Nature Neuroscience

Published by Springer Nature
Online ISSN: 1546-1726
Learn more about this page
Recent publications
Article
2020, The Author(s), under exclusive licence to Springer Nature America, Inc. Recent success in identifying gene-regulatory elements in the context of recombinant adeno-associated virus vectors has enabled cell-type-restricted gene expression. However, within the cerebral cortex these tools are largely limited to broad classes of neurons. To overcome this limitation, we developed a strategy that led to the identification of multiple new enhancers to target functionally distinct neuronal subtypes. By investigating the regulatory landscape of the disease gene Scn1a, we discovered enhancers selective for parvalbumin (PV) and vasoactive intestinal peptide-expressing interneurons. Demonstrating the functional utility of these elements, we show that the PV-specific enhancer allowed for the selective targeting and manipulation of these neurons across vertebrate species, including humans. Finally, we demonstrate that our selection method is generalizable and characterizes additional PV-specific enhancers with exquisite specificity within distinct brain regions. Altogether, these viral tools can be used for cell-type-specific circuit manipulation and hold considerable promise for use in therapeutic interventions.
 
Article
In nature, animals display defensive behaviors that reflect the spatiotemporal distance of threats. Laboratory-based paradigms that elicit specific defensive responses in rodents have provided valuable insight into the brain mechanisms that mediate the construction of defensive modes with varying degrees of threat imminence. In this Review, we discuss accumulating evidence that the central nucleus of the amygdala (CeA) plays a key role in this process. Specifically, we propose that the mutually inhibitory circuits of the CeA use a winner-takes-all strategy that supports transitioning across defensive modes and the execution of specific defensive behaviors to previously formed threat associations. Our proposal provides a conceptual framework in which seemingly divergent observations regarding CeA function can be interpreted and identifies various areas of priority for future research. Moscarello and Penzo propose that mutually inhibitory circuits within the central nucleus of the amygdala implement a ‘winner-takes-all’ mechanism that guides transitions across defensive modes defined by threat-imminence theory.
 
Article
The ε4 allele of the apolipoprotein E (APOE) gene, a genetic risk factor for Alzheimer’s disease, is abundantly expressed in both the brain and periphery. Here, we present evidence that peripheral apoE isoforms, separated from those in the brain by the blood–brain barrier, differentially impact Alzheimer’s disease pathogenesis and cognition. To evaluate the function of peripheral apoE, we developed conditional mouse models expressing human APOE3 or APOE4 in the liver with no detectable apoE in the brain. Liver-expressed apoE4 compromised synaptic plasticity and cognition by impairing cerebrovascular functions. Plasma proteome profiling revealed apoE isoform-dependent functional pathways highlighting cell adhesion, lipoprotein metabolism and complement activation. ApoE3 plasma from young mice improved cognition and reduced vessel-associated gliosis when transfused into aged mice, whereas apoE4 compromised the beneficial effects of young plasma. A human induced pluripotent stem cell-derived endothelial cell model recapitulated the plasma apoE isoform-specific effect on endothelial integrity, further supporting a vascular-related mechanism. Upon breeding with amyloid model mice, liver-expressed apoE4 exacerbated brain amyloid pathology, whereas apoE3 reduced it. Our findings demonstrate pathogenic effects of peripheral apoE4, providing a strong rationale for targeting peripheral apoE to treat Alzheimer’s disease.
 
HFD consumption for 1 week induces compulsive sucrose-seeking behavior in adult mice
a, Schematic of the experimental approach. b,c, Bar graphs showing the number of lever presses (b) and the ratio of lever presses in test sessions to lever presses in baseline (BL) sessions (c) in chow (n = 10) and HFD (n = 11) mice. d, Quantification of the time spent exploring the tray (left) and lever (right) for chow (n = 10) and HFD (n = 11) mice in test sessions. e, Quantification of the ratio of reward approaches to lever presses for chow (n = 10) and HFD (n = 11) mice in BL and test sessions. f,g, Quantifications of the probability of (f) and time spent (g) freezing in chow (n = 8) and HFD (n = 9) mice in response to visual cues in test sessions. h, Quantification of percentage freezing in chow (n = 6) and HFD (n = 8) mice during a 10-second period before (pre-CS+) and after (CS+) presentation of a conditioned light stimulus in a cued fear-conditioning paradigm. i, Quantification of the foot-shock currents required to elicit flinching, jumping and vocalizing in chow (n = 8) and HFD (n = 8) mice. *P < 0.05, **P < 0.01 and ***P < 0.001. Data are mean ± s.e.m. See Supplementary Tables 1 and 2 for further details of statistical data analysis.
Source data
Excitation of aPVT excitatory neurons promotes compulsive sucrose-seeking behavior in mice
a, Representative images showing c-fos⁺ aPVT cells in chow and HFD mice. Bar, 100 μm. b, Bar graph showing the number of c-fos⁺ cells in the aPVT in chow (n = 15 sections/5 mice) and HFD (n = 19 sections/6 mice) mice after test sessions. c, Representative images showing c-fos⁺ neurons are co-expressed with CaMKIIɑ. Bars, 100 μm. d, Bar graph showing that most c-fos⁺ neurons are co-expressed with CaMKIIɑ (n = 10 sections/3 mice). e, Diagrams of the experimental setup (left), the viral vector strategy and optical fiber placement (middle) and representative image showing GCaMP6f-expressing neurons within the aPVT (right) (n = 25 mice). Bar, 500 μm. f, Average GCaMP6f responses before and after lever presses (dotted lines) from CaMKIIɑ⁺ neurons in the aPVT in chow mice and HFD mice. Red bars above the graph denote a significant difference in ΔF/F. Permutation tests, P = [0.003, 0.01). Shaded areas represent s.e.m. g, Heat map of ΔF/F for all individual trials. h, Quantification of the area under the curve (AUC) of GCaMP6f responses in a period of 6 seconds after lever presses in chow (n = 16) and HFD (n = 13) mice in baseline and test sessions. i, Schematic of the experimental approach (left) and representative image showing ChR2-expressing neurons in the aPVT and optical fiber placement (right) (n = 6 mice). Bar, 500 μm. j,k, Bar graphs showing the number of lever presses in control and ChR2 mice in baseline (j, n = 7 Ctrl and 9 ChR2 mice) and test (k, n = 7 Ctrl and 11 ChR2 mice) sessions. l, Schematic (left) of the viral vector strategy and representative image (right) (n = 7 mice) showing the hM4Di-expressing neurons in the aPVT. Bar, 100 μm. m,n, Bar graphs showing the number of lever presses (m) and the ratio of test lever presses to baseline lever presses (n) in control (n = 17) and hM4Di (n = 15) mice fed HFD. *P < 0.05 and ***P < 0.001. Data are mean ± s.e.m. See Supplementary Tables 1 and 2 for further details of statistical analysis. Ctrl, control.
Source data
HFD-induced microglia proliferation in the aPVT promotes compulsive sucrose-seeking in mice
a, Representative images of Iba-1⁺ cells in the aPVT in mice fed chow, chow containing PLX3397, HFD and HFD containing PLX3397. Bar, 100 μm. b, Quantification of the number of Iba-1⁺ cells in the aPVT from the four groups of mice shown in a (n = 24 sections/5 mice, n = 22 sections/6 mice, n = 23 sections/6 mice and n = 30 sections/6 mice for Chow, Chow + PLX3397, HFD and HFD + PLX3397 groups, respectively). c,d, Bar graphs showing the number of lever presses (c) and the ratio of lever presses in test sessions to lever presses in baseline (BL) sessions (d) in chow and HFD mice with or without PLX3397 treatment (n = 11, 12, 13 and 12 mice for Chow, Chow + PLX3397, HFD and HFD + PLX3397 groups, respectively). e, Schematic (left) of infusion of an anti-CSF-1 antibody in the aPVT and representative image (right) (n = 8 mice) showing the cannula placement and Iba-1⁺ cells (green). Bar, 500 μm. f,g, Representative images (f) and quantification (g) of Iba-1⁺ cells in the aPVT in HFD mice after infusion of ACSF (n = 9 sections/4 mice) or the anti-CSF-1 antibody (n = 13 sections/5 mice) in the aPVT. Bar, 200 μm. h,i, Bar graphs showing the number of lever presses (h) and the ratio of lever presses in test sessions to lever presses in BL sessions (i) in HFD mice infused with ACSF (n = 11) or the anti-CSF-1 antibody (n = 11). *P < 0.05, **P < 0.01 and ***P < 0.001. Data are mean ± s.e.m. See Supplementary Tables 1 and 2 for further details of statistical analysis.
Source data
Article
Overconsumption of palatable food may initiate neuroadaptive responses in brain reward circuitry that may contribute to eating disorders. Here we report that high-fat diet (HFD) consumption impedes threat-cue-induced suppression of sucrose-seeking in mice. This compulsive sucrose-seeking was due to enhanced cue-triggered neuronal activity in the anterior paraventricular thalamus (aPVT) resulting from HFD-induced microglia activation. Thus, metabolic inflammation in the aPVT produces an adaptive response to threat cues, leading to compulsive food-seeking.
 
