bioRxiv Subject Collection: Neuroscience's Journal

Microglia and myeloid cell populations of the developing mouse retina
Microglia make important contributions to central nervous system (CNS) development, but the breadth of their distinct developmental functions remain poorly understood. The mouse retina has been a key model system for understanding fundamental mechanisms controlling assembly of the CNS. To gain insight into where and how microglia might influence retinal development, here we identified molecularly unique myeloid cell populations that are selectively present during development, and characterized their anatomical locations. Development-specific transcriptional states were identified using single-cell and single-nucleus RNA-sequencing (RNA-seq) of purified retinal myeloid cells across multiple timepoints. Transcriptional states were validated in vivo by histological staining for key RNA and/or protein markers. Several of these development-specific myeloid populations have been described before in brain RNA-seq atlases but not validated in vivo, while others are unique to our retinal dataset. We identify two closely-related microglial populations, labeled by the Spp1 and Hmox1 genes, that are distinguished mainly by activity of the NRF2 transcription factor. Both types are present selectively within the developing retinal nerve fiber layer where they engulf neurons and astrocytes undergoing developmental cell death. Hmox1+ microglia were also localized selectively at the wavefront of developing vasculature during retinal angiogenesis, suggesting that developmental events associated with angiogenesis modulate NRF2 activity and thereby induce microglia to switch between the Spp1+ and Hmox1+ states. Overall, our results identify transcriptional profiles that define specific populations of retinal microglia, opening the way to future investigations of how these programs support microglial functions during development.

Leukemia Inhibitory Factor as a late-stage treatment for delayed white matter loss in concussive head injury
Leukemia Inhibitory Factor (LIF) is an injury-induced cytokine that peaks 48 hours after a traumatic brain injury (TBI). Juvenile LIF haplodeficient mice exhibit desynchronized glial responses, increased neurodegeneration, decreased axonal conductivity and behavioral deficits after a concussive head injury. Given the necessity of LIF during the acute recovery phase after injury, we hypothesized that intranasal (IN) LIF treatment would prevent neurodegeneration when administered during the chronic recovery period from a mild TBI (mTBI). Here, young adult male CD1 mice were subjected to a midline, closed-head frontal cortex injury using a flat metal impactor with a 3mm tip to induce a mTBI. In the 6-8 weeks post-mTBI, known to precede axonal atrophy in this mTBI model, two doses of 40 ng and 100 ng of LIF were administered twice daily, 5 days/week for two consecutive weeks. Sensorimotor functions were assessed at 4 and 8 weeks post mTBI, followed by ex-vivo brain magnetic resonance imaging at 9.4T and histopathology. mTBI mice showed sensorimotor deficits at 4 weeks, which worsened by 8 weeks post-injury. IN-LIF treatment prevented the progressive sensorimotor loss seen in the vehicle-treated controls. Increased mean diffusivity (MD) and decreased fractional anisotropy (FA) were observed in the corpus callosum and prefrontal cortex of mTBI brains. In a dose-dependent manner, IN-LIF prevented the mTBI-induced MD increase and FA decrease. Histologically, there was significantly less astrogliosis, microgliosis and axonal injury in the IN-LIF treated mice vs. controls. These results support the therapeutic potential of IN-LIF to reduce delayed neurodegeneration and improve neurological outcomes after mTBIs.

Unreal? A Behavioral, Physiological, and Computational Model of the Sense of Reality
An intriguing aspect of the human mind is our knowledge that our perceptions may be false. Our frequent exposure to non-veridical perceptions such as those found in dreams, illusions and hallucinations cause us to examine the actuality of our sensory experiences. As such humans continuously monitor the veridicality of their perceptions in a process termed the Sense of Reality (SoR). Moreover, the Sense of Reality is a central criterion in the assessment of neurological and psychiatric health. The scientific study of hallucinatory experiences has been hindered by their transitory, subjective, ineffable nature and by the fact that they typically co-occur with psychiatric, medical or psychedelic states. Despite the critical role of SoR in daily life and in pathologies little is known regarding its cognitive, physiological and computational underpinnings. Here we employed a novel immersive virtual reality paradigm to induce Virtual Hallucinations (VH) simulating the phenomenology of visual hallucinations found in psychiatric, neurological, and pharmacological conditions. Combining psychophysics, physiological recordings and computational modeling in one exploratory (n = 31) and one preregistered experiment (n = 32) we examined responses to VH of varying magnitudes and domains. Judgments of SoR varied depending on the magnitude and domain of VH. These were accompanied by distinct motor, pupillometric and cardiac responses. Finally, sense of reality can be well explained by a computational model in which decisions of reality are based on comparison of current sensory experience to an internal model of the world. Our results shed some light on the age-old question: how do we know what is real?

Mapping individualized multi-scale hierarchical brain functional networks from fMRI by self-supervised deep learning
The brain's multi-scale hierarchical organization supports functional segregation and integration. Characterizing the hierarchy of individualized multi-scale functional networks (FNs) is crucial for understanding these fundamental brain processes. It provides promising opportunities for both basic neuroscience and translational research in neuropsychiatric illness. However, current methods typically compute individualized FNs at a single scale and are not equipped to quantify any possible hierarchical organization. To address this limitation, we present a self-supervised deep learning (DL) framework that simultaneously computes multi-scale FNs and characterizes their across-scale hierarchical structure at the individual level. Our method learns intrinsic representations of fMRI data in a low-dimensional latent space to effectively encode multi-scale FNs and their hierarchical structure by optimizing functional homogeneity of FNs across scales jointly in an end-to-end learning manner. A DL model trained on fMRI scans from the Human Connectome Project successfully identified individualized multi-scale hierarchical FNs for unseen individuals and generalized to two external cohorts. Furthermore, the individualized hierarchical structure of FNs was significantly associated with biological phenotypes, including sex, brain development, and brain health. Our framework provides an effective method to compute multi-scale FNs and to characterize the inter-scale hierarchy of FNs for individuals, facilitating a comprehensive understanding of brain functional organization and its inter-individual variation.

