NSR, normal sinus rhythm; QT, QT interval; RAD, right axis deviation; RBBB, right bundle branch block; RVH, right ventricular hypertrophy; SA, sinus arrhythmia; SB, sinus bradycardia; TWI, T-wave inversion; xAI, explainable artificial intelligence. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=123 SRC="FIGDIR/small/26346628v1_ufig1.gif" ALT="Figure 1"> View larger version (60K): org.highwire.dtl.DTLVardef@124f232org.highwire.dtl.DTLVardef@98ba6forg.highwire.dtl.DTLVardef@f7fdedorg.highwire.dtl.DTLVardef@14012b3_HPS_FORMAT_FIGEXP M_FIG C_FIG

Obsessive-compulsive disorder (OCD) is a disabling condition that is characterized by disruptions in distributed brain circuit dynamics. However, current network studies predominantly evaluate these circuits by measuring functional synchrony (connectivity) between pairs of regions of interest, potentially overlooking complex higher-order interactions. In this study, we applied a Hodge Laplacian topological framework to investigate these higher-order interactions in OCD. Using a large-scale resting-state fMRI dataset from the ENIGMA-OCD consortium (1,024 OCD patients and 1,028 healthy controls across 28 sites worldwide), we identified significant disruptions in topological loops spanning frontoparietal, default mode, and sensorimotor networks. Crucially, the edges constituting these abnormal loops largely lacked significant pairwise differences, highlighting higher-order multi-nodal disturbances. Subgroup analyses revealed that these disruptions were most pronounced in adult, medicated, and high-severity OCD patients. Our findings suggest that OCD pathology involves abnormal recurrent higher-order multi-region interactions, providing new insights into the brains functional organization and offering potential biomarkers for clinical application.

Fluctuations in attentional states, such as mind-wandering (MW), are associated with critical variability in task performance. While fMRI studies highlight the opposing roles of task-positive (e.g., dorsal attention network) and task-negative (e.g., default mode network) systems, the electrophysiological mechanisms underlying these dynamics remain poorly understood. Using intracranial electrocorticography in humans performing a sustained attention task, we identified global oscillatory dynamics linked to attentional shifts. MW was characterized by (i) reduced theta ({theta}) and alpha ({square}) power, (ii) decreased aperiodic signal components, indicating a shift toward cortical inhibition, (iii) enhanced phase synchronization across networks, and (iv) strengthened {theta} phase-behavior correlations ({rho}). These features support a non-network-specific framework in which low-frequency {theta} dynamics--captured by both {theta} power and {rho}--are associated with attentional fluctuations, while aperiodic offset relates to attentional state indirectly through its association with {rho} (Structural Equation Modeling: power[-&gt;]state {beta} = -0.118, p = 0.002; {rho}[-&gt;]state {beta} = 0.246, p < 0.001; offset[-&gt;]{rho} {beta} = -0.222, p < 0.001). Our study provides a unified neurophysiological framework for understanding how spontaneous neural activity can drive attentional fluctuations and performance variability, with implications for research on attention, learning, and neuropsychiatric disorders.

Light-sheet fluorescence microscopy enables deep optical sectioning of large, cleared biological tissues. However, effective clearing of collagen-rich tissues remains a persistent technical challenge. Moreover, standardized workflows integrating three-dimensional imaging with computational analysis of collagen architecture are currently unavailable. Here, we present an integrated pipeline combining optimized tissue clearing, volumetric light-sheet imaging, and deep learning-based feature extraction of collagen architecture. Using experimental desmoid tumors as a proxy for collagen-dense tissues, we optimised DISCO-based clearing incorporating Fast Green FCF for collagen labelling. We achieved optical transparency and full 3D visualization of collagen architectures in desmoid tumors, human skin biopsies, and fibrotic mouse lung & liver tissues, including FFPE samples. Using the Benchtop mesoSPIM platform, we acquired high-resolution volumetric datasets and validated multimodal collagen imaging through two-photon microscopy with the Schmidt-Voigt objective. To enable automated feature extraction from these large volumetric datasets, we developed ColNet, a U-Net model for automated collagen fiber segmentation. ColNet demonstrated robust generalization across diverse human and mouse tissues without retraining or hyperparameter adjustment. This integrated workflow provides a foundation for future quantitative assessment of cell-extracellular matrix dynamics in a fibrotic context.

