A single-cell atlas deconstructs heterogeneity across multiple models in murine traumatic brain injury and identifies novel cell-specific targets

Abstract

Traumatic brain injury (TBI) heterogeneity remains a critical barrier to translating therapies. Identifying final common pathways/molecular signatures that integrate this heterogeneity informs biomarker and therapeutic-target development. We present the first large-scale murine single-cell atlas of the transcriptomic response to TBI (334,376 cells) across clinically relevant models, sex, brain region, and time as a foundational step in molecularly deconstructing TBI heterogeneity. Results were unique to cell populations, injury models, sex, brain regions, and time, highlighting the importance of cell-level resolution. We identify cell-specific targets and previously unrecognized roles for microglial and ependymal subtypes. Ependymal-4 was a hub of neuroinflammatory signaling. A distinct microglial lineage shared features with disease-associated microglia at 24 h, with persistent gene-expression changes in microglia-4 even 6 months after contusional TBI, contrasting all other cell types that mostly returned to naive levels. Regional and sexual dimorphism were noted. CEREBRI, our searchable atlas (https://shiny.crc.pitt.edu/cerebri/), identifies previously unrecognized cell subtypes/molecular targets and is a leverageable platform for future efforts in TBI and other diseases with overlapping pathophysiology.

Keywords: ependymal cells; heterogeneity; microglia; single-cell transcriptomic atlas; traumatic brain injury.

Figure 1:

Single-cells isolated with mRNA sequenced from murine cortex post-TBI. (A) Workflow of injury models/conditions and tissue sampling from naïve and injured brain (Supplemental-Table-1). Tissue was isolated from naïve or left pericontusional brain 24h post-injury (rCHI, CCI, CCI+HS), and 7d and 6mo post-CCI (Allen-Mouse Brain bregma values −0.755 to −2.75). Tissue was also obtained from females and contralateral mirror-image brain-tissue 24h post-CCI. Tissue was micro-dissected, single-cells were isolated, sequenced and computationally analyzed. (B) UMAP clustering of 23 cell populations (334,376 cells). (C) – (F) UMAPs in: (C) naïve/injury models 24h post-rCHI, CCI, and CCI+HS; (D) sex (CCI); (E) contralateral-region (CCI), and (F) time (CCI, 7d, 6mo). (Supplemental-Table-1). G) Gene-expression violin plot of umbrella cell-type markers (i) and immunofluorescence of an exemplar ependymal marker (Enpp2) across models (ii). (Supplemental-Figure-S1 for microglial/ependymal subtypes). H) Immunohistochemistry-based characterization of histological model severity with representative images of IBA-1+ (microglia) and Ly6g+ (neutrophil) infiltration across models (n=14). 24h and 72h are demonstrated per model across several brain-regions (row1=IBA1+ CA-1 hippocampus, row2=IBA1+ thalamus, row-3=Ly6g+ pericontusion, scale bar=20 μm). UMAP feature plots of Ly6g+ cells (neutrophils-1, Supplemental-Figure-S2) (I) Box plot of select cell-populations in (i) TBI-models (naïve=black, rCHI=gray, CCI=red, CCI+HS=maroon), (ii) sex (female=green, male=red), (iii) space/brain-region (contralateral cortex=blue, ipsilateral cortex=red), and (iv) stacked bar chart showing temporal changes across models. (Supplemental-Figure-S5).

Figure 2:

Gene-expression changes across TBI-models agnostic to cell-type. (A) –(C) Venn diagrams showing overlapping DEG counts in any cell-type across TBI-models (p-value_adjusted<0.05, log₂FC ≥0.5, naïve=reference) including (A) upregulated DEGs (B) downregulated DEGs (C) upregulated DEGs at three timepoints post-CCI. DEGs overlapping with all models are listed. (D) Ordinal heatmap of upregulated DEGs (y-axis) by TBI-condition (x-axis) demonstrating number of cell-types where DEGs are upregulated (p-value_adjusted<0.05, log₂FC>1). Most DEGs are upregulated in a single cell-type (blue). (E) Ordinal heatmap of downregulated DEGs (y-axis) by TBI-condition (x-axis) demonstrating number of cell-types where DEGs are downregulated (p-value_adjusted<0.05, log₂FC>1). Most DEGs are downregulated in a single cell-type (blue). (F) Upset plot showing number of overlapping DEGs shared by different cell-types after CCI (naïve=reference). Black=two cell-types, Green=three cell-types, Blue= four cell-types. (G) – (H) Functional enrichment of upregulated DEGs in (D) using g:Profiler highlights select gene ontology augmented in rCHI versus naïve (G), CCI versus naïve (H).