Sulfotransferase 4A1 (SULT4a1): A Novel Neuroprotective Protein in Stroke
SULT4a1, a member of the cytosolic sulfotransferase family, is predominantly expressed in neurons and plays potentially vital roles in regulating neural survival and function. SULT4a1 protects against mitochondrial dysfunction and oxidative stress. SULT4a1 levels decrease in experimental stroke models and may play a critical neuroprotective role in mitigating neuronal injury caused by oxygen-glucose deprivation (OGD) and ischemic stroke, as shown in a transient middle cerebral artery occlusion (tMCAO) mouse model. In this study, we investigated the neuroprotective role of SULT4a1 in OGD and tMCAO and highlighted its expression pattern and involvement in maintaining mitochondrial function and reducing oxidative stress, two early pathophysiological features in stroke and related neuronal injury. Our data show that decreased SULT4a1 expression in OGD conditions and in the tMCAO mouse brain leads to enhanced neuronal damage, emphasizing the importance of SULT4a1 in preserving neuronal integrity. Loss of SULT4a1 alone was sufficient to decrease mitochondrial function in mouse cortical neurons. Notably, overexpression of SULT4a1 preserved mitochondrial function, reduced the loss of mitochondrial membrane potential, and diminished oxidative stress, as evidenced by lower reactive oxygen species (ROS) production and reduced protein carbonylation. These results indicate that modulating SULT4a1 expression in stroke could offer a promising strategy for preventing neuronal damage. Indeed, overexpression of SULT4a1 via stereotaxic injection of AAV9 into the mouse brain mitigated tMCAO-related brain injury and functional deficits over time. The findings of this study indicate that SULT4a1 may protect neurons in stroke and related brain injury, possibly by maintaining mitochondrial function and redox homeostasis through mechanisms that are still unknown. It is likely that SULT4a1 regulates neuroprotective processes in both the mitochondria and the cytosol. However, further research is needed to clarify the specific molecular pathways involved in its neuroprotective function.

Bioinformatics-Driven Identification and Prioritization of PTSD Targets Based on Published Multi-omic Data
Efforts in recent years to uncover neurobiological mechanisms underlying post-traumatic stress disorder (PTSD) have yielded an expanding candidate pool of targets from genomic and transcriptomic data. However, not all candidates are disease-causing, related to pathological mechanisms, clinically relevant, nor druggable by conventional means. An effective method to systematically identify and prioritize high-con[fi]dence, high-impact targets in the central nervous system (CNS) is required to de-risk resource-intensive experimental validation of disease mechanisms and accelerate the development of novel treatments. Here, we describe methods and implementation of a novel 3-phased, biologically rationalized, and quantitative prioritization strategy to identify and rank PTSD-associated targets based on con[fi]dence of association to PTSD and estimated CNSrelevant pathogenicity. Phase 1 was designed to identify and advance targets con[fi]dently associated with PTSD through their expression in CNS tissues. Putative targets derived from 29 transcriptomic and genomic analyses of PTSD were evaluated for: 1. Replication in independent cohorts, 2. Observation of differential expression in PTSD CNS tissues, and 3. Demonstration of consistent direction of effect. This strategy resulted in 177 targets that passed criteria for advancement. Phase 1-selected targets were enriched for PTSD relevant traits including irritability, emotional symptoms, and insomnia (FDR <0.05). Phase 2 advanced targets with additional evidence of association to pathological CNS phenotypes. DisGeNET gene-disease association scores were applied to each Phase 1-selected target to assign a con[fi]dence score indicating that a target was associated with CNS-relevant pathology using criteria for moderate or strong evidence of CNS disease association. Phase 2 advanced 55 of the 177 (31.1%) targets. The Phase 2 target pool was enriched for CNS phenotypic abnormalities (FDR<0.05). Finally, Phase 3 enabled target prioritization by annotating targets with a composite pathogenicity score. Components of the pathogenicity score included metrics derived from drug trial databases, predicted loss-of-function intolerance, and connectivity within a protein-protein interaction network de[fi]ned by PTSD-associated targets. The resulting 55 targets were ultimately prioritized by the sum of Phase 2 and Phase 3 scores, where top-ranked targets had strong evidence to support both association with PTSD in brain and high pathogenicity estimates in a CNS-relevant context. Biologically, top-ranked targets implicate transmitter systems (GABA, histamine, and estrogen), structural regulation of neurites, and protein homeostasis. Future work will be required to experimentally validate the utility of the high priority PTSD targets we identi[fi]ed as well as to demonstrate the general applicability of this methodology. Ultimately, we anticipate that the three phased approach will enable ef[fi]cient de-risking of PTSD and other poorly understood CNS disorders.

Scalp EEG predicts intracranial brain activity in humans
Inferring deep brain activity from noninvasive scalp recordings remains a fundamental challenge in neuroscience. Here, we analyzed concurrent scalp and intracranial recordings from 1918 electrode contacts across 20 patients affected by drug-resistant epilepsy undergoing intracranial depth electrode monitoring for pre-surgical evaluation to establish predictive relationships between surface and deep brain signals. Using regularized and cross-validated linear regression within subjects, we demonstrate that scalp recordings can predict spontaneous intracranial activity, with accuracy varying by region, depth, and frequency. Low-frequency signals (<12 Hz) were most predictable, with our models explaining approximately 10% of intracranial signal variance across contacts. Prediction accuracy decreased with contact depth, particularly for high-frequency signals. Using Bayesian modeling with leave-one-patient-out cross-validation, we observed generalizable prediction of activity in mesial temporal, prefrontal, and orbitofrontal cortices, explaining 10-12% of low-frequency signal variance. This scalp-to-intracranial mapping derived from spontaneous activity was further validated by its correlation with scalp responses evoked by direct electrical stimulation. These findings support the development of improved inverse models of brain activity and potentially more accurate scalp-based markers of disease and treatment response.