Halisarca dujardinii is a marine sponge known for its ability to completely regenerate via cell reaggregation. Here we present the first chromosome-level genome assembly of H. dujardinii, generated using Oxford Nanopore long reads, Illumina short reads, and Hi-C data. The final assembly spans 226.5 Mbp and is organized into 21 chromosome-scale scaffolds, representing the full nuclear genome, with an assembly N50 of 10.1 Mbp. In addition, we report the complete mitochondrial genome for H. dujardinii, assembled as a circular molecule and annotated to contain 14 protein-coding genes. We provide a comprehensive genome annotation comprising 14,565 nuclear protein-coding genes, of which 85.6% are functionally annotated, along with repetitive elements and non-coding RNAs. Transcript models were refined using extensive bulk and single-cell RNA sequencing data, enabling accurate annotation of 3' untranslated regions. This new genomic resource provides a valuable foundation for investigating the molecular basis of sponge regeneration, as well as for comparative studies of sponge evolution and marine biology.

Simultaneous EEG-fMRI offers a powerful way to study brain dynamics, but combining the two modalities in a common whole-brain model remains challenging. Here, I developed a fusion modeling framework for resting-state simultaneous EEG-fMRI that emphasized careful multimodal alignment, construction of a stable shared feature space, and cross-validated, reproducibility-based model selection. Using 15 eyes-open resting-state runs from 12 healthy adults in an open simultaneous EEG-fMRI dataset, I constructed a no-lag, 15-TR-minimum fusion dataset comprising 3550 retained TRs and 124.25 min of usable data. A leave-one-subject-out cross-validation sweep supported a parsimonious three-state fusion hidden Markov model. In the final full-data solution, one state emerged as a dominant backbone state with the highest occupancy, strongest persistence, and clearest canonical BOLD network organization. Two lower-occupancy states behaved as transient alternatives: one appeared as a broadly attenuated version of the backbone state, whereas the other showed more selective network reweighting. The states also differed in their descriptive cross-modal BOLD-EEG structure, suggesting that electrophysiological and hemodynamic network expression may align differently across latent brain states. These results provide both a practical whole-brain EEG-fMRI fusion workflow and a biologically interpretable account of low-order resting-state brain dynamics.

Fluorescence lifetime imaging ophthalmoscopy permits in vivo assessment of retinal metabolism but has remained limited by insufficient cellular resolution in the human eye. Here we present adaptive optics-enhanced fluorescence lifetime imaging ophthalmoscopy (AOFLIO), a method for single-cell-resolved, in vivo structural and metabolic imaging of the human retinal pigment epithelium (RPE). Through real-time correction of ocular wavefront aberrations, precisely synchronized adaptive optics reflectance and lifetime image acquisition via a phase-locked loop-based timing architecture, and sub-pixel photon registration that localizes individual autofluorescence photons with high spatial precision, AOFLIO directly resolves the RPE cell mosaic and measures autofluorescence decay using the same photons, enabling direct structural-functional correlation at the single-cell level. We demonstrate single-cell RPE lifetime mapping in healthy subjects and reveal altered metabolic signatures and fine characterization of RPE metabolic in age-related macular degeneration. AOFLIO establishes a platform for cellular-scale metabolic imaging in the living human eye.

Advances in spectral flow cytometry have enabled the simultaneous measurement of dozens of markers across millions of cells within a single experiment. Despite the increasing maximum perplexity achievable in spectral panels, panel design remains constrained. A central obstacle is signal spread-- unmixed fluorescence signal misattributed to unrelated channels--which reduces the resolution of cell populations. Here we introduce the Residual Model, a robust, scalable, and interpretable model-based approach for spread prediction during panel design. The Residual Model integrates statistical features derived from single-color controls and predicts spread under Ordinary Least Squares unmixing, the most widely used unmixing method. We demonstrate its reliable predictive performance across 141 single-color control samples measured on two instruments. To facilitate practical application, we developed the USERM R package, which implements the Residual Model and provides an out-of-box solution for interactive spread prediction and visualization.