Study summary
We performed single-nuclei RNA-seq on brain samples from 192 genotyped donors. We mapped cis-eQTLs for eight major brain cell types and identified a total of 7,607 cis-eQTL genes. We identified cell-type-specific genetic effects and leveraged our results to identify risk genes for brain disorders.
Cis-eQTL discoveries
a, Number of cis-eQTLs per cell type (5% FDR). b, Number of cis-eQTLs (5% FDR) versus number of single nuclei belonging to the cell types. c, Histogram of the replication P values for each SNP–gene pair of our cis-eQTL discoveries (aggregated for glia and neurons) in a large cortical eQTL study¹⁴. The proportion of replicating SNP–gene pairs at 5% FDR is displayed as well as the pi1 statistic and the proportion of replicating eQTL with the same direction of effect. d, Enrichment of the discovered cis-eQTLs around the TSS. e, LOEUF scores⁵¹ for (1) genes with no detected eQTL (at either the cell type level or the ‘tissue-like’ level); (2) genes with an eQTL detected at the cell type level but not at the ‘tissue-like’ level; (3) genes with an eQTL detected at the ‘tissue-like’ level but not at the cell type level; and (4) genes with eQTLs detected both at the cell type level and the ‘tissue-like’ level. All eQTLs detected in neuronal cell types or glial cell types were aggregated here (individual cell types are shown in Supplementary Fig. 9) (more constrained genes have lower LOEUF scores). The black horizontal bars indicate the medians. P values were obtained using a two-sided Wilcoxon test and are not adjusted for multiple testing. f, Examples of cis-eQTLs in astrocytes and microglia (n = 189/187 biologically independent individuals, respectively). The two-sided P value was obtained using fastQTL and is corrected for multiple testing (Methods). g, Examples of cis-eQTLs with fine-mapped SNPs (causal probability of 0.87 for AHI1 and 0.53 for GLUD1) (n = 178/187 biologically independent individuals, respectively). The two-sided P value was obtained using fastQTL and is corrected for multiple testing (Methods). Box plots show the median and first and third quartiles, and whiskers extend up to 1.5 times the interquartile range. LoF, loss of function.
Cell-type-specific genetic effects on gene expression
a, Enrichment of cell-type-specific ATAC-seq peaks (derived from snATAC-seq¹⁷) around our cis-eQTLs. b, Number of significant cell-type-specific genetic effects (5% FDR). ‘Cell type specific (at least one)’ shows the number of cis-eQTLs that have a different genetic effect for the same SNP–gene pair in at least one cell type. ‘Cell type specific (all)’ shows the number of cis-eQTLs that have a different genetic effect for the same SNP–gene pair in all cell types, whereas ‘Cell type specific (aggregate)’ shows the number of genes that have a different genetic effect in the discovered cell types than the aggregate genetic effect of all other cell types. c, Estimates of the proportions of cis-eQTLs that have a different genetic effect in another cell type. Estimates were computed on the interaction P value distributions using the pi1 statistic²⁵. d, Example of a cell-type-specific cis-eQTL for CHRM5 in oligodendrocytes (n = 192 biologically independent individuals); e, RNF150 in microglia (n = 187 biologically independent individuals); and f, FBN2 in microglia (n = 187 biologically independent individuals). Each dot represents an individual. The displayed two-sided P values (Wald z-statistics) are the interaction P values testing whether the genetic effect in the discovered cell type (top-left cell type) is different from the genetic effects in all other cell types adjusted for multiple testing (Methods). Box plots show the median and first and third quartiles, and whiskers extend up to 1.5 times the interquartile range.
Co-localization results
a, Number of co-localized genes per loci. At most loci, a single gene co-localizes in a single cell type. b, PP of co-localization for co-localized genes (PP > 0.7) in the top cell type (rank = 1) and other cell types (ranked). The two-sided P values were obtained using a Wilcoxon sum-rank test (n = 23/19/65/116 for AD/PD/MS/SCZ, respectively). Each dot represents a co-localized gene. BIN1, TMEM163, NR1H3 and RERE are highlighted as examples. PPs of shared genetic signal between GWAS and eQTLs for AD (c), PD (d), MS (e) and SCZ (f). The closest gene to the top GWAS signal is indicated on the left; the co-localized gene is indicated on the right. The beta column indicates the effect size of the top GWAS SNP at the locus. The risk column indicates whether an increase in gene expression leads to an increase in disease risk (red) or a decrease in disease risk (blue). The LOEUF⁵¹ column indicates whether the gene is constrained (low score) or not (high score). Genes that co-localized with the same trait in the Metabrain study¹⁴ are highlighted with an asterisk. Co-localized genes with a PP > 0.7 are shown for AD and PD. Co-localized genes with a PP > 0.8 are shown for MS and SCZ (for better visualization). All results are available in Supplementary Table 5 and in the shinyApp. Box plots show the median and first and third quartiles, and whiskers extend up to 1.5 times the interquartile range.
Epigenomic overlap of GWAS SNPs around co-localized genes
a, Genomic map indicating the location of AD GWAS SNPs (r² > 0.8 with index SNP) overlapping a microglia-specific enhancer, connected to the PICALM promoter through PLAC-seq. b, Genomic map indicating the location of AD GWAS SNPs (r² > 0.8 with index SNP) overlapping a microglia-specific enhancer, connected to the INPP5D promoter through PLAC-seq. c, Genomic map indicating the location of an AD GWAS SNP (r² > 0.8 with index SNP) overlapping the promoter of APH1B in oligodendrocytes. d, Genomic map indicating the location of an AD GWAS SNP (r² > 0.8 with index SNP) overlapping an oligodendrocyte enhancer located within the CR1 gene. e, Genomic map indicating the location of PD GWAS SNPs (r² > 0.8 with index SNP) overlapping a microglia enhancer located within GPNMB. f, Genomic map indicating the location of MS GWAS SNPs (r² > 0.8 with index SNP) overlapping an astrocytic enhancer located downstream of NR1H3. g, LocusZoom plots showing AD GWAS associations and eQTL associations (for the PICALM gene in microglia) at the PICALM locus. h, LocusZoom plots showing AD GWAS associations and eQTL associations (for the INPP5D gene in microglia) at the INPP5D locus. cM, centimorgan.
Article
To date, most expression quantitative trait loci (eQTL) studies, which investigate how genetic variants contribute to gene expression, have been performed in heterogeneous brain tissues rather than specific cell types. In this study, we performed an eQTL analysis using single-nuclei RNA sequencing from 192 individuals in eight brain cell types derived from the prefrontal cortex, temporal cortex and white matter. We identified 7,607 eGenes, a substantial fraction (46%, 3,537/7,607) of which show cell-type-specific effects, with strongest effects in microglia. Cell-type-level eQTLs affected more constrained genes and had larger effect sizes than tissue-level eQTLs. Integration of brain cell type eQTLs with genome-wide association studies (GWAS) revealed novel relationships between expression and disease risk for neuropsychiatric and neurodegenerative diseases. For most GWAS loci, a single gene co-localized in a single cell type, providing new clues into disease etiology. Our findings demonstrate substantial contrast in genetic regulation of gene expression among brain cell types and reveal potential mechanisms by which disease risk genes influence brain disorders. Bryois et al. mapped genetic variants regulating gene expression in eight major brain cell types. They found a large number of cell-type-specific genetic effects and leveraged their results to identify novel putative risk genes for brain disorders.
 
Profiling of genetically defined cortical pyramidal neuron classes
a, Schematic of experimental design. b, Representative coronal sections (n ≥ 3 biological replicates) showing correct laminar location of tdTomato⁺ cells in the somatosensory cortex and FACS plots of Tle4-CThPNs and Cux2-CPNs at P7. Scale bar, 100 μm. Also see Extended Data Fig. 1. c–e, Developmental dynamics of differentially expressed genes (c), differentially enriched ATAC-seq peaks (d) and DMRs (e). f, Fraction of dynamic features classified into each overall category. g, 2D MDS plots for each modality.
Divergent regulatory principles of early and late stages of neuronal development
a, Schematic of analyses. b, Summary of gene expression clusters from Fig. 1c. c, TF expression specificity within the brain for TFs in different categories of cluster, from expression data from the Allen Brain Atlas (left). Higher values indicate more specific expression. TF expression specificity across 397 mouse cell types from the FANTOM5 project (right). Box and midline: 25th, 50th and 75th percentiles; whiskers: 1.5× interquartile range from box. d, Summary of ATAC peak clusters from Fig. 1d. e, Fraction of open chromatin regions (ATAC-seq peaks) in different categories of clusters that overlap CGIs or annotated promoter regions. f, Fraction of DMRs in ATAC clusters of different categories that are static or dynamic over the time course. g, Fraction of ATAC-seq peaks from different classes of ATAC clusters that overlap DNAse hypersensitivity sites (DHSs) across 35 cell types from the mouse ENCODE project. Adult (8-week) whole-brain and telencephalon samples highlighted in red; embryonic (E14.5 and E18.5) brain samples are highlighted in blue. Also see Extended Data Fig. 5f. Error bars: mean ± s.e.m. h, Open chromatin specificity across ATAC peak clusters (nEarly = 10,488, nLate = 17,570), as the concordance of ATAC peaks with an scATAC-seq panel of 85 mouse primary tissues and cell types¹⁷. Box plot as in c. i, Distribution of phyloP sequence conservation across all placental mammals for early developmental and late developmental ATAC clusters. Box plot as in c. j, Density of known TF binding motifs within ATAC clusters of different types. Box plot as in c. k, Average sequence entropy within ATAC clusters of different types. Box plot as in c. l, Summary of characteristics from c and g–j for each of the shared developmental ATAC peak clusters individually. Each arm of the plot represents an individual metric (center = low, edge = high). Also see Extended Data Fig. 5b.
ScRNA-seq demonstrates a developmental shift in specificity of shared gene expression programs across multiple neuronal subclasses in both mouse and marmoset
a, UMAP representation of gene expression profiles from 60,989 single cells from mouse cortex at P1, P7 and P21, color-coded by major cell type. Left: UMAP plots showing cell distribution by age. b, UMAP representation of 36,592 single nuclei from marmoset cortex at P0 and Y2, color-coded by major cell type. Left: UMAP plots showing cell distribution by age. c, Representative marker genes for major cell types in the mouse data. Also see Supplementary Fig. 6a,c. d, Representative marker genes for major cell types in the marmoset data. Also see Supplementary Fig. 8a,c. e, Developmental dynamics of clusters of differentially expressed genes across the mouse excitatory and inhibitory neuronal populations (cell type indicated by color-coded bar at top, corresponding to colors in a). f, Developmental dynamics of clusters of differentially expressed genes across the marmoset excitatory and inhibitory neuronal populations (cell type indicated by color-coded bar at top, corresponding to colors in b). g, Mouse gene expression specificity in the shared-early and shared-late gene clusters, within the mouse brain (from expression data from the Allen Brain Atlas; left), and across 397 mouse cell types (from the FANTOM5 project; right). Higher values indicate more specific expression. Box and midline: 25th, 50th and 75th percentiles; whiskers: 1.5× interquartile range from box. h, Marmoset gene expression specificity in the shared-early and shared-late gene clusters, within the mouse brain (from expression data from the Allen Brain Atlas; left), across 397 mouse cell types (from the FANTOM5 project; center) and across cell populations from our mouse single-cell dataset (right). Higher values indicate more specific expression. Box plot as in g.
Developmental shift in gene regulatory principles is common to multiple neuronal subclasses and is conserved across mouse and marmoset
a, UMAP representation of ATAC chromatin accessibility profiles from 19,145 single cells from mouse cortex at P1, P7 and P21, color-coded by major cell type. Left: UMAP plots with cell distribution by age. b, UMAP representation of ATAC chromatin accessibility profiles from 15,919 single cells from marmoset cortex at P0 and Y2, color-coded by major cell type. Left: UMAP plots with cell distribution by age. c, Developmental dynamics of clusters of differentially accessible ATAC peaks across the mouse excitatory and inhibitory neuronal populations (cell type indicated by color-coded bar at top, corresponding to colors in a). Color scale: average peak normalized read count. d, Summary of CGI overlap, promoter overlap, TF motif density, tissue specificity and sequence conservation for each of the mouse shared developmentally regulated clusters, as in Fig. 2k. Also see Extended Data Fig. 9a–c. e, Developmental dynamics of clusters of differentially accessible ATAC peaks across the marmoset excitatory and inhibitory neuronal populations (cell type indicated by color-coded bar at top, corresponding to colors in b). Color scale: average peak normalized read count. f, Summary of CGI overlap, promoter overlap, TF motif density, tissue specificity and sequence conservation for each of the marmoset shared developmentally regulated clusters. Also see Extended Data Fig. 9d–f. g, Overlap between genes in the mouse and marmoset single-cell gene expression clusters, showing the four most similar pairs. Also see Extended Data Fig. 10b. h, Overlap between accessible regions in the mouse and marmoset single-cell ATAC chromatin accessibility clusters, showing the five most similar pairs. Also see Extended Data Fig. 10d. Color scale: average peak normalized read count.
Article
Mammalian neocortical neurons span one of the most diverse cell type spectra of any tissue. Cortical neurons are born during embryonic development, and their maturation extends into postnatal life. The regulatory strategies underlying progressive neuronal development and maturation remain unclear. Here we present an integrated single-cell epigenomic and transcriptional analysis of individual mouse and marmoset cortical neuron classes, spanning both early postmitotic stages of identity acquisition and later stages of neuronal plasticity and circuit integration. We found that, in both species, the regulatory strategies controlling early and late stages of pan-neuronal development diverge. Early postmitotic neurons use more widely shared and evolutionarily conserved molecular regulatory programs. In contrast, programs active during later neuronal maturation are more brain- and neuron-specific and more evolutionarily divergent. Our work uncovers a temporal shift in regulatory choices during neuronal diversification and maturation in both mice and marmosets, which likely reflects unique evolutionary constraints on distinct events of neuronal development in the neocortex. The mechanisms underlying neuron specification and maturation are unclear. Here the authors provide an integrated epigenomic and transcriptomic analysis of mouse and marmoset neocortical neuronal classes. Pan-neuronal programs active during early development are more evolutionary conserved but not neuron-specific, whereas pan-neuronal programs active during later stages of maturation are more neuron- and species-specific.
 
