Spatially mapped single-cell chromatin accessibility

Abstract

High-throughput single-cell epigenomic assays can resolve cell type heterogeneity in complex tissues, however, spatial orientation is lost. Here, we present single-cell combinatorial indexing on Microbiopsies Assigned to Positions for the Assay for Transposase Accessible Chromatin, or sciMAP-ATAC, as a method for highly scalable, spatially resolved, single-cell profiling of chromatin states. sciMAP-ATAC produces data of equivalent quality to non-spatial sci-ATAC and retains the positional information of each cell within a 214 micron cubic region, with up to hundreds of tracked positions in a single experiment. We apply sciMAP-ATAC to assess cortical lamination in the adult mouse primary somatosensory cortex and in the human primary visual cortex, where we produce spatial trajectories and integrate our data with non-spatial single-nucleus RNA and other chromatin accessibility single-cell datasets. Finally, we characterize the spatially progressive nature of cerebral ischemic infarction in the mouse brain using a model of transient middle cerebral artery occlusion.

Fig. 1. sciMAP-ATAC schematic and performance.

a sciMAP-ATAC workflow. Cryosectioning of alternating 20 µm (histological) and 100–300 µm (sciMAP-ATAC) slices are obtained. Thin (20 µm) slices are stained and imaged for use in spatial registration (scale bar, 1 mm) to a reference atlas (Allen Mouse Brain Atlas: http://atlas.brain-map.org/atlas?atlas=1&plate=100960312, ref. ). Thick (100–300 µm) slices are carried through high-density microbiopsy punching (100–500 µm diameter) in the cryostat chamber. Punches are placed directly into wells of a microwell plate for nuclei isolation, and washed prior to splitting into multiple wells for indexed transposition and the sci-ATAC-seq workflow. b Four punch volumes were assessed for nuclei yield using either a 250 or 500 μm diameter punch on a 200 or 300 μm thick section. Total nuclei isolated for each punch is shown on the left, and normalized for tissue voxel volume on the right, representing the efficiency of extraction from each punch, for punches with dimensions 250 ×200 μm (n = 48), 250 × 300 μm (n = 15), 500 × 200 μm (n = 46), and 500 × 300 μm (n = 7). Center line represents median, lower and upper hinges represent first and third quartiles, and whiskers extend from hinge to ±1.5 × IQR. c Passing reads per cell from sci-ATAC-seq (n = 4102 cells examined from a single mouse brain experiment) and sciMAP-ATAC (n = 15,552 cells examined from two independent mouse brain experiments), which are comparable at the level of depth sequenced. Center line represents median, lower and upper hinges represent first and third quartiles, whiskers extend from hinge to ±1.5 × IQR, individual cells represented as colored dots. d ATAC read signal at transcription start sites (TSSs) and surrounding base pairs (bps) for sci-ATAC-seq and sciMAP-ATAC. Enrichment for sci-ATAC-seq is greater than that of sciMAP-ATAC, likely due to increased processing time of isolated nuclei prior to transposition. e UMAP of sciMAP-ATAC and sci-ATAC-seq libraries from mouse brain group closely together. Asterisk indicates a population of 734 cells, derived from spinal cord, which was not sampled during microbiopsy punching. Source data are provided as a Source data file.

Fig. 2. sciMAP-ATAC reveals spatially distinct cell type composition in the mouse somatosensory cortex.

a Experiment schematic of sciMAP-ATAC in the mouse somatosensory cortex. b DAPI and SATB2 staining of SSp cortex from sciMAP-ATAC histological section (scale bar, 50 µm) in reference to matched rerference atlas image (Allen Mouse Brain Atlas: http://atlas.brain-map.org/atlas?atlas=1&plate=100960312, ref. ). c UMAP of 7779 cells colored by punch location category. Each category contains cells from 32 spatially distinct tissue punches. d UMAP as in c, colored by cell type (OPC oligodendrocyte precursor cells, Int Olig intermediate oligodendrocytes, Mat Olig mature oligodendrocytes, Astro astrocytes, Endo endothelia, Micro microglia, MSN medium spiny neurons, GABA GABAergic (inhibitory) neurons, Glut glutamatergic (excitatory) neurons). e ATAC-seq profiles for cells aggregated by cell type for marker genes; colored by cell type as in d. f Aggregate cell type composition over punches belonging to the broad region categories; colored by cell type as in d. g Cell type composition for each of the 96 individual punches split by broad region category; colored by cell type as in d. Source data are provided as a Source data file.