Optimizing Biophysical Large-Scale Brain Circuit Models With Deep Neural Networks
Biophysical modeling provides mechanistic insights into brain function, from single-neuron dynamics to large-scale circuit models bridging macro-scale brain activity with microscale measurements. Biophysical models are governed by biologically meaningful parameters, many of which can be experimentally measured. Some parameters are unknown, and optimizing their values can dramatically improve adherence to experimental data, significantly enhancing biological plausibility. Previous optimization methods -- such as exhaustive search, gradient descent, evolutionary strategies and Bayesian optimization -- require repeated, computationally expensive numerical integration of biophysical differential equations, limiting scalability to population-level datasets. Here, we introduce DELSSOME (DEep Learning for Surrogate Statistics Optimization in MEan field modeling), a framework that bypasses numerical integration by directly predicting whether model parameters produce realistic brain dynamics. When applied to the widely used feedback inhibition control (FIC) mean field model, DELSSOME achieves a 2000x speedup over Euler integration. By embedding DELSSOME within an evolutionary optimization strategy, trained models generalize to new datasets without additional tuning, enabling a 50x speedup in FIC model estimation while preserving neurobiological insights. The massive acceleration facilitates large-scale mechanistic modeling in population-level neuroscience, unlocking new opportunities for understanding brain function.

Nociceptor clock genes control excitability and pain perception in a sex- and time-dependent manner
Nociception is critical for pain perception and survival, and begins with the activation of nociceptors, specialized sensory neurons located in the dorsal root ganglia (DRGs). Both sex and circadian rhythms, governed by clock genes, seem to play a significant role in modulating pain perception. However, the potential interaction between circadian rhythms and sex differences in nociception at the peripheral level has been largely overlooked. Here we first report that DRGs from mice express clock genes in a time- and sex-dependent manner. Using whole-cell recordings in whole-mounted DRGs and optogenetic stimulation of Nav1.8-expressing neurons, we demonstrate that male nociceptors exhibit reduced excitability during the night, while female nociceptor excitability remains stable across time points. Disruption of the core clock gene Bmal1 in Nav1.8-expressing neurons not only diminished nociceptor activity but also abolished the nighttime reduction in heat sensitivity, highlighting a pivotal role for the molecular clock in regulating nociception. Transcriptomic analyses, voltage-clamp recordings, and pharmacological experiments identified the voltage-gated chloride channel ClC-2, controlled by Bmal1, as a key mediator for the observed fluctuations in male nociceptor excitability. This work opens new avenues for chronobiology-inspired strategies in pain management, tailored to sex-specific mechanisms.

Visual pathway origins: an electron microscopic volume for connectomic analysis of the human foveal retina
With over 1014 synapses, the human brain presents a seemingly insurmountable challenge to a nano-scale circuit-level understanding of its diverse neural systems. The foveal retina however presents a feasible site for a complete connectome of a key human CNS structure. Foveal cells and circuits are miniaturized and compressed to densely sample the visual image at highest resolution to initiate form, color and motion perception. Here we use computational methods first applied to the fly brain to provide a draft connectome of all neurons in a foveal volume. We found synaptic connections distinct to humans linking short-wavelength sensitive cones to color vision pathways. Moreover, by reconstructing excitatory synaptic pathways arising from cone photoreceptors we found that 96% of foveal ganglion cells contribute to only three major pathways to the brain. This new resource reveals unique features of a human neural system and opens a door to its complete connectome.

Competition between tool and hand motion impairs movement planning in limb apraxia
Tool use is a complex motor planning problem. Prior research suggests that planning to use tools involves resolving competition between different action representations. We reasoned that competition may be exacerbated with tools for which the motions of the tool and the hand are incongruent, such as pinching the fingers to open a clothespin. If this hypothesis is correct, then we should observe particularly marked deficits in planning the use of incongruent as compared to congruent tools in individuals with limb apraxia, a common disorder after left-hemisphere stroke (LCVA) that is associated with abnormal action competition. In a first experiment, we asked 34 individuals with chronic LCVA and 16 matched neurotypical controls to use novel tools that we developed in which the correspondence between the motions of the hand and tool-tip were either congruent or incongruent. Individuals with LCVA also completed background assessments to quantify apraxia severity. We observed increased planning time for incongruent as compared to congruent tools as a function of the severity of apraxia, particularly when the tools were first introduced. Exploratory lesion-symptom mapping analyses revealed that lesions to posterior portions of the tool-use network were associated with impaired planning for incongruent tools. A second experiment on the same individuals with LCVA revealed that the ability to demonstrate the use of conventional tools was impaired for tools rated as more incongruent by a normative sample. Together, these findings suggest tool-hand incongruence evokes action competition and influences the tool-use difficulties experienced by people with apraxia.

Connecting the Dots: A Computational Framework linking Molecular Regulation, Synaptic Plasticity, and Brain Disorders
The complexity of brain function emerges from multiscale interactions spanning molecular, synaptic, and circuit levels. While substantial progress has been made in each domain, a unified framework linking molecular perturbations to large-scale brain dysfunctions remains elusive. Here, we introduce a computational framework that bridges molecular regulation and neuronal plasticity, providing mechanistic insights into disease pathogenesis. By creating a Boolean network model of molecular regulatory pathways, we identify two stable states-synaptic potentiation and depression-governed by AMPA receptor trafficking and consistent with Hebbian learning principles. Through in silico knockout simulations mimicking genetic anomalies, our model demonstrates how molecular perturbations cascade across scales, disrupting synaptic plasticity and ultimately driving diverse neurological dysfunctions. The model's predictions align with a broad range of experimental and clinical data. Our framework establishes a mechanistic foundation for understanding the hierarchical causality of brain disorders, offering a powerful tool for dissecting disease mechanisms and informing targeted therapies.