Article
Resting-state functional magnetic resonance imaging (MRI) has yielded seemingly disparate insights into large-scale organization of the human brain. The brain’s large-scale organization can be divided into two broad categories: zero-lag representations of functional connectivity structure and time-lag representations of traveling wave or propagation structure. In this study, we sought to unify observed phenomena across these two categories in the form of three low-frequency spatiotemporal patterns composed of a mixture of standing and traveling wave dynamics. We showed that a range of empirical phenomena, including functional connectivity gradients, the task-positive/task-negative anti-correlation pattern, the global signal, time-lag propagation patterns, the quasiperiodic pattern and the functional connectome network structure, are manifestations of these three spatiotemporal patterns. These patterns account for much of the global spatial structure that underlies functional connectivity analyses and unifies phenomena in resting-state functional MRI previously thought distinct.
 
Article
Studies investigating the neural mechanisms by which associations between cues and predicted outcomes control behavior often use associative learning frameworks to understand the neural control of behavior. These frameworks do not always account for the full range of effects that novelty can have on behavior and future associative learning. Here, in mice, we show that dopamine in the nucleus accumbens core is evoked by novel, neutral stimuli, and the trajectory of this response over time tracked habituation to these stimuli. Habituation to novel cues before associative learning reduced future associative learning, a phenomenon known as latent inhibition. Crucially, trial-by-trial dopamine response patterns tracked this phenomenon. Optogenetic manipulation of dopamine responses to the cue during the habituation period bidirectionally influenced future associative learning. Thus, dopamine signaling in the nucleus accumbens core has a causal role in novelty-based learning in a way that cannot be predicted based on purely associative factors.
 
Article
Frontotemporal dementia (FTD) is the second most prevalent form of early-onset dementia, affecting predominantly frontal and temporal cerebral lobes. Heterozygous mutations in the progranulin gene (GRN) cause autosomal-dominant FTD (FTD-GRN), associated with TDP-43 inclusions, neuronal loss, axonal degeneration and gliosis, but FTD-GRN pathogenesis is largely unresolved. Here we report single-nucleus RNA sequencing of microglia, astrocytes and the neurovasculature from frontal, temporal and occipital cortical tissue from control and FTD-GRN brains. We show that fibroblast and mesenchymal cell numbers were enriched in FTD-GRN, and we identified disease-associated subtypes of astrocytes and endothelial cells. Expression of gene modules associated with blood–brain barrier (BBB) dysfunction was significantly enriched in FTD-GRN endothelial cells. The vasculature supportive function and capillary coverage by pericytes was reduced in FTD-GRN tissue, with increased and hypertrophic vascularization and an enrichment of perivascular T cells. Our results indicate a perturbed BBB and suggest that the neurovascular unit is severely affected in FTD-GRN. Frontotemporal dementia (FTD) is a common early-onset dementia caused by heterozygous mutations in the progranulin gene (GRN). Gerrits et al. demonstrate blood–brain barrier dysfunction and a severely affected neurovasculature in FTD-GRN.
 
The language network in native speakers of diverse languages
Activation maps for the Alice language localizer contrast (Native-language > Degraded-language) in the LH of a sample participant for each language (see Extended Data Fig. 2 for RH maps and details of the image generation procedure). The general topography of the language network in speakers of 45 languages is similar, and the variability observed is similar to the variability that has been reported for the speakers of the same language¹⁰ (Extended Data Fig. 8). A significance map was generated for each participant by FreeSurfer; each map was smoothed using a Gaussian kernel of 4-mm FWHM and thresholded at the 70th percentile of the positive contrast for each participant. The surface overlays were rendered on the 80% inflated white–gray matter boundary of the fsaverage template using FreeView/FreeSurfer. Opaque red and yellow correspond to the 80th and 99th percentiles of positive-contrast activation for each participant, respectively. (These maps were used solely for visualization; all the analyses were performed on the data analyzed in the volume (see Extended Data Fig. 3)).
The probabilistic overlap map for the Native-language > Degraded-language contrast. This map was created by binarizing and overlaying the 86 participants’ individual maps (like those shown in Fig. 1). The value in each vertex corresponds to the proportion of participants for whom that vertex belongs to the language network (see Extended Data Fig. 8 for a comparison between this probabilistic atlas versus atlases based on native speakers of the same language).
a, Percent BOLD signal change across the LH language fROIs (see inset for the RH language fROIs) for the three language conditions of the Alice localizer task (native language, acoustically degraded native language and unfamiliar language), the spatial WM task and the math task. The language fROIs show robust functional selectivity for language processing. Here and in the other panels, the dots correspond to languages (n = 45 in all panels), and the labels mark the averages for each language family (n = 12; AfAs, Afro-Asiatic; AuAs, Austro-Asiatic; Aust, Austronesian; Drav, Dravidian; IndEu, Indo-European; Japn, Japonic; Korn, Koreanic; AtCo, Atlantic-Congo; SinT, Sino-Tibetan; Turk, Turkic; Ural, Uralic; Isol, Isolate). Here and in other panels, box plots include the first quartile (lower hinge), third quartile (upper hinge) and median (central line); upper and lower whiskers extend from the hinges to the largest value no further than 1.5 times the interquartile range; darker-colored dots correspond to outlier data points. For each statistical comparison reported in the text, a two-tailed t-test was used (see Supplementary Table 2 for results of linear mixed-effects models); no correction for the number of comparisons was applied (because each test addressed a distinct question). b, Three measures that reflect LH lateralization of the language network: i-strength of activation (effect sizes for the Native-language > Degraded-language contrast); ii-extent of activation (number of voxels within the union of the language parcels at a fixed threshold for the Native-language > Degraded-language contrast; a whole-brain version of this analysis yielded a similar result: t(44) = 5.79, P < 0.001); and iii-inter-region functional correlations during two naturalistic cognition paradigms (i-story comprehension in the participant’s native language; ii-resting state). The LH language network shows greater selectivity for language processing relative to a control condition, is more spatially extensive and is more strongly functionally integrated than the RH language network. c, Inter-region functional correlations for the LH and RH language network and the MD network during two naturalistic cognition paradigms (i-story comprehension in the participant’s native language and ii-resting state). The language and the MD networks are each strongly functionally integrated but are robustly dissociated from each other (pairs of fROIs straddling network boundaries show little/no correlated activity).
Article
To understand the architecture of human language, it is critical to examine diverse languages; however, most cognitive neuroscience research has focused on only a handful of primarily Indo-European languages. Here we report an investigation of the fronto-temporo-parietal language network across 45 languages and establish the robustness to cross-linguistic variation of its topography and key functional properties, including left-lateralization, strong functional integration among its brain regions and functional selectivity for language processing. fMRI reveals similar topography, selectivity and inter-connectedness of language brain areas across 45 languages. These properties may allow the language system to handle the shared features of languages, shaped by biological and cultural evolution.
 