Developmental stuttering is usually framed as a sensorimotor disorder, yet it manifests in communicative situations that engage motivational, self-referential, and regulatory processes. In this case-control study, we combined a socio-economic decision task outside the scanner with a socially modulated speech task during fMRI to test how listener presence and self-referential speech shape neural activity in 34 adults who stutter and 32 controls. Both groups valued talking with another person about themselves more than talking with another person about someone else or talking to themselves. On the neural level, listener-directed (vs. private) speech and self-disclosure (vs. guessing the preferences of a famous other) elicited stronger responses in the ventral striatum and medial prefrontal cortex, extending social valuation effects previously reported in fluent speakers to adults who stutter. Within the stuttering group, individual differences revealed a systematic reweighting of socially modulated activation as a function of symptom burden: higher anticipation of stuttering and greater overall impact were associated with stronger engagement of motivational circuitry, greater recruitment of frontal evaluative-control regions, and reduced contextual differentiation within speech-language cortex. Stuttering anticipation and lived experience gradually shift the balance between control, language, and motivational salience-processing systems, contributing to the disorders marked heterogeneity and context sensitivity. These findings indicate system-level signatures of the interaction between social context and symptom severity, rather than isolated motor deficits, in developmental stuttering. More generally, they reveal how recurrent experiences shape brain activity through the interplay between language, motivational and control systems that governs human social interactions.

High-impact physical activity delivers musculoskeletal and systemic health benefits yet remains controversial for people with hip or knee arthroplasties due to concerns about implant loading and longevity. This study provides the first three-dimensional kinematic and kinetic characterisation of high-impact tasks in high-functioning adults with total hip arthroplasty (THA), total knee arthroplasty (TKA), or unicompartmental knee arthroplasty (UKA), compared with healthy controls. High-functioning adults with a joint arthroplasty (THA=11, TKA=4, UKA=3) participated. Healthy comparison data (n=70) were obtained from prior studies adopting the same analytical framework. They completed running, 45{degrees} change of direction (COD), countermovement jumps (CMJs), and unilateral/bilateral hopping. An 8-segment biomechanical model was used to quantify 3D joint angles and internal moments. These were time-normalised and modelled using Generalized Additive Models with covariates for group, age, and task intensity. THA had greater hip adduction angles during running 3.65{degrees} (95%CI 0.73{degrees}, 6.57{degrees}) and COD45 14.12{degrees} (95%CI 3.22{degrees}, 25.02{degrees}), compared to controls. THA participants recruited greater hip extensor moment during loading response in running, propulsion during CMJ, and in unilateral and bilateral hops, compared to controls. People with a TKA and UKA exhibited a greater internal rotation angle during running, without a concomitant increase in moments at initial contact and toe-off, compared to healthy controls. Also, TKA and UKA participants experienced reductions in knee extensor moments across all high-impact tasks, such as a 1.86 to 1.89 Nm/kg reduction during running. High-functioning individuals with a joint arthroplasty can perform demanding tasks, but often through movement strategies that may compromise long-term implant durability.

A much-replicated finding in human brain imaging is a distributed "multiple-demand" or MD system, increasing in activity for many kinds of cognitive demand, and centrally involved in cognitive control. MD regions are proposed to encode a distributed mental model of critical task events, bound together in the roles and relationships needed to direct action selection. Though previous data hint at a corresponding network in the macaque, there has been no direct comparison to human data. Here we used functional magnetic resonance imaging to measure whole brain activation in a multi-step saccadic maze task, compared to a control requiring similar moves but without goal-based decisions. Human data were a close match to the canonical MD network, extended to include adjacent regions and in particular much of the canonical dorsal attention network. Monkey data suggested correspondences in dorsomedial frontal, lateral and medial parietal, insula/orbitofrontal and posterior temporal cortex. In lateral frontal cortex there was just a single, largely dorsal activation patch, in contrast to multiple distinct human patches. In macaque as in human, together with previous data, our findings suggest an extended and strongly interconnected brain network recruited by increased cognitive challenge.

Cryogenic electron microscopy (cryo-EM) has become a widely used technique for determining the three-dimensional structures of biological macromolecules. Despite its advantages, building accurate structural models from cryo-EM data remains challenging, particularly at non-atomic resolutions. Here, we present CryoZeta, a de novo structure modeling program that leverages a diffusion-based generative deep neural network to integrate cryo-EM map density features with a biomolecular structure prediction pipeline similar to Alphafold3. By jointly leveraging sequence information and density-based features, CryoZeta generates highly accurate structural models that are consistent with the experimental map density. Evaluated on benchmark datasets covering protein complexes, protein-nucleic acid assemblies, and nucleic acid-only systems at resolutions up to 10 [A], CryoZeta consistently outperforms existing cryo-EM modeling methods in atomic accuracy. These results highlight the benefits of directly incorporating cryo-EM density into modern structure prediction pipelines and establish the method as a robust tool for automated, high-fidelity modeling from cryo-EM maps.