Optogenetic disruption of neural dynamics in the prefrontal cortex impaired spatial learning
Cortical neural activity is highly dynamical at several temporal scales, property that has been postulated to be critical for the emergence of specific patterns supporting cognitive operations. During spatial learning, task-associated activity patterns gradually develop in the medial prefrontal cortex (mPFC) as the subject acquires experience. If neural activity dynamics is required in the mPFC for spatial learning is still unclear. Here we show that optogenetic entrainment of neural activity the mPFC disrupted local oscillatory and single-neuron dynamics. When applied during spatial training, optogenetic entrainment impaired behavioral performance and navigation strategy progression, a hallmark of spatial learning supported by the mPFC. Also, optogenetic entrainment blocked the emergence of learning-related activity patterns such as cross-frequency coupling and firing patterns signaling efficient goal approaching. Importantly, during memory retrieval, training-stimulated mice showed impaired performance in the absence of optogenetic stimulation. This evidence show that neural activity dynamics in the mPFC is crucial for spatial learning.

Strategies used by two memories to share space in a common neural network
How distinct memories are encoded into the same network space without destructive interference is not well-understood. Here, we utilized Tritonia diomedea's escape swim network to explore how two sequentially acquired forms of non-associative learning, sensitization and habituation, are encoded into the same network. Behavioral experiments showed them to alter non-identical sets of behavioral features, suggesting they utilize somewhat independent sites of plasticity within the network. Large-scale voltage-sensitive dye recordings revealed two findings. First, both forms of learning, which occur sequentially in the 10-trial training protocol used, act to produce a change in the number of pedal neurons firing during the dorsal phase of the motor program, with sensitization producing an increase, and habituation a decrease in their number. The number of neurons participating in the ventral phase was unaffected. Second, sensitization produced an enhancement of burst intensity specific to the ventral phase neurons, while habituation was associated with a decrease in burst intensity in both phases. Using injected current pulses, intracellular recordings revealed that sensitization acted to increase the excitability of neurons firing in both phases, whereas habituation only acted to reduce excitability in ventral phase neurons. These excitability changes were associated with different mechanisms - reduced spike frequency accommodation in the ventral phase neurons, and depolarization of the resting potential in the dorsal phase neurons. These findings of partially different storage sites and mechanisms for two different non-associative memories illuminate a potential network strategy for minimizing destructive interference when storing multiple memories into the same network.

YPEL controls synapse development through p62-Nrf2 antioxidant pathway in Drosophila neuromuscular junction
The Yippee-like (YPEL) genes are highly conserved among all eukaryotic species, yet the molecular and cellular pathways that YPEL regulate are poorly understood. Human and animal studies suggest that YPEL is involved in nervous system functions. Here we report that YPEL is necessary for synapse development in neuromuscular junction and motor functions in Drosophila. YPEL interacts with the Drosophila p62, Refractory to sigma P, which forms cytoplasmic proteinaceous bodies for selective autophagy and signaling. YPEL overexpression decreased p62 bodies, while increased p62 bodies were observed in YPEL mutant neurons. Suppressing p62 bodies by reducing p62 gene dosage significantly alleviated both synaptic and locomotion defects in YPEL mutants, suggesting that YPEL acts through p62 for synapse development. On the other hand, suppressing p62 bodies via autophagy did not restore synapse development in YPEL mutants. Interestingly, reduced levels of reactive oxygen species were found in YPEL mutants, which is consistent with the role of p62 in promoting the nuclear factor erythroid 2-related factor 2 (Nrf2) antioxidant pathway. Overexpressing Drosophila Nrf2, Cap n collar (Cnc), phenocopied the synaptic and locomotion deficits in YPEL mutants. Importantly, both synaptic and locomotion defects were completely rescued by knocking down Cnc in YPEL mutant motor neurons. Taken together, our study demonstrates that YPEL negatively controls p62 - Nrf2 antioxidant pathway for neuromuscular synapse development and locomotion.

Cell-type specific repertoire of responses to natural scenes in primate retinal ganglion cells
At least 20 distinct retinal ganglion cells (RGC) types have been identified morphologically in the primate retina, but our understanding of the distinctive visual messages they send to various targets in the brain remains limited. Here, we use large-scale multi-electrode array recordings to examine how multiple functionally-distinct RGC types in the macaque retina respond to flashed natural images. Responses to white noise visual stimulation were used to functionally identify 936 RGCs of 12 types in three recordings. Each cell type was confirmed by the mosaic organization of receptive fields, and 7 cell types were cross-identified between recordings. The average kinetics of light response in each RGC type as well as the repertoire of distinct firing patterns that each type produces were examined across thousands of natural images. The kinetics of the average response across images were highly stereotyped among cells of each cell type and distinct for cells of different types. Moreover, the full repertoires of firing patterns produced by different cell types, assessed by their latency and duration, were generally quite distinct with only a few exceptions. Together these data provide an overview of the range of responses to natural images transmitted from the primate retina to the brain.

Learning evoked centrality dynamics in the schizophrenia brain: Entropy, heterogeneity and inflexibility of brain networks
BackgroundBrain network dynamics are responsive to task induced fluctuations, but such responsivity may not hold in schizophrenia (SCZ). We introduce and implement Centrality Dynamics (CD), a method developed specifically to capture task-driven dynamic changes in graph theoretic measures of centrality. We applied CD to fMRI data in SCZ and Healthy Controls (HC) acquired during a learning paradigm.

MethodsfMRI (3T Siemens Verio) was acquired in 88 participants (49 SCZ). Time series were extracted from 246 functionally defined cerebral nodes. We applied a dynamic widowing technique to estimate 280 partially overlapping connectomes (30,135 region-pairs in each connectome). In each connectome we calculated every nodes Betweenness Centrality (BC) before building 246 unique time series (representing a nodes CD) from a nodes BC in successive connectomes. Next, in each group nodes were clustered based on similarities in CD.