Article
A large body of evidence has indicated that the phasic responses of midbrain dopamine neurons show a remarkable similarity to a type of teaching signal (temporal difference (TD) error) used in machine learning. However, previous studies failed to observe a key prediction of this algorithm: that when an agent associates a cue and a reward that are separated in time, the timing of dopamine signals should gradually move backward in time from the time of the reward to the time of the cue over multiple trials. Here we demonstrate that such a gradual shift occurs both at the level of dopaminergic cellular activity and dopamine release in the ventral striatum in mice. Our results establish a long-sought link between dopaminergic activity and the TD learning algorithm, providing fundamental insights into how the brain associates cues and rewards that are separated in time. The authors found that dopamine signals move gradually from time of reward to time of cue through intermediate timepoints, similarly to the evaluation signals used in temporal difference learning. These findings thereby fill a gap between computational theories and the brain.
 
Article
Sleep has a complex micro-architecture, encompassing micro-arousals, sleep spindles and transitions between sleep stages. Fragmented sleep impairs memory consolidation, whereas spindle-rich and delta-rich non-rapid eye movement (NREM) sleep and rapid eye movement (REM) sleep promote it. However, the relationship between micro-arousals and memory-promoting aspects of sleep remains unclear. In this study, we used fiber photometry in mice to examine how release of the arousal mediator norepinephrine (NE) shapes sleep micro-architecture. Here we show that micro-arousals are generated in a periodic pattern during NREM sleep, riding on the peak of locus-coeruleus-generated infraslow oscillations of extracellular NE, whereas descending phases of NE oscillations drive spindles. The amplitude of NE oscillations is crucial for shaping sleep micro-architecture related to memory performance: prolonged descent of NE promotes spindle-enriched intermediate state and REM sleep but also associates with awakenings, whereas shorter NE descents uphold NREM sleep and micro-arousals. Thus, the NE oscillatory amplitude may be a target for improving sleep in sleep disorders. Kjaerby and Andersen et al. show that norepinephrine (NE) plays profound roles in shaping sleep micro-architecture. NE slowly oscillates during sleep, with NE oscillatory amplitude being a major determinant of spindle-dependent memory consolidation and awakenings.
 
Article
A new study shows that infra-slow cortical norepinephrine oscillations shape the micro-structure of sleep and transitions to micro-arousals, wakefulness or rapid eye movement (REM) sleep. Prolonged descending phases of these oscillations promote the occurrence of spindle-rich intermediate sleep, which is involved in memory consolidation.
 
Article
Studying the natural wanderings of the living brain is extremely challenging. Bolt et al. describe a new framework for considering the brain’s intrinsic activity based on the geophysical concepts of standing and traveling waves.
 
Article
Optogenetic actuators enable highly precise spatiotemporal interrogation of biological processes at levels ranging from the subcellular to cells, circuits and behaving organisms. Although their application in neuroscience has traditionally focused on the control of spiking activity at the somatodendritic level, the scope of optogenetic modulators for direct manipulation of presynaptic functions is growing. Presynaptically localized opsins combined with light stimulation at the terminals allow light-mediated neurotransmitter release, presynaptic inhibition, induction of synaptic plasticity and specific manipulation of individual components of the presynaptic machinery. Here, we describe presynaptic applications of optogenetic tools in the context of the unique cell biology of axonal terminals, discuss their potential shortcomings and outline future directions for this rapidly developing research area. This Review provides a comprehensive overview of presynaptic applications of optogenetic tools, including the associated challenges, current limitations and future directions for this approach.
 
Experimental paradigm
a, Left, depth electrodes (6–12 per patient) implanted in patients with epilepsy for clinical monitoring, each consisting of eight 1.5-mm iEEG contacts along the shaft and eight 40-μm microwires protruding from the distal tip, recording LFP and spiking activities. Right, two representative pre-implant magnetic resonance images co-registered with post-implant computed tomography used to localize electrodes from the same individual. b, A topographic display (flat cortical map) of all sites where neuronal activity was recorded (each circle denotes one iEEG macroelectrode or a bundle of microwires) along with the probability of observing an auditory response in wakefulness (number of responses/number of stimuli played, color bar on right). LH, Left Hemisphere; RH, Right Hemisphere. c, Representative time–frequency representation (spectrogram) of iEEG recorded in one individual during a full-night sleep study with intermittent auditory stimulation. Warm colors (for example, red, see color bar on far right) mark increased power in specific time–frequency windows (frequency shown on left side of y axis). Superimposed hypnogram (black trace) marks the time-course of sleep/wake states (shown on right side of y axis). Note that NREM stages N2 and N3 are associated with increased power in spindle (10–15 Hz) and slow (<4 Hz) frequency ranges.
Robust auditory spiking responses across the temporal lobe during NREM sleep
a, Left, representative spiking response of neuronal unit in response to word in the primary auditory cortex. The top row shows the action potential waveform (left inset, mean ± s.d.) and the anatomical location of the recorded unit (right inset, circle in MRI sections), while the grayscale soundwave spectrograms are shown above the raster (lighter shades denote stronger power). Pink, wakefulness; green, NREM sleep. Vertical dotted black lines mark stimulus onset and offset. Horizontal bars above peri-stimulus time histogram (PSTH) time-courses indicate automatically detected response intervals for which response magnitude was compared quantitatively. Right, same format for a unit in higher-order auditory cortex (planum polare) responding to music. b, Scatter plot of auditory spiking response magnitudes during NREM sleep (y axis) versus wakefulness (x axis), together with a histogram of gain values comparing response magnitudes (upper-right corner along the diagonal). N = 312 responses/55 clusters/7 patients. Each data point represents the averaged response across stimuli and trials per cluster. Mean and P value were calculated using a nested mixed model analysis (Methods) (confidence interval (CI) (−43.381, −12.064), P = 0.018). c, Gain values of spiking response magnitudes (NREM versus wakefulness) in each region exhibiting auditory responses. The position of each circle denotes its anatomical location shown on a standard (Montreal Neurological Institute (MNI)) brain template, the circle’s color represents the average gain detected in that region (color bar on bottom right), and the circle’s size reflects the number of responses detected in that region. The letters A and B mark the locations of the representative units shown in panels a and b.
Source data
High-gamma auditory responses and entrainment to 40-Hz click-trains during NREM sleep
a, Representative spectrogram of induced LFP high-frequency power in response to music during wakefulness (left) and NREM sleep (right). Color bar on right. Black rectangles represent time–frequency regions-of-interest used for subsequent quantification. Top, grayscale soundwave spectrograms (lighter shades denote stronger power). b, Time-courses of high-gamma (80–200 Hz) responses shown in a. Horizontal bars and vertical black lines as in Fig. 2a. c, Scatter plot of high-gamma responses (% increase from baseline) during NREM sleep (y axis) versus wakefulness (x axis). Gain histogram in upper-right inset as in Fig. 2b; black and gray lines represent distributions for LFP and iEEG data, respectively. Each data point represents the averaged response across stimuli and trials per electrode (n = 556 responses/74 LFP microwires, black dots (CI (−23.732, 8.426), P = 0.276); 320 responses/55 iEEG channels, white circles (CI (−0.461, 55.422), P = 0.205); six patients). d, High-gamma gain values (NREM versus wakefulness) in each region exhibiting auditory high-gamma responses. Circle positions, color and size as in Fig. 2c. The letter A marks location of the representative microwire shown in panel a and b. e, Representative time-courses of LFP high-gamma responses showing a tight relationship with the sound envelope of auditory stimulus. f, Robust correlation between LFP high-gamma responses and the sound envelope in both wakefulness and NREM sleep. LFP: r(NREM-wake) = −0.002, CI (−0.02, 0.03), P = 0.88 for n = 406 responses/64 microwires/6 patients; iEEG: r(NREM-wake) = 0.04, CI (−0.07, −0.01), P = 0.006 for n = 210 responses/40 macroelectrodes/6 patients. g, Scatter plot of the degree of response attenuation in NREM sleep (y axis) versus latency of gamma LFP response (x axis) in each microwire (n = 25); Pearson correlation coefficient: r = 0.73, P < 0.001 by permutation test. Cyan dots mark adjacent microwires that exhibit different sleep attenuations and latencies. h, iEEG ITPC in response to a 40-Hz click-train in wakefulness (top), and associated Event-Related Potential (ERP) (bottom). i, Scatter plot of ITPC in response to 40-Hz click-trains during NREM sleep (y axis) versus wakefulness (x axis). Inset and format as in panel c. Each data point in scatter represents the averaged response across stimuli and trials per electrode. n = 84 LFP microwires/12 patients (black dots, CI (−40.9, 8.7), P = 0.176) and n = 325 iEEG macroelectrodes/13 patients (white circles, CI (−49.9, −2.0), P = 0.036). Mean and P values were calculated using a nested mixed model analysis for panels c and f and a one-level mixed model for panel i (Methods).
Source data
NREM sleep disrupts auditory-induced LFP ABD
a, Representative spectrogram of auditory-induced LFP power (<50 Hz) in response to music during wakefulness (left) and NREM sleep (right). Colder colors (for example, blue) denote a decrease in power (dB scale, color bar on right). Black rectangles represent time–frequency regions-of-interest used for subsequent quantification. b, Time-course of induced alpha–beta (10–30 Hz) power dynamics shown in a. Pink, wakefulness; green, NREM sleep. Horizontal pink bars above the time-course indicate automatically detected response intervals (Methods) for which the response magnitude was compared quantitatively (significant decreases were not detected in sleep). Vertical black lines mark stimulus onset and offset. c, Scatter plot of all auditory-induced ABD responses (% power decrease below baseline) during NREM sleep (y axis) versus wakefulness (x axis), together with a histogram of gain values comparing response magnitude (upper-right corner along the unity diagonal; black and gray lines in top-right inset represent distributions for LFP and iEEG data, respectively). Each data point in scatter represents the averaged response across stimuli and trials per electrode. n = 244 responses/57 LPF microwires/7 patients (black dots, CI (−84.434, −2.258), P = 0.042) and n = 188 responses/29 iEEG electrodes/5 patients (white dots, CI (−92.899, −70.678), P < 0.001). Mean and P values were calculated using a nested mixed model analysis. d, ABD gain values (NREM versus wakefulness) in each region exhibiting such responses. The position of each circle represents its anatomical location shown on a standard (MNI) brain template, the circle’s color reflects the average gain detected in that region (color bar on right) and the circle’s size reflects the number of responses detected in the region. The letter A marks the location of the representative microwire shown in panel a. e, Scatter plot of ABD gain values (y axis) versus latency of ABD (x axis) in each microwire (n = 18). Pearson correlation coefficient r = 0.54, P < 0.001 by permutation test. Cyan dots mark adjacent microwires that exhibit different sleep attenuations and latencies.
Source data
Auditory responses in REM sleep
a, Two representative raster plots (top) and PSTHs (bottom) of spiking response of neuronal units to auditory stimuli (left, click-train; right, word) in the primary auditory cortex. Pink, wakefulness; green, REM sleep. Vertical dotted black lines mark stimulus onset and offset. Horizontal bars above the PSTH time-courses indicate automatically detected response intervals (Methods) for which the magnitude of the response was compared quantitatively. b, Scatter plot of auditory spiking response magnitudes during REM sleep (y axis) versus wakefulness (x axis), together with a histogram of gain values comparing response magnitudes (upper-right corner along the diagonal). n = 141 responses/25 clusters/2 patients (CI (−31.763, −2.739), P = 0.022). c, Scatter plot of high-gamma responses to auditory stimuli during REM sleep (y axis) versus wakefulness (x axis), with a histogram of gain values comparing response magnitude (upper-right corner along the unity diagonal; black and gray lines in top-right inset represent gain distributions for LFP and iEEG data, respectively). Each data point represents the averaged response across stimuli and trials per electrode. n = 286 responses/33 LFP channels/2 patients (CI (−34.726, −12.838), P < 0.001) and n = 197 responses/30 iEEG channels/3 patients (CI (8.328, 22.630), P < 0.001)). d, Scatter plot of ABD responses to auditory stimuli in REM sleep (y axis) versus wakefulness (x axis). Histograms in top-right inset represent gain distributions above. n = 154 responses/32 LFP channels/3 patients (CI (−75.132, −52.207), P < 0.001) and n = 217 responses/36 iEEG channels/4 patients (CI (−78.814, −55.867), P < 0.001). e, Scatter plot of ITPC in response to 40-Hz click-trains during REM sleep (y axis) versus wakefulness (x axis), with a histogram of gain values as above. n = 60 LFP microwires/8 patients and n = 326 iEEG electrodes/9 patients. Each data point represents the averaged response across trials per electrode. Mean and P values were calculated using a one-level mixed model analysis (Methods); ***P < 0.001.
Source data
Article
During sleep, sensory stimuli rarely trigger a behavioral response or conscious perception. However, it remains unclear whether sleep inhibits specific aspects of sensory processing, such as feedforward or feedback signaling. Here, we presented auditory stimuli (for example, click-trains, words, music) during wakefulness and sleep in patients with epilepsy, while recording neuronal spiking, microwire local field potentials, intracranial electroencephalogram and polysomnography. Auditory stimuli induced robust and selective spiking and high-gamma (80–200 Hz) power responses across the lateral temporal lobe during both non-rapid eye movement (NREM) and rapid eye movement (REM) sleep. Sleep only moderately attenuated response magnitudes, mainly affecting late responses beyond early auditory cortex and entrainment to rapid click-trains in NREM sleep. By contrast, auditory-induced alpha–beta (10–30 Hz) desynchronization (that is, decreased power), prevalent in wakefulness, was strongly reduced in sleep. Thus, extensive auditory responses persist during sleep whereas alpha–beta power decrease, likely reflecting neural feedback processes, is deficient. More broadly, our findings suggest that feedback signaling is key to conscious sensory processing. Intracortical recordings in humans reveal that auditory stimulation during sleep induces robust spiking and high-gamma responses, whereas alpha–beta desynchronization—likely reflecting neural feedback processes—is reduced compared to wakefulness.
 