Fig. 3. sciMAP-ATAC enables the analysis and comparison of cells and cell types from individual spatial positions.

a Topic weight matrix for cells present only in a single punch (F5, inner cortex punch), annotated by cell type (bottom); colored by cell type from the full dataset (Fig. 2d). b UMAP of cells from punch F5 showing spatially distinct groupings for cell type; colored by cell type from the full dataset (Fig. 2d). c Isolated analysis of cells from Punch F5 for peak calling, topic modeling, and visualized via UMAP; colored by cell type from the full dataset (Fig. 2d). d Two major clusters identified from the isolated analysis of punch F5 (Glut glutamatergic (excitatory) neurons). e Transcription factor motif enrichments for the isolated analysis of punch F5, indicating that cluster 1 (n = 44 cells) is made up of glutamatergic neurons and cluster 2 (n = 45 cells) is made up of other cell types. Center line represents median, lower, and upper hinges represent first and third quartiles, whiskers extend from hinge to ±1.5 × IQR, individual cells represented as colored dots. f UMAP of all glutamatergic neuron cells from two adjacent punches (C5, inner cortex, and B5, outer cortex) after topic modeling on the isolated cell profiles. g Transcription factor motif enrichments for glutamatergic cells from adjacent punches from inner cortex (n = 39 cells) and outer cortex (n = 30 cells) shown in f; colored by individual punch as in f. Two-sided Mann–Whitney U test with Bonferroni–Holm correction. Center line represents median, lower and upper hinges represent first and third quartiles, whiskers extend from hinge to ±1.5 × IQR, individual cells represented as colored dots. h Motif enrichments across glutamatergic neurons across all punch pairs. TFME transcription factor motif enrichment. Source data are provided as a Source data file.

Fig. 4. sciMAP-ATAC trajectories through the human primary visual cortex.

a sciMAP-ATAC punching schematic showing one of three adjacent sections from one individual. A total of 21 eight-punch trajectories (T) spanning the cortex were produced. b UMAP of cells colored by position within their respective trajectory as in a. Top right shows the same UMAP with all cells grayed out with the exception of cells from the third trajectory from section 2. Bottom right shows all cells grayed out with the exception of cells from a single punch; the outermost cortical position (1) from the third trajectory of the second section. c UMAP as in b colored by cell type (OPC oligodendrocyte precursor cells, Olig oligodendrocytes, Astro astrocytes, Micro microglia, GABA GABAergic (inhibitory) neurons, Glut glutamatergic (excitatory) neurons). d ATAC-seq profiles for cells aggregated by cell type for marker genes; colored by cell type as in c. e Aggregate cell type composition across the 21 trajectories (n = 4547 cells over 188 independent punches); colored by cell type as in d. Data are presented as mean values ± SD. f Cell type composition for each of the 188 individual punches split by trajectory position. Punches from the WM indicated by an asterisk are aggregated by section. Colored by cell type as in d. Source data are provided as a Source data file.

Fig. 5. Integration of sciMAP-ATAC with snRNA-seq and scTHS-seq human VISp datasets.

a Co-embedding of sciMAP-ATAC and scTHS-seq cell profiles from Lake et al. using Signac in a joint UMAP. Top right shows only scTHS-seq cells colored by cell type identified in Lake et al. and bottom shows sciMAP-ATAC cells colored by our called cell types as in Fig. 4c, except for glutamatergic neurons which are colored by spatial positions 1–8 (Glut glutamatergic (excitatory) neurons, GABA GABAergic (inhibitory) neurons, Astro astrocytes, Micro microglia, Olig oligodendrocytes, OPC oligodendrocyte precursor cells, Endo endothelial cells, NA not applicable—no cell type provided). b Co-embedding of sciMAP-ATAC and snRNA-seq transcriptional profiles from Lake et al. using Signac. Top right shows only snRNA-seq cells. Abbreviations as in a, but with the addition of Per = pericytes, and glutamatergic (excitatory) neurons (Ex) are colored by subtype identified in Lake et al.. Bottom right shows only sciMAP-ATAC cells, with glutamatergic neurons colored by spatial position 1–8. c Confusion matrix representing the percent agreement in predicting the cell type of a cell from one dataset using the other between sciMAP-ATAC and scTHS-seq cells. d As in c, but between sciMAP-ATAC and snRNA-seq. Spatial agreement between excitatory neuron subtypes identified in the snRNA-seq data correspond to the spatial positioning of cells within our sciMAP-ATAC dataset. Source data are provided as a Source data file.