ResultsClustering gave rise to fewer sub-networks in SCZ, and these were formed by nodes with greater functional heterogeneity. These sub-networks also showed greater ApEn (indicating greater stochasticity) but lower amplitude variability (suggesting less adaptability to task-induced dynamics). Higher ApEn was associated with worse clinical symptoms.

LimitationsCentrality Dynamics is a new method for network discovery in health and schizophrenia but will need further extension to other tasks and psychiatric conditions, before we achieve a fuller understanding of its promise.

ConclusionThe brains functional connectome is not static under task-driven conditions, and characterizing the dynamics of the connectome will provide new insight on the dysconnection syndrome that is schizophrenia. Centrality Dynamics provides novel characterization of task-induced changes in the brains connectome and shows that in the schizophrenia brain, learning-evoked sub-network dynamics were less responsive to learning evoked changes and showed greater stochasticity.

Decoding the Moving Mind: Multi-Subject fMRI-to-Video Retrieval with MLLM Semantic Grounding
Decoding dynamic visual information from brain activity remains challenging due to inter subject neural heterogeneity, limited per subject data availability, and the substantial temporal resolution gap between fMRI signals (0.5 Hz) and video dynamics (30 Hz). Current approaches face persistent challenges in addressing these temporal mismatches, demonstrate limited capacity to integrate subject specific neural patterns with shared representational frameworks, and lack adequate semantic granularity for aligning neural responses with visual content. To bridge these gaps, we propose a framework addressing these limitations through three innovations: (1) a Dynamic Temporal Alignment module that resolves temporal mismatches via exponentially weighted multiframe fusion with adaptive decay coefficients; (2) a Brain Mixture of Experts architecture that combines subject specific extractors with shared expert layers through parameter efficient tri modal contrastive learning; and (3) a Multi-perspective Semantic Hyper Anchoring module that resolves cross subject attention bias via multi-dimensional semantic decomposition, leveraging multimodal LLMs for fine grained video semantic extraction enabling the model to match individual attention patterns as different subjects naturally focus on distinct aspects of the same visual stimulus. This module boosts Top 10/Top 100 retrieval by 17.7%/6.6%. Experiments on two video fMRI datasets demonstrate state of the art performance, with 39%/30% improvements in Top 10/Top 100 accuracy over single subject baselines and 27% gains against multi subject models. The framework exhibits remarkable few shot adaptability, retaining 97% performance when using only 10% training data for new subjects. Visualization analysis confirms this generalization capability stems from effective disentanglement of subject specific and shared neural representations.

Gaze-centered spatial coding of touch on a hand-held tool
Humans possess the remarkable ability to project tactile sensations outside their body and onto a hand-held tool that they are using. Despite nearly a century of research, the computations underlying this projection have not been adequately addressed. In the present study, we employed model-driven psychophysics to fill this gap. We hypothesized that tool-based sensory projection involves the remapping of touch from sensory feedback in the hand into an egocentric coordinate system. We first formalized the computational steps underlying tactile remapping. We designed a novel tool-sensing experiment that allowed us to rigorously test this model. In this task, participants contacted an object with a hand-held rod and then judged whether the object was above or below where they were currently looking. This comparison would only be possible if touch on the tool was projected outside the hand. Crucially, both object location and gaze position varied independently, allowing us to characterize the hand-to-space-to-gaze remapping process. Model-based curve fitting provided strong evidence that all participants in our task projected touch outside their body and into gazecentered coordinates. Crucially, the resolution of this projection was similar to what has been found for touch on the body. These findings provide the first step towards characterizing the computations underlying the spatial projection of touch on external objects, highlighting the incredible versatility of the sensorimotor system.

Differential Effects of Short-term and Long-term Deep Brain Stimulation on Striatal Neuronal Excitability in a Dystonia Animal Model
Deep brain stimulation (DBS) is, by now, one of the standard treatment options for movement disorders like dystonia or Parkinsons Disease. Although its clinical effectiveness is established, the exact mechanisms by which DBS influences neural motor networks are not fully understood. The present study explores the development of adaptive network mechanisms with DBS in the dtsz hamster model, an in-vivo model exhibiting spontaneous dystonic episodes, by comparing functional impacts of short-term and long-term DBS on medium spiny neurons (MSNs) and synaptic transmission in the striatum. In this electrophysiological study, we uncovered contrasting changes in neuronal excitability and synaptic dynamics following short-term versus long-term DBS. Short-term DBS enhanced neuronal firing responses, while long-term DBS diminished them. Regarding synaptic alterations, both short-term and long-term DBS significantly shifted spontaneous EPSC occurrences to longer intervals, with this effect, however, being more pronounced in short-term DBS, leading to a significant decrease in mEPSC frequency. Notably, acetylcholine application effectively reversed this effect, restoring mEPSC frequency more efficiently again in tissue subjected to short-term DBS compared to long-term DBS. These observations indicate that the therapeutic benefits of DBS in dystonia may involve both immediate and adaptive mechanisms, which has implications for improving stimulation parameters and treatment protocols. The findings shed light on the temporal specificity of DBS effects and highlight the importance of understanding synaptic mechanisms to enhance therapeutic outcomes for dystonia patients.

Connectomic traces of Hebbian plasticity in the entorhinal-hippocampal system
The key model of how we learn and memorize is Hebbian learning in the hippocampus, via long-term potentiation of synapses, allowing the storage of associations, linkage to places, and their consolidation into imprinted episodes. Learning is therefore expected to change synaptic weights. With the notion of hippocampal circuits being the primary site of learning in the mammalian brain, it has been assumed that all synapses in these circuits are constantly exposed to synaptic plasticity, and possibly in a learned state. However, a testing of these hypotheses is so far missing. In particular, the systematic mapping of synaptic weight distributions, and their relation to Hebbian preconditions has not been achieved yet in the hippocampal-entorhinal system. Here, we report such a systematic connectomic mapping of synaptic weight distributions and their relation to same-axon same-dendrite paired synaptic configurations across the hippocampal-entorhinal system. By analyzing millions of synapses and tens of thousands of paired synaptic configurations from 3D EM-based automated circuit reconstructions in hippocampal areas CA3, CA1, and layers 2 and 3 of the medial entorhinal cortex (MEC), we found systematic and unique synaptic weight distributions, with almost 50% (but not 100%) of synaptic weights in CA1 being in a Hebbian-consistent state, CA3 uniquely exhibiting small synaptic weights with indications of learned states, and MEC resembling previous data from other isocortices, with only up to 20% learned synaptic configurations. We further analyzed the sublayer-specificity of these weight distributions, finding the molecular layer of CA1 and the lower layers of CA3 being the most unique site of potential learning. Together, this data provides a first systematic synaptic weight analysis of the key neuronal system involved in memory formation in mammalian brains.