Article
During cortical development, the generation of neurons from progenitors coincides with increasing vascularization and oxygen levels in the fetal brain and the transition from anaerobic to aerobic metabolism. Dong et al. identify lactate, a metabolite of anaerobic glycolysis, as a multifunctional regulator that coordinates synchrony of cortical neurogenesis and angiogenesis.
 
Article
The meninges, comprising the leptomeninges (pia and arachnoid layers) and the pachymeninx (dura layer), participate in central nervous system (CNS) autoimmunity, but their relative contributions remain unclear. Here we report on findings in animal models of CNS autoimmunity and in patients with multiple sclerosis, where, in acute and chronic disease, the leptomeninges were highly inflamed and showed structural changes, while the dura mater was only marginally affected. Although dural vessels were leakier than leptomeningeal vessels, effector T cells adhered more weakly to the dural endothelium. Furthermore, local antigen-presenting cells presented myelin and neuronal autoantigens less efficiently, and the activation of autoreactive T cells was lower in dural than leptomeningeal layers, preventing local inflammatory processes. Direct antigen application was required to evoke a local inflammatory response in the dura. Together, our data demonstrate an uneven involvement of the meningeal layers in CNS autoimmunity, in which effector T cell trafficking and activation are functionally confined to the leptomeninges, while the dura remains largely excluded from CNS autoimmune processes. This work shows a distinct involvement of the meningeal layers in CNS autoimmunity. In animal models and in patients with multiple sclerosis, the leptomeninges were highly inflamed and showed structural changes, while the dura was only marginally affected.
 
Article
Regaining arm control is a top priority for people with paralysis. Unfortunately, the complexity of the neural mechanisms underlying arm control has limited the effectiveness of neurotechnology approaches. Here, we exploited the neural function of surviving spinal circuits to restore voluntary arm and hand control in three monkeys with spinal cord injury, using spinal cord stimulation. Our neural interface leverages the functional organization of the dorsal roots to convey artificial excitation via electrical stimulation to relevant spinal segments at appropriate movement phases. Stimulation bursts targeting specific spinal segments produced sustained arm movements, enabling monkeys with arm paralysis to perform an unconstrained reach-and-grasp task. Stimulation specifically improved strength, task performances and movement quality. Electrophysiology suggested that residual descending inputs were necessary to produce coordinated movements. The efficacy and reliability of our approach hold realistic promises of clinical translation. This paper describes a neurotechnology that interacts with neural circuits in the spinal cord to restore arm and hand control after injury. With this implant, monkeys with paralysis recovered the ability to reach and grasp just a few days after injury.
 