Fig. 6. sciMAP-ATAC shows spatial epigenetic patterns of glutamatergic neurons.

a Isolation and UMAP visualization of human VISp glutamatergic neurons from all cells (top right), colored by punch position. An interactive, three-dimensional UMAP embedding is available as Supplementary Data 4. b ATAC-seq profiles for glutamatergic neurons along trajectory positions for layer (L)-specific marker genes CALB1 (layers 2 and 3), LMO4 (layer 5), and CTGF (layer 6b); colored by punch position as in a. c Cells from section 1, Trajectory 4 (T1.4, top) are shown in color on the UMAP of all cells, with other cells shown in gray (bottom); colored by position as in a. d UMAP of glutamatergic neurons from Trajectory 1.4 after topic modeling on the isolated cells; colored by position along the trajectory as in a. e DNA-binding motif enrichment for layer-specific factors for Trajectory 1.4 shown in d, with cells split by their positions along the trajectory. Source data are provided as a Source data file.

Fig. 7. sciMAP-ATAC applied to a mouse model of ischemic injury.

a Experimental design using a mouse MCAO model of ischemic injury. Mice were sacrificed 3 days post surgery (dps) and brains flash-frozen in TFM. Alternating thin (20 µm) and thick (200 µm) sections were processed using IHC to define infarction (red outline) and peri-infarct area (pink outline) and sciMAP-ATAC punching schematic, respectively. b GFAP IHC of a 20 µm coronal section of an ischemic mouse brain. Punch positions along the 5–8 axis (core-to-border) are indicated. Background corrected GFAP fluorescence along the 5–8 axis is shown to the right for stroke and contralateral hemispheres (n = 10). Data are presented as linear fitted model ± SEM; boxplot center line represents median, lower and upper hinges represent first and third quartiles, and whiskers extend from hinge to ±1.5 × IQR, (scale bar, 1 mm) c. UMAP of cells colored by the three conditions. d UMAP as in c, colored by clusters assigned to cell type (Olig oligodendrocytes, Astro astrocytes, Micro/MΦ microglia/macrophage, GABA GABAergic (inhibitory) neurons, Glut glutamatergic (excitatory) neurons). e Cell × topic matrix colored by normalized topic weights, as in c, d and annotated by conditions and cell type as given at the bottom reveals substantially divergent topic weighting in cells from the stroke punches (left). Topic 30, enriched specifically in the stroke cells belonging to the chromatin-disrupted cluster, has peaks enriched for ontologies associated with ischemic injury with reperfusion. Colored by −log₁₀ false discovery rate (FDR) Q-value, height by log₂ fold enrichment (right). Source data are provided as a Source data file.

Fig. 8. Spatially progressive epigenetic remodeling in ischemic injury.

a Volcano plot of Z-scored transcription factor (TF) motif enrichment slope change across punch positions 5–8 (Δslope = slope_stroke − slope_{contralateral}) by −log10 p value of the two-way ANOVA from the interaction of TF motif enrichment per punch by condition (stroke, contralateral) without multiple comparison correction. Colored by significance (N.S. not significant, Sig. Δ slope = significant change in slope, Sig. Δ Slope and p = significant change in slope and significant p value). b Top hits for significantly different changes in TF motif enrichment over space as compared between stroke (red) and contralateral (blue); KLF9 (top) and BHLHA15 (bottom). −Log10 p value of the two-way ANOVA from the interaction of TF motif enrichment per punch by condition without multiple comparison correction. Data are presented as linear fitted model ± SEM. c KLF9 TF motif enrichment over space reveals cell type contribution to KLF9 enrichment from infarct core to peri-infarct area. Cell types as defined in Fig. 7d. d Comparison of TF motif enrichment at the infarct border (punch position 8) between stroke (punch 40) and contralateral (punch 48) single-cell profiles. Oligodendrocyte (Olig) TF motif enrichment shown for BCL11B and RXRG for punch 40 (n = 4 cells) and punch 48 (n = 6 cells). Glutamatergic neuron (Glut) TF motif enrichment shown for KLF4 and KLF7 for punch 40 (n = 5 cells) and punch 48 (n = 3 cells). Two-sided Mann–Whitney U test with Bonferroni–Holm correction. Center line represents median, lower and upper hinges represent first and third quartiles, whiskers extend from hinge to ±1.5 × IQR, individual cells represented as colored dots. Source data are provided as a Source data file.