High-throughput measurements of neuronal activity in single human iPSC-derived glutamate neurons
Induced pluripotent stem cell (iPSC)-derived neurons provide a promising platform for studying neuronal function and modeling central nervous system (CNS) diseases. However, functional analysis of large populations of iPSC-derived neurons has been challenging. Here, we developed a high throughput strategy targeting N-methyl-D-aspartate receptors (NMDA-R) to enhance neuronal activity and reveal functional phenotypes in human iPSC-induced glutamatergic neurons (iGlut). Using a genetically encoded calcium indicator (GCaMP8f), we first demonstrate that using artificial cerebrospinal fluid (ACSF) lacking Mg2+ (Mg2+-free) significantly increases neuronal firing, and that firing is enhanced by a potentiator (glycine) but inhibited by the NMDA-R antagonist AP-V. Similarly, multi-electrode array (MEA) recordings also show robust firing in Mg2+-free ACSF. Lastly, single-cell patch-clamp electrophysiology confirms the high firing rates in Mg2+-free ACSF across multiple iPSC donor lines and also reveals iPSC donor-specific tonic and bursting firing phenotypes. This new methodology provides a scalable, high-throughput method to study neuronal activity in iGlut neurons while preserving single-cell resolution. The strategy also reveals different functional phenotypes, enabling detailed characterization of iGlut neurons in diverse applications such as CNS disease modeling and drug screening. These findings establish a versatile framework for future studies of neuronal network dynamics and individual excitability in iPSC-derived neuronal cultures.

Harmonizing Inter-Site Differences in T1-Weighted Images Using CycleGAN
IntroductionWhen neuroimaging studies using magnetic resonance imaging (MRI) are conducted across multiple centers, they often encounter inter-site differences in MRI equipment and protocols leading to biases and confounding effects in MRI measurements. There are existing techniques for correcting these site effects, i.e., harmonization, but they have limitations, including the need for preprocessing of MRI data, which involves processes such as spatial normalization. Deep learning-based methods have emerged as potential alternatives that can handle site effects without the need for preprocessing steps. In this study, we propose a novel method based on the generative adversarial network (GAN) framework, CycleGAN, that effectively addresses inter-site differences in T1-weighted images with minimal preprocessing requirements. We compare the harmonization efficacy of CycleGAN with that of the commonly used method ComBat.

MethodsWe trained the proposed CycleGAN method and the comparative ComBat method using data from 40 subjects at each of two sites. To evaluate the effectiveness of the two methods, we used data from nine subjects who underwent imaging at both sites. We assessed harmonization performance at the image level using the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). Additionally, we evaluated harmonization results at the feature level by analyzing regional cortical thickness and volume data. Cohens d was employed to quantify the differences between feature values.

ResultsAt the image level, the ComBat method decreased the median baseline SSIM value from 0.86 (interquartile range [IQR], 0.02) to 0.84 (IQR, 0.02), whereas the proposed CycleGAN method maintained the SSIM value at 0.86 (IQR, 0.02). For PSNR, the baseline value was 18.33 (IQR, 1.78), which decreased to 15.30 (IQR, 2.20) after applying ComBat, but increased to 19.58 (IQR, 3.12) with the proposed CycleGAN method. These findings indicate that CycleGAN preserved the structural and signal similarity of the images. At the feature level, the effect size for cortical thickness decreased from 0.97 (IQR, 1.79) to 0.91 (IQR, 1.54) after applying ComBat, whereas the proposed CycleGAN method yielded an effect size of 1.05 (IQR, 1.14). For cortical volume, the effect size decreased from 0.95 (IQR, 1.78) to 0.69 (IQR, 1.00) after applying ComBat, and decreased to 0.88 (IQR, 0.74) with the CycleGAN method. Compared with baseline, Cohens d was significantly lower with both ComBat (p = 0.000002) and CycleGAN (p = 0.028) with no significant difference between the two methods, indicating similar performance of the two methods under the study conditions.

ConclusionThe results underscore the ability of CycleGAN to harmonize data without explicit normalization and emphasize the potential impact of the normalization process on harmonization procedures.

Our findings suggest that CycleGAN holds promise as a harmonization technique in multi-site neuroimaging studies.

A data-driven approach to identifying and evaluating connectivity-based neural correlates of consciousness
Identifying the neural correlates of consciousness remains a major challenge in neuroscience, requiring theories that bridge between subjective experience and measurable neural correlates. However, theoretical interpretation of empirical evidence is often post hoc and susceptible to confirmation bias. Building upon the adversarial collaboration mediated by the COGITATE Consortium, we present a generalizable approach for the data-driven identification, evaluation, and theoretical modeling of connectivity-based neural correlates of consciousness. Using the same magnetoencephalography (MEG) dataset and accompanying pre-registered hypotheses from the COGITATE Consortium, we systematically compared 246 functional connectivity (FC) measures between regions predicted to underlie conscious vision by Integrated Information Theory (IIT) and/or Global Neuronal Workspace Theory (GNWT). We identified a family of FC measures based on the barycenter--tracking the center of mass between two signals--as the top-performing stimulus decoding measures that generalize across regions central to predictions of both IIT and GNWT. To interpret these findings within a theoretical framework, we developed neural mass models that recapitulate the neural dynamics hypothesized to underlie conscious perception by each theory. Comparing simulated barycenter values from these models against empirically measured MEG data revealed that the GNWT-based model, featuring delayed ignition dynamics, better captured observed connectivity patterns than the IIT-based model, which relied on highly synchronized sensory dynamics. Beyond dataset-specific conclusions and limitations, we introduce a framework for systematically identifying and testing candidate neural correlates of consciousness in an unbiased and interpretable manner.