Oligodendrocytes display major transcriptomic alterations in the 5xFAD mouse model
a, Single-cell expression heat map of non-immune (CD45⁻) cells from 5xFAD (n = 9) and WT (n = 9) mice. b, Two-dimensional uniform manifold approximation and projection (UMAP) embedding of the scRNA-seq data. Dots are colored based on the scRNA-seq clustering. c, Number of DEGs between 5xFAD and WT mice across the major cell types. d, Specificity of the DEGs in oligodendrocytes. e, Power analysis—number of DEGs based on the number of cells used for the differential expression analysis. Red, downsampling of oligodendrocytes; orange, astrocytes; green, endothelial cells. Red dots correspond to the mean number of DEGs, and the bar represents the standard deviation. RBC, red blood cell. OEC, olfactory ensheathing cell.
Identification of a disease-associated oligodendrocyte state
a, Experimental strategy to study oligodendrocytes. Illustration was created with BioRender. b, Expression heat map of oligodendrocytes from 5xFAD and WT mice at various ages, GalC⁺ sorted and pooled together with the previously sequenced oligodendrocytes. n = 33 independent mice (across ages—6–8 months: n = 4 5xFAD, 4 WT; 10–11 months: n = 6 5xFAD, 5 WT; 15 months: n = 3 5xFAD, 3 WT; 24 months: n = 4 5xFAD, 4 WT). c, Two-dimensional UMAP embedding of the scRNA-seq oligodendrocyte data. Dots are colored based on the scRNA-seq clustering. Top, cells from WT mice; bottom, cells from 5xFAD mice. d, Z-score of the enrichment test comparing cluster frequency between 5xFAD and WT mice. P values were computed by fitting a Poisson regression and corrected using multiple testing (Methods). e, Proportion of DOLs among oligodendrocytes between 5xFAD and WT mice (left panel) and across ages (right panel). P values were computed using the Kruskal–Wallis test. Large bars correspond to the median and small bars to the interquartile range (IQR). n = 33 independent mice (across ages—6–8 months: n = 4 5xFAD, 4 WT; 10–11 months: n = 6 5xFAD, 5 WT; 15 months; n = 3 5xFAD, 3 WT; 24 months: n = 4 5xFAD, 4 WT). f, Volcano plot corresponding to the differential expression analysis between DOLs and the rest of the oligodendrocytes. g, Results of the promoter analysis by iRegulon. FDR, false discovery rate; TPM, transcripts per million..
DOLs are independent of the dementia’s etiology
a, Experimental strategy in the scRNA-seq dataset from Lee et al.³¹. b, Two-dimensional UMAP embedding of the oligodendrocytes in each mouse model. Cluster 11 cells are in orange. c, Comparison of the mean transcriptional profile of oligodendrocytes from cluster 11 with the rest of the oligodendrocytes. DOL genes are colored in orange. d, Result of GSEA analysis performed on the log2FC list between oligodendrocytes from cluster 11 and the rest of the oligodendrocytes using the DOL genes as a query gene set. P values were computed by performing a GSEA as described by Subramanian et al.⁵⁶. e, Proportion of DOLs among oligodendrocytes across mouse strains (n = 3 for each strain). Large bars correspond to the median and small bars to the IQR. f, Proportion of DAM among microglia across mouse strains (n = 3 for each strain). Large bars correspond to the median and small bars to the IQR. g, log2FC of DOL genes in cultured oligodendrocytes after treatment with Aβ at different aggregation stages.
DOL signature in non-AD pathologies
a, Analytical approach used to identify DOL signature in additional CNS pathologies. b, Volcano plot corresponding to the differential expression analysis between oligodendrocytes from control (CFA) and EAE spinal cord (Falcão et al.³⁵). DOL genes are colored in orange. c, Volcano plot corresponding to the differential expression analysis between oligodendrocytes from control (CFA) and peak EAE spinal cord (Wheeler et al.³⁶). DOL genes are colored in orange. d, Volcano plot corresponding to the differential expression analysis between oligodendrocytes from young and old mice (Dulken et al.³⁷). DOL genes are colored in orange. e, Kinetics of the mean expression level of key DOL genes across various EAE stages. P values were computed using a binomial regression with complementary log–log link function (Methods). f, Intensity of the DOL-like signature across the sections of the three LPS-treated mice (left). Hematoxylin and eosin (H&E) staining of the corresponding slides (right). FDR, false discovery rate; TPM, transcripts per million.
Spatial analysis of DOLs in mouse and human brain sections
Immunohistochemistry of DOL-like cells and analysis of spatial association with areas enriched with plaques in 5xFAD mice (a–f). a, Representative image of DOLs (OLIG2⁺ (red) SERPINA3N⁺(cyan)) from cortex of 16-month-old 5xFAD and WT mice. In 5xFAD, DOLs are in proximity to Aβ plaques (yellow); magnification ×63, and scale bar corresponds to 20 μm. Arrowheads point to OLIG2⁺SERPINA3N⁺ cells. b, Description of the experimental and computational approach used to study DOL location in mouse and human brains. Illustration was created with BioRender. c, Proportion of SERPINA3N⁺ cells in 5xFAD (n = 4) and WT (n = 2) mice. Large bars correspond to the median and small bars to the IQR. d, Proportion of SERPINA3N⁺ cells among OLIG2⁺ cells in 5xFAD and WT mice. Large bars correspond to the median and small bars to IQR. e, Normalized L functions of the SERPINA3N⁺OLIG2⁺ cells in 5xFAD sections. Each curve corresponds to a unique sample. f, Estimated interaction functions within SERPINA3N⁺OLIG2⁺ cells (blue curves) or between them and plaques (red curves) in 5xFAD mice. Each curve corresponds to a unique sample. Analysis of Visium data from postmortem human AD brains (g,h). g, Intensity of the macrophage (left) and inflammatory (right) topics across the four samples. n = 4 biologically independent samples (two AD and two NDC). The thick line corresponds to the median, and the bottom and upper limits of the box correspond to the first and third quartiles, respectively. The lower and upper whiskers correspond to the lowest and highest values, respectively, within the range of the first (third) quartile minus (plus) 1.5 times the IQR. h, H&E staining of the AD2 sample (left panel), spatial pattern of the macrophage (middle panel) and inflammatory (right panel) topics. Immunohistochemistry of postmortem human AD brains (i–k). i, Representative image of postmortem AD patient (right) and NDC (left) temporal cortex stained for oligodendrocyte marker (CC1, red), SERPINA3 (cyan), Aβ (thioflavin S, yellow) and nuclei (Hoechst, blue); magnification ×20, and scale bar corresponds to 50 μm. Arrowheads point to CC1⁺SERPINA3⁺ cells. Inset marks the position of the right image in ×63 magnification, and scale bar corresponds to 20 μm. j, Proportion of SERPINA3N⁺ cells in AD (n = 8) and NDC (n = 8) samples. Large bars correspond to the median and small bars to the IQR. P values were computed using a Wilcoxon rank test. k, Association between MMSE score and proportion of SERPINA3⁺ cells. The dashed line corresponds to an ordinary least square linear. P values were computed by testing the significance of the contribution of the MMSE score to predict the proportion of SERPINA3⁺ cells using a likelihood ratio test (LRT). CNN, convolutional neural network.
Article
Alzheimer’s disease (AD) is a complex neurodegenerative disease, perturbing neuronal and non-neuronal cell populations. In this study, using single-cell transcriptomics, we mapped all non-immune, non-neuronal cell populations in wild-type and AD model (5xFAD) mouse brains. We identified an oligodendrocyte state that increased in association with brain pathology, which we termed disease-associated oligodendrocytes (DOLs). In a murine model of amyloidosis, DOLs appear long after plaque accumulation, and amyloid-beta (Aβ) alone was not sufficient to induce the DOL signature in vitro. DOLs could be identified in a mouse model of tauopathy and in other murine neurodegenerative and autoimmune inflammatory conditions, suggesting a common response to severe pathological conditions. Using quantitative spatial analysis of mouse and postmortem human brain tissues, we found that oligodendrocytes expressing a key DOL marker (SERPINA3N/SERPINA3 accordingly) are present in the cortex in areas of brain damage and are enriched near Aβ plaques. In postmortem human brain tissue, the expression level of this marker correlated with cognitive decline. Altogether, this study uncovers a shared signature of oligodendrocytes in central nervous system pathologies.
 
Microglia and infiltrating immune cells in brain tissue from patients with DRE
a, Position and phenotype of clusters on the t-SNE map. Color represents the cluster ID. b, Surface epitope expression of lineage-specific cellular markers quantified using antibody staining with the CITE-seq protocol was overlaid on the t-SNE map to identify the cluster phenotype. Color intensity reflects cellular surface epitope protein expression. c, Stacked bar chart shows the frequency of infiltrating, resident (microglial) immune cells and NVU cells from brain tissues of patients with DRE. Bar color reflects cell types as indicated in the figure. OL, olfactory lobe; FL, frontal lobe; TL, temporal lobe. P1.A, occipital cortex; P1.B, occipital core; P2, frontal lobe; P3.A, posterior mid-temporal gyrus; P3.B, superior frontal gyrus; P4, posterior mid-temporal gyrus; P5.A, posterior mid-temporal gyrus; P5.B, lateral mid-temporal gyrus; P6.A, mid-temporal gyrus; P6.B, posterior temporal gyrus; P6.C, lateral temporal gyrus.
Expression of pro-inflammatory and microglial-specific genes in patients with DRE, non-neurological disease controls and brain tissue of patients with ASD
a, Distribution of microglial cells from non-neurological disease controls (CON) and patients with ASD on a t-SNE map. b,c, snRNA-seq dataset (b) and scRNA-seq dataset (c) show expression of microglial marker genes overlaid on a t-SNE map. d–f, Expression of pro-inflammatory cytokines and chemokine marker genes. Normalized gene expression levels were overlaid on the t-SNE map. g–i, Multispectral Opal dye IHC imaging of brain tissue sections from control tissue (g) and DRE lesion tissue (h). i, Magnified image from the DRE tissue section. FFPE (5-µm) tissue sections were stained with panels of antibodies for microglia (AIF-1), macrophages (CD68), T cells (CD3), neurons (microtubule-associated protein 2 (MAP2)), astrocytes (GFAP) and the pro-inflammatory cytokine IL-1b. After staining with all the antibodies, sections were stained with 4,6-diamidino-2-phenylindole (DAPI) for a nuclear stain. Tissues were imaged using the Vectra 3 imaging system with a 40× view finder. White boxes and numbers in h correspond to the magnified image in i. FFPE brain tissue sections from controls (n = 4) and DRE lesions (n = 4) were stained and imaged, and a representative image from one sample is shown. The colors that represent the antibody and nuclear stain are shown. Data from Velmeshev et al.¹⁹ (a,b,e); data from the Masuda et al.¹⁸ (c,f).
Infiltrating immune cells in the epileptic human brain and their interaction with NVU cells
a, Cluster positions are shown on a t-SNE map where color represents cluster identity. b, Surface epitope expression of lineage-specific cellular markers quantified using antibody staining with the CITE-seq protocol was overlaid on a t-SNE map to identify the cluster phenotypes. Color intensity reflects the expression of cellular surface proteins. c, LR interaction network between NVUs and immune cell clusters is shown as a directed network graph. Network nodes reflect cluster ID, and edges shown as colored arrows reflect the potential interaction between cognate LR pairs. Arrow direction shows signaling from ligands (arrow tail) to receptors (arrowhead). Thickness and color of the arrows reflect the number of LR pairs found between the two nodes. Node ID is shown as imm_ (for immune cell clusters) and micro_20, micro_13,and micro_22 for NVU clusters. Node colors show the cell type. d, Network of all ligand and receptor gene pairs found enriched between clusters of NVUs and immune cells. Here nodes represent ligands (blue circles) and receptors (yellow circles), and edges show cognate ligand and receptor interactions. Recently updated gene names include CYR61 (CCN1), CTGF (CCN2), NOV (CCN3), MLLT4 (AFDN), PVRL2 (NECTIN2). Arrow directions show signaling from ligands (arrow tail) to receptors (arrowhead). Arrow colors show the number of cluster pairs for which an LR pair was enriched.
LR genes significantly modulated in a TLE epilepsy mouse model
a,b, LR network genes enriched in human epileptic brain foci (Fig. 3d and Extended Data Fig. 9) were investigated for differential gene regulation in a TLE mouse model compared with control mice. RNA-seq data from hippocampal brain tissue from mice with TLE (n = 100) and control mice (n = 100) were analyzed. Expression of significantly differentially regulated (exact test and FDR < 0.05) ligand (a) and receptor (b) genes is shown as a heatmap. Each row of the heatmap shows a mouse gene and its human ortholog gene (shown in uppercase letters), and each column represents data from an individual mouse.
Direct interaction of microglia and infiltrating T cells in brain tissue from patients with refractory epilepsy
a, Doublet cell clusters are shown on the t-SNE map. Numbers and colors on the t-SNE map show the cluster ID. b, Surface epitope protein expression of major lineage markers are overlaid on the t-SNE map. c, Gene expression specific to NK cells is overlaid on the t-SNE map. d,f, Gene expression profile of physically interacting CD4⁺ T and microglial cells (d) and CD8⁺ T and microglial cells (f). The bar at the bottom (blue, microglia; yellow, T cells) shows the estimated mixing factor. Heatmap and mixing factor bars are ordered with increasing mixing factor value for T cells. Left, colored bar indicates the ratio of expected gene expression in microglia versus T cells. The top ten genes specific to T cells (lower microglial/T cell ratio) and specific to microglia (higher microglial/T cell ratio) are shown in the heatmap. e,g, Real gene expression values in microglia and T cells. Heatmaps were plotted for 500 randomly drawn cells, 250 each from microglia and T cells. h,i, Immune cells isolated from DRE tissue were formalin fixed, and cells were cytocentrifuged with Cytospin on slides for staining of CD3, AIF-1 and IL-1b. DAPI was used for the nuclear stain. Stained slides were imaged using a Vectra 3.0 imaging microscope. h, T cell (CD3⁺)–microglial (AIF-1) immune cell complex. i, Representative T cell–microglial immune cell complex producing IL-1b from the brain tissue of three patients with DRE. j, T cell–microglial immune cell complex from one of the FFPE tissue sections stained with a panel of six antibodies. A CD3⁺ T cell (orange) in physical interaction with an AIF-1⁺ microglia (yellow) is indicated with a white arrow, and IL-1b proteins are shown (red). k, Bivariate flow cytometry plot with gating for CD45, CD11b and CD3. Left, live gated cells with SSC on the y axis and CD45 expression on the x axis. Right, CD45hi-gated cells with CD3 expression on the y axis and CD11b expression on the x axis. CD115 levels were overlaid on the bivariate plot, where expression is indicated from low (green) to high (red). Flow cytometry analysis and plots were created using FlowJo software.
Article
Epileptogenic triggers are multifactorial and not well understood. Here we aimed to address the hypothesis that inappropriate pro-inflammatory mechanisms contribute to the pathogenesis of refractory epilepsy (non-responsiveness to antiepileptic drugs) in human patients. We used single-cell cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) to reveal the immunotranscriptome of surgically resected epileptic lesion tissues. Our approach uncovered a pro-inflammatory microenvironment, including extensive activation of microglia and infiltration of other pro-inflammatory immune cells. These findings were supported by ligand–receptor (LR) interactome analysis, which demonstrated potential mechanisms of infiltration and evidence of direct physical interactions between microglia and T cells. Together, these data provide insight into the immune microenvironment in epileptic tissue, which may aid the development of new therapeutics. Single-cell analysis of immune cells from surgically resected human epileptic brain tissues showed heterogeneity and pro-inflammatory signaling in microglia and evidence for direct interaction of microglia with T cells.
 