Knockout of P2Y12 receptor facilitates microglia-neuron body-to-body interactions and accelerates prion disease
Microglia continuously monitor neuronal health through somatic purinergic junctions, where microglial processes establish dynamic contacts with neuronal cell bodies. The P2Y12 receptor is a key component of these junctions, essential for intercellular communication between ramified microglia and neurons under homeostatic conditions. While P2Y12 has long been considered a marker of homeostatic microglia, its potential role in reactive microglia during neurodegenerative disease remains largely unexplored. In this study, we demonstrate that P2Y12 deletion significantly reduces microglia-neuron process-to-body contacts in adult mice, consistent with previous findings. However, unexpectedly, P2Y12 loss markedly increases microglia-neuron body-to-body contacts, revealing an alternative mode of microglia-neuron communication independent of P2Y12. In prion-infected mice, P2Y12 expression persists in reactive, amoeboid microglia during advanced disease stages, including those engaging in extensive neuronal envelopment. Notably, P2Y12 loss increases the prevalence of envelopment events and accelerates disease progression. These findings redefine the role of P2Y12 in neurodegeneration, suggesting that its progressive decline lowers the threshold for microglia-neuron body-to-body interactions, ultimately influencing disease trajectory.

Low affinity noradrenergic signaling promotes passive coping during reinforcement behavior
BackgroundThe fundamental decision processes that allow us to avoid negative outcomes (e.g., injury) and seek relevant positive outcomes (e.g., food) can be disrupted by stress. A critical component of the stress response is increased noradrenergic signaling, particularly at lower affinity receptors (1 and {beta}). It has been shown that healthy noradrenergic signaling is essential for cue-driven reward seeking in males. Whether noradrenergic components of the stress response alter negative reinforcement behavior is still unknown.

MethodsWe used manipulated noradrenergic signaling at low affinity (1, {beta}) and high affinity adrenergic receptors (2) during an active avoidance and reward seeking task in male and female rats.

ResultsWe found that modulating noradrenergic activity on low affinity 1 and {beta} receptors but not high affinity 2 receptors decreased reinforcement behavior. Increased 1, but to a greater extent increased {beta} receptor activity shifted negative reinforcement behavior away from active avoidance and towards passive coping by increasing decision thresholds. We found sex differences in the impact of increased 1 and {beta} receptor activity on reward seeking where only males decreased their positive reinforcement behavior.

ConclusionsStress related noradrenergic signaling deceased active avoidance behavior by shifting behavioral strategies towards passive coping. Reward seeking was more robust behavior after noradrenergic modulation in females but not males. Collectively these results identify sex differences in noradrenergic stress response behaviors and highlight low affinity adrenoreceptors as relevant therapeutic targets to mitigate the impact of stress on goal-directed behavior.

The Physiological Component of the BOLD Signal: Impact of Age and Heart Rate Variability Biofeedback Training
Aging is associated with declines in autonomic nervous system (ANS) function, including reduced heart rate variability (HRV), impaired neurovascular coupling, and diminished cerebrovascular responsiveness: factors that may contribute to cognitive decline and neurodegenerative diseases. Understanding how aging alters physiological signal integration in the brain is crucial for identifying potential interventions to promote brain health. This study examines age-related differences in how cardiac and respiratory fluctuations influence the blood oxygenation level-dependent (BOLD) signal, using two independent resting-state fMRI datasets with concurrent physiological recordings from younger and older adults. Our findings reveal significant age-related reductions in the percent variance of the BOLD signal explained by heart rate (HR), respiratory variation (RV), and end-tidal CO2, particularly in regions involved in autonomic regulation, including the orbitofrontal cortex, anterior cingulate cortex, insula, basal ganglia, and white matter. Cross-correlation analysis also revealed that younger adults exhibited stronger HR-BOLD coupling in white matter, as well as a more rapid BOLD response to RV and CO2 in gray matter. Additionally, we investigated the effects of heart rate variability biofeedback (HRV-BF) training, a non-invasive intervention designed to modulate heart rate oscillations. The intervention altered physiological-BOLD coupling in an age- and training-dependent manner: older adults who underwent HRV-BF to enhance HR oscillations exhibited a shift toward younger-like HR-BOLD coupling patterns, while younger adults who trained to suppress HR oscillations showed increased CO2-BOLD coupling. These findings suggest that HRV-BF may help mitigate age-related declines in autonomic or cerebrovascular function. Overall, this study underscores the role of physiological dynamics in brain aging and highlights the importance of considering autonomic function when interpreting BOLD signals. By demonstrating that HRV-BF can modulate physiological-BOLD interactions, our findings suggest a potential pathway for enhancing cerebrovascular function and preserving brain health across the lifespan.

Atypical Alpha Oscillatory EEG Dynamics in Children with Angelman Syndrome
Objectives: Biomarkers of atypical brain development are crucial for advancing clinical trials and guiding therapeutic interventions in Angelman syndrome (AS). Electroencephalography (EEG) captures well-characterized developmental changes in peak alpha frequency (PAF) that reflect underlying neural circuit maturation and may provide a sensitive metric for mapping atypical neural trajectories in AS. Method: We analyzed EEG recordings from 159 children with AS (ages 1 to 15 years) and 185 age-matched typically developing (TD) controls. PAF was quantified using a well-established curve-fitting method applied to 1/f-corrected power spectra. To validate robustness, we further evaluated PAF using an alternative prominence-based peak detection approach across varying detection thresholds. Results: Significant disruptions in PAF were evident in children with AS. While over 90% of EEGs from TD children exhibited a clear alpha peak, fewer than 50% of EEGs from children with AS showed a detectable PAF. Furthermore, when PAF was present, its frequency was significantly lower in AS children and did not show the typical age-related increases observed in TD children. Validation analyses confirmed consistently lower rates of PAF detection in AS across varying sensitivity thresholds, demonstrating the robustness of these results. Conclusions: PAF is a robust and developmentally sensitive marker of disrupted neural maturation in children with Angelman syndrome. As a quantifiable and sensitive measure of neural disruptions in AS, PAF has the potential to complement and enhance existing clinical trial outcome assessments by providing an objective index of underlying brain function. Future analyses will explore individual differences related to PAF in AS, to better understand mechanistic insights to guide targeted therapeutic strategies.

The role of the right language network and the multiple-demand network in verbal semantics: Insights from an Activation Likelihood Estimation Meta-analysis of 561 Functional Neuroimaging Studies
Language processing has been traditionally associated with a network of fronto-parietal and temporal regions in the left hemisphere. Nevertheless, the right language network (frontal, temporal, and parietal regions homologous to the left language network) and the Multiple-Demand Network (MDN) are often involved in verbal semantic processing as well, however their role remain poorly understood. This is in part due to the inconsistent engagement of these latter two networks across linguistic tasks. To explore the factors driving networks recruitment of right language network and MDN during verbal semantic processing, we conducted a large-scale Activation Likelihood Estimation meta-analysis of neuroimaging studies. We examined whether the right language network is influenced by verbal stimulus type (sentences/narratives versus single words/word pairs) and whether this may be due to differences in semantic control demands and/or the presence of social content in the stimuli. Additionally, we investigated whether MDN recruitment depends on external task demands rather than semantic control demands. Our main findings revealed greater engagement of the right language network during semantic processing of sentence/narrative stimuli, with distinct regions reflecting different functions: increased semantic control demands recruit the right inferior frontal gyrus. Instead, social content processing during a semantic task engages the right Anterior Temporal Lobe, as well as the right posterior middle temporal gyrus. Finally, semantic processing engages the MDN, but only when external task (rather than semantic) demands increase.

Epigenetic memory astrocytes are likely an artifact of immune cell contamination
Innate immune memory, in which prior immune stimuli can "train" certain immune cells to respond more aggressively to subsequent challenges, is crucial for immune system plasticity in disease. Lee et al. [1] recently described a similar kind of immune memory state in astrocytes which they termed "epigenetic memory astrocytes". The discovery of astrocytes with immune memory could have tremendous importance in understanding and treating neurological disease. However, the RNA-seq data and in vitro experiments presented by Lee et al. to claim astrocytes possess pro-inflammatory immune memory show signs of immune cell contamination. Further, astrocyte-specific knockout of Ep300, the purported epigenetic regulator of this memory, did not reduce expression of any memory astrocyte signature genes. The FIND-seq signature used to verify the presence of epigenetic memory astrocytes in experimental autoimmune encephalomyelitis (EAE) also shows signs of immune cell contamination, and the cells identified as memory astrocytes in previously published EAE single-cell RNA-seq data are misannotated macrophages. Lastly, we find the purported epigenetic memory astrocytes identified in single-nucleus RNA-seq data of multiple sclerosis (MS) tissue are an artifact of ambient RNA, low quality nuclei, and non-astrocyte contamination. We conclude that the epigenetic memory astrocyte signature is likely driven by immune cell contamination and the existence of astrocyte immunological memory is insufficiently evidenced. We caution that astrocyte transcriptomic, epigenomic, and functional assays must take care to exclude contamination by immune cells, especially when evaluating the potential of astrocytes to perform immunological functions.

Cardiorespiratory fitness and cardiometabolic health are associated with distinct cognitive domains in cognitively healthy older adults.
BackgroundAging is associated with progressive cognitive decline, as well as increased prevalence of cardiometabolic risk factors and reduced cardiorespiratory fitness. In fact, reduced cardiometabolic health and cardiorespiratory fitness are both associated with a decline in cognitive functioning. This study examines the common and distinct contributions of these cardiovascular health factors on cognitive variability across different domains in cognitively healthy older adults.

Methods and ResultsWe apply structural equation modelling (SEM) to model cross-sectional relationships between cardiometabolic health, cardiorespiratory fitness and performance across multiple cognitive domains in an age-restricted sample of healthy older adults from the ACTIVate Study (n=345; 60-70 yrs). Participants completed a series of cognitive and clinical assessments (including brachial blood pressure, heart rate, blood-based metabolic markers). We designed a cognitive model (Model 1) with four latent factors that are differentially impacted by aging (Processing Speed, Executive Function, Verbal Memory and Crystallized Ability) and used it to test effects on cognition of two theory-driven dimensions of cardiovascular health: Cardiorespiratory Fitness (Model 2) and Cardiometabolic Health (Model 3). Model 4 included both predictors and examined their joint and distinct effects on these cognitive domains. When controlling for their joint variance, Cardiometabolic Health and Cardiorespiratory Fitness showed evidence consistent with a double dissociation on cognitive domains. Specifically, cardiorespiratory fitness significantly predicted processing speed (r=0.28, p<0.05) and executive function (r=0.66, p<0.05), but not verbal memory and crystallized ability. In contrast, cardiometabolic health predicted crystallized ability (r=0.31, p<0.05) and verbal memory (r=0.28, p<0.05), but not executive function and processing speed.

ConclusionsThis study shows the first evidence that cardiorespiratory fitness and cardiometabolic health are associated with distinct cognitive domains in a large cross-sectional, age-restricted and high functioning cohort. These findings emphasize the importance of healthy aging approaches that target both health literacy and lifestyle behaviors to promote functional capacity across the lifespan.