Article
Progressive multiple sclerosis (MS) is characterized by unrelenting neurodegeneration, which causes cumulative disability and is refractory to current treatments. Drug development to prevent disease progression is an urgent clinical need yet is constrained by an incomplete understanding of its complex pathogenesis. Using spatial transcriptomics and proteomics on fresh-frozen human MS brain tissue, we identified multicellular mechanisms of progressive MS pathogenesis and traced their origin in relation to spatially distributed stages of neurodegeneration. By resolving ligand–receptor interactions in local microenvironments, we discovered defunct trophic and anti-inflammatory intercellular communications within areas of early neuronal decline. Proteins associated with neuronal damage in patient samples showed mechanistic concordance with published in vivo knockdown and central nervous system (CNS) disease models, supporting their causal role and value as potential therapeutic targets in progressive MS. Our findings provide a new framework for drug development strategies, rooted in an understanding of the complex cellular and signaling dynamics in human diseased tissue that facilitate this debilitating disease. By complementing spatial transcriptomics with high-resolution proteomics, Kaufmann et al. tracked a gradient of disease severity across the brains of patients with progressive multiple sclerosis, uncovering new therapeutic opportunities to slow disease.
 
Phylogenetic relations of KCRs with other known ChRs
Branches are colored to distinguish between different ChR families. Leaves corresponding to the KCRs characterized in this study are shown as thick red lines. A full list of other ChR sequences used to create the tree can be found in ref. ⁴². The gray circles show ultrafast bootstrap support values above 95%.
KCR photocurrents evoked by continuous light and 2P excitation
a, Action spectra of photocurrents in the visible range (mean ± s.e.m., n = 6 and n = 10 cells for HcKCR1 and HcKCR2, respectively) constructed from measurements of their initial slopes (Methods). b, Photocurrent traces recorded from HcKCR1 in response to 1-s light pulses (540 nm, 6.3 mW mm⁻²) upon 20-mV voltage increments in the Na⁺ bath. d, Photocurrent traces recorded from HcKCR1 as in b but in the K⁺ bath. c, IV curves (mean ± s.e.m., n = 7 cells from three independent experiments) in the Na⁺ (red) and K⁺ (blue), pH 7.4. f, IV curves (mean ± s.e.m., n = 7 cells from three independent experiments) in the NMDG⁺ bath, pH 7.4 (black) and K⁺ bath, pH 9.4 (magenta). e, Relative permeabilities (PX/PK); lines, mean ± s.e.m. (n = 7 cells from three independent experiments for Li⁺, Cs⁺, NMDG⁺, Mg²⁺ and Ca²⁺, n = 8 cells for Rb⁺ and n = 10 cells for Na⁺, each from four independent experiments); symbols, data from individual cells. g, Photocurrent traces recorded upon 2P excitation (HcKCR1, 1,040 nm, 3.8 mW; HcKCR2, 1,000 nm, 3.5 mW). h, Dependence of 2P photocurrents on the laser power (mean ± s.e.m., n = 7 and n = 8 cells for HcKCR1 and HcKCR2, respectively; each from three independent experiments). i, Action spectra of photocurrents under 2P excitation (mean ± s.e.m., n = 6 and n = 10 cells, each from three independent experiments, for HcKCR1 and HcKCR2, respectively). Source data are provided.
Source data
Determination of HcKCR1 unitary conductance
a, Current noise in the dark (black) and under 540 nm, 60 μW mm⁻² illumination (red) recorded from a HcKCR1-transfected HEK293 cell. RMS, root mean square. b, Power spectra of the noise in the dark (black) and under illumination (red). c, Difference (light minus dark) power spectrum (black) and its Lorenzian fit (red). γ, unitary conductance. For more detail, see Methods.
Photocurrents and photochemical conversion upon single quantum excitation
a, Photocurrent traces (thin solid lines) recorded from HcKCR1 in response to laser flashes upon 20-mV voltage increments superimposed with their multiexponential approximations (thick dashed lines). b, Voltage dependence of the three kinetic components of channel currents. c, Dependence of the channel current Vrev with 130 mM KCl in the bath and pipette on bath pH (mean ± s.e.m., n = 3, 8 and 6 cells from three independent experiments for pH 5.4, 7.4 and 9.4, respectively). d, Timecourse of the peak current recovery (mean ± s.e.m., n = 5 cells from three independent experiments). e, Transient absorbance changes (blue and black) and channel current (red). f, Photocurrent traces in the absence of permeant metal cations at bath pH 7.4. g, Voltage dependence of active current at different bath pH in the absence of permeant metal cations. h, Transient absorbance changes at 400 nm (blue), compared with active and passive (channel) currents (red and black, respectively). i, Photocurrent traces in the absence of permeant metal cations at bath pH 9.4. Source data are provided.
Source data
Photoactivation of HcKCR1 in neurons generates robust photocurrents and efficiently suppresses neuronal firing
a, Fluorescent images of a cortical slice showing HcKCR1-EYFP and tdTomato expression in layer 2/3 neurons. Similar results were observed in 14 slices from two male and two female mice at the age of 3–4 weeks. Cortical layers were identified by 4′,6-diamidino-2-phenylindole staining. L, layer. b, Photocurrents traces of a HcKCR1-expressing neuron in response to a 1-s 565 nm light pulse (13.1 mW mm⁻²) at holding voltages increased in 5-mV steps. c, The fast τ of photocurrent desensitization at the indicated voltages. P = 0.2 by the two-tailed Wilcoxon test. d, Ratios of photocurrent at the end of illumination to the peak photocurrent. ***P = 0.0004 by the two-tailed paired t-test. e,f, IV curves of peak photocurrent (e) and photocurrent at the end of illumination (f) in individual neurons indicated by different colors. P < 0.0001, R² > 0.98 for all linear regressions. g, Reversal potentials calculated from the data in e and f. ****P < 0.0001 by the two-tailed paired t-test. h, Membrane voltage traces of a HcKCR1-expressing neuron in response to 0.2 (left) or 0.5 nA (right) current injections without (top) and with (bottom) 565 nm light pulses. i, The frequencies of action potentials evoked by different current injections with (magenta) and without (black) photoactivation. P > 0.9999 for 0.1 nA and **P = 0.039 for 0.2–0.5 nA by the multiple Wilcoxon matched-pairs signed rank test with Bonferroni–Dunn multiple corrections. Data in c,d,g and i are expressed as mean ± s.e.m., n = 9 neurons in c,d and g and n = 8 neurons in i from one male and one female mouse at the age of 3–4 weeks. Statistics source data are provided.
Source data
Article
Channelrhodopsins are used widely for optical control of neurons, in which they generate photoinduced proton, sodium or chloride influx. Potassium (K+) is central to neuron electrophysiology, yet no natural K+-selective light-gated channel has been identified. Here, we report kalium channelrhodopsins (KCRs) from Hyphochytrium catenoides. Previously known gated potassium channels are mainly ligand- or voltage-gated and share a conserved K+-selectivity filter. KCRs differ in that they are light-gated and have independently evolved an alternative K+ selectivity mechanism. The KCRs are potent, highly selective of K+ over Na+, and open in less than 1 ms following photoactivation. The permeability ratio PK/PNa of 23 makes H. catenoides KCR1 (HcKCR1) a powerful hyperpolarizing tool to suppress excitable cell firing upon illumination, demonstrated here in mouse cortical neurons. HcKCR1 enables optogenetic control of K+ gradients, which is promising for the study and potential treatment of potassium channelopathies such as epilepsy, Parkinson’s disease and long-QT syndrome and other cardiac arrhythmias. The authors report a functional class of channelrhodopsins that are highly selective for K+ over Na+. These light-gated channels, named ‘kalium channelrhodopsins’, enable robust inhibition of mouse cortical neurons with millisecond precision.
 
Article
Proper neural progenitor behavior in conjunction with orderly vasculature formation is fundamental to the development of the neocortex. However, the mechanisms coordinating neural progenitor behavior and vessel growth remain largely elusive. Here we show that robust metabolic production of lactate by radial glial progenitors (RGPs) co-regulates vascular development and RGP division behavior in the developing mouse neocortex. RGPs undergo a highly organized lineage progression program to produce diverse neural progeny. Systematic single-cell metabolic state analysis revealed that RGPs and their progeny exhibit distinct metabolic features associated with specific cell types and lineage progression statuses. Symmetrically dividing, proliferative RGPs preferentially express a cohort of genes that support glucose uptake and anaerobic glycolysis. Consequently, they consume glucose in anaerobic metabolism and produce a high level of lactate, which promotes vessel growth. Moreover, lactate production enhances RGP proliferation by maintaining mitochondrial length. Together, these results suggest that specific metabolic states and metabolites coordinately regulate vasculature formation and progenitor behavior in neocortical development. Dong et al. show that proliferative neural progenitors in the developing neocortex preferentially undergo anaerobic glycolysis and generate a high level of lactate that coordinately regulates vasculature outgrowth and progenitor behavior.
 
Article
Microglia are the resident macrophages of the CNS that serve critical roles in brain construction. Although human brains contain microglia by 4 weeks gestation, an understanding of the earliest microglia that seed the brain during its development remains unresolved. Using time-lapse imaging in zebrafish, we discovered a mrc1a⁺ microglia precursor population that seeds the brain before traditionally described microglia. These early microglia precursors are dependent on lymphatic vasculature that surrounds the brain and are independent of pu1⁺ yolk sac-derived microglia. Single-cell RNA-sequencing datasets reveal Mrc1⁺ microglia in the embryonic brains of mice and humans. We then show in zebrafish that these early mrc1a⁺ microglia precursors preferentially expand during pathophysiological states in development. Taken together, our results identify a critical role of lymphatics in the microglia precursors that seed the early embryonic brain.
 
Article
The cerebellum, a primary brain structure involved in the control of sensorimotor tasks, also contributes to higher cognitive functions including reward, emotion and social interaction. Although the regulation of these behaviors has been largely ascribed to the monoaminergic system in limbic regions, the contribution of cerebellar dopamine signaling in the modulation of these functions remains largely unknown. By combining cell-type-specific transcriptomics, histological analyses, three-dimensional imaging and patch-clamp recordings, we demonstrate that cerebellar dopamine D2 receptors (D2Rs) in mice are preferentially expressed in Purkinje cells (PCs) and regulate synaptic efficacy onto PCs. Moreover, we found that changes in D2R levels in PCs of male mice during adulthood alter sociability and preference for social novelty without affecting motor functions. Altogether, these findings demonstrate novel roles for D2R in PC function and causally link cerebellar D2R levels of expression to social behaviors. Cutando and colleagues show that changes in D2 dopamine receptor levels in cerebellar Purkinje cells alter sociability and preference for social novelty without affecting motor functions.
 
Article
Sleep disturbances frequently occur in neurodevelopmental disorders such as autism, but the developmental role of sleep is largely unexplored, and a causal relationship between developmental sleep defects and behavioral consequences in adulthood remains elusive. Here, we show that in mice, sleep disruption (SD) in adolescence, but not in adulthood, causes long-lasting impairment in social novelty preference. Furthermore, adolescent SD alters the activation and release patterns of dopaminergic neurons in the ventral tegmental area (VTA) in response to social novelty. This developmental sleep function is mediated by balanced VTA activity during adolescence; chemogenetic excitation mimics, whereas silencing rescues, the social deficits of adolescent SD. Finally, we show that in Shank3-mutant mice, improving sleep or rectifying VTA activity during adolescence ameliorates adult social deficits. Together, our results identify a critical role of sleep and dopaminergic activity in the development of social interaction behavior. Bian et al. show that sleep during adolescence is crucial for shaping the preference for novel social stimulation in adulthood. This developmental role of sleep is mediated by balanced levels of VTA activity during adolescent brain development.
 
Article
Epilepsy is the most common childhood neurological disease, and nearly 20% of affected children develop drug-resistant childhood epilepsy (DRCE). Using single-cell analysis methods, Kumar et al. have identified pro-inflammatory interactions between microglia and T cells in brain tissue from individuals with DRCE. This work may help to identify therapeutic targets for DRCE.
 
Article
Can studying individual differences in brain structure and function reveal individual differences in behavior? Analyses of MRI data from nearly 50,000 individuals may suggest that the possibility is fleeting. Although sample size is important for brain-based prediction, researchers can take other steps to build better biomarkers. These include testing model generalizability across people, datasets, and time points and maximizing model robustness by optimizing brain data acquisition, behavioral measures, and prediction approaches.
 
Article
Reward expectations based on internal knowledge of the external environment are a core component of adaptive behavior. However, internal knowledge may be inaccurate or incomplete due to errors in sensory measurements. Some features of the environment may also be encoded inaccurately to minimize representational costs associated with their processing. In this study, we investigated how reward expectations are affected by features of internal representations by studying behavior and dopaminergic activity while mice make time-based decisions. We show that several possible representations allow a reinforcement learning agent to model animals’ overall performance during the task. However, only a small subset of highly compressed representations simultaneously reproduced the co-variability in animals’ choice behavior and dopaminergic activity. Strikingly, these representations predict an unusual distribution of response times that closely match animals’ behavior. These results inform how constraints of representational efficiency may be expressed in encoding representations of dynamic cognitive variables used for reward-based computations.
 
Article
Neural computations are currently investigated using two separate approaches: sorting neurons into functional subpopulations or examining the low-dimensional dynamics of collective activity. Whether and how these two aspects interact to shape computations is currently unclear. Using a novel approach to extract computational mechanisms from networks trained on neuroscience tasks, here we show that the dimensionality of the dynamics and subpopulation structure play fundamentally complementary roles. Although various tasks can be implemented by increasing the dimensionality in networks with fully random population structure, flexible input–output mappings instead require a non-random population structure that can be described in terms of multiple subpopulations. Our analyses revealed that such a subpopulation structure enables flexible computations through a mechanism based on gain-controlled modulations that flexibly shape the collective dynamics. Our results lead to task-specific predictions for the structure of neural selectivity, for inactivation experiments and for the implication of different neurons in multi-tasking. Neural computations are envisioned as arising from either distinct function subpopulations or distributed collective dynamics. Dubreuil and Valente et al. examined recurrent neural networks trained on various cognitive tasks and found that a mixed-selective yet non-random subpopulation structure enabled flexible responding through gain-modulated latent dynamics.
 
Article
The solutions found by neural networks to solve a task are often inscrutable. We have little insight into why a particular structure emerges in a network. By reverse engineering neural networks from dynamical principles, Dubreuil, Valente et al. show how neural population structure enables computational flexibility.
 
Article
Learning induces the formation of new excitatory synapses in the form of dendritic spines, but their functional properties remain unknown. Here, using longitudinal in vivo two-photon imaging and correlated electron microscopy of dendritic spines in the motor cortex of mice during motor learning, we describe a framework for the formation, survival and resulting function of new, learning-related spines. Specifically, our data indicate that the formation of new spines during learning is guided by the potentiation of functionally clustered preexisting spines exhibiting task-related activity during earlier sessions of learning. We present evidence that this clustered potentiation induces the local outgrowth of multiple filopodia from the nearby dendrite, locally sampling the adjacent neuropil for potential axonal partners, likely via targeting preexisting presynaptic boutons. Successful connections are then selected for survival based on co-activity with nearby task-related spines, ensuring that the new spine preserves functional clustering. The resulting locally coherent activity of new spines signals the learned movement. Furthermore, we found that a majority of new spines synapse with axons previously unrepresented in these dendritic domains. Thus, learning involves the binding of new information streams into functional synaptic clusters to subserve learned behaviors. Learning induces formation of dendritic spines, but their functional properties are unknown. The authors show that new spines bind new presynaptic inputs into preexisting spine clusters, generating locally coherent inputs representing learned behaviors.
 
Article
Autophagy is markedly impaired in Alzheimer’s disease (AD). Here we reveal unique autophagy dysregulation within neurons in five AD mouse models in vivo and identify its basis using a neuron-specific transgenic mRFP-eGFP-LC3 probe of autophagy and pH, multiplex confocal imaging and correlative light electron microscopy. Autolysosome acidification declines in neurons well before extracellular amyloid deposition, associated with markedly lowered vATPase activity and build-up of Aβ/APP-βCTF selectively within enlarged de-acidified autolysosomes. In more compromised yet still intact neurons, profuse Aβ-positive autophagic vacuoles (AVs) pack into large membrane blebs forming flower-like perikaryal rosettes. This unique pattern, termed PANTHOS (poisonous anthos (flower)), is also present in AD brains. Additional AVs coalesce into peri-nuclear networks of membrane tubules where fibrillar β-amyloid accumulates intraluminally. Lysosomal membrane permeabilization, cathepsin release and lysosomal cell death ensue, accompanied by microglial invasion. Quantitative analyses confirm that individual neurons exhibiting PANTHOS are the principal source of senile plaques in amyloid precursor protein AD models.
 
Top-cited authors
Hongkui Zeng
  • Allen Institute for Brain Science
Susan M Sunkin
  • Allen Institute for Brain Science
Allan Jones
  • Allen Institute for Brain Science
Lydia Ng
  • Allen Institute for Brain Science
Eric J Nestler
  • Icahn School of Medicine at Mount Sinai