Figure 3:

Cell-Specific DEGs, pathways and pseudotime in TBI-models (A) Heatmap of DEGs (naïve=reference, log₂FC≥0.5, p-value_adjusted<0.05 ) in (i) rCHI (ii) CCI-microglia, (iii) CCI-neuro-glial-vascular and ependymal cells. Select DEGs annotated. (Supplemental-Figure-S6). (iv) Dual-labelled immunofluorescence of select markers differentially expressed in specific cell-types post-CCI (bottom-panel) versus naïve (top-panel) including CCL3, IL-1β, and SPP1 (n=14) co-stained with established cell-type markers. Minimal-no expression in naïve microglia (IBA1+), endothelial-cells (Cd31+) or astrocytes (GFAP+). Post-CCI, consistent with gene expression changes, immunofluorescent CCL3 is increased in a subpopulation of IBA1+ microglia (with altered morphology), IL-1β is increased in a subpopulation of endothelial cells, SPP1 is increased in astrocytes. White arrowheads=increased expression. (Supplemental-Figure-S6). (B) Chord diagrams highlighting key biological processes and associated genes altered in TBI-models. Naïve=reference. (B-i) Microglia-7 post-rCHI. (B-ii) Microglia-4 24h post-CCI. (Supplemental-Figure-S6). (C) UMAP depicting four microglial pseudotime-lineages. Key microglial-subtypes altered post-rCHI and -CCI stem from microglial lineage-1. Microglia-5= root of all lineages. (D) Pseudotime heatmap showing gene-expression dynamics across microglial lineage-1. Annotation bars above the heatmap demonstrate grouping by model and cell-type. Microglia at extremes of pseudotime are predominantly in CCI±HS with different processes upregulated versus rCHI and Naïve. Microglia-5 is the earliest in pseudotime. Genes are grouped by expression profiles with hierarchical clustering. Gene transcripts (right) with select gene-ontology biological processes and molecular functions (left) for each gene-cluster. (E) Individual expression profiles of top-genes altered along pseudotime based on (i) pseudotime-differentiation state and (ii) microglial cell-type. Each dot=single cell. Solid line=loess regression. Pseudotime (differentiation state) increases from left to right (x-axis). Y-axis=expression levels. (F) Heatmap of IL-1 cell-communication post-CCI identifying relative importance of cell-populations as senders, receivers, mediators, and influencers of IL-1 signaling highlighting central role of ependymal-4 (Supplemental-Figure-S6). (G) Circle plot of IL-1 signaling showing cell-populations that communicate with ependymal-4 post-CCI. (H) Pseudotime UMAP identifying ependymal-4 as lineage root—early and undifferentiated.

Figure 4:

Temporal cell-specific transcriptomic changes post-CCI. (A) Dotplot of DEGs (y-axis) by cell-type (x-axis) 7d post-CCI (naïve=reference, log₂FC≥1.25, p-value_adjusted<0.05) Red=upregulated, blue=downregulated. Dot size =% of cells. Microglia-4 has the most DEGs (Figure-1A) (B) Dotplot of DEGs (y-axis) by cell-type (x-axis) 6mo post-CCI (naïve=reference, log₂FC≥1.25, p-value_adjusted<0.05) Red=upregulated, blue=downregulated. Dot size =% of cells. Microglia-4 has persistent differential expression of multiple genes (Figure-1A, Figure-4A). (C) Heatmap of DEGs (y-axis, naïve=reference) in microglia-4 at 24h, 7d and 6mo post-CCI (x-axis=time, log₂FC≥1, p-value_adjusted<0.05). (D) Chord diagrams highlighting key biological processes and associated genes altered in Microglia-4 at (i) 7d (ii) 6mo post-CCI (Figure 3B).

Figure 5:

Sex-based transcriptomic differences post-CCI. (A) Heatmap of significant DEGs in female CCI (male=reference) across neuro-glial-vascular and ependymal cells (left), peripheral immune cells (top-right) and microglia (bottom-right). Log₂FC threshold≥0.5, p-value_adjusted<0.05. (B-C): Chord diagrams of gene set enrichment analyses highlighting key biological processes and associated genes altered in microglia-4 (B) and neurons (C) in females post-CCI. (Reference=male CCI). (D) Bar graph (top) showing differences in number and strength of cell-signaling pathways in females versus males post-CCI. Cell-communication signaling pathways ranked based on differences in relative information flow (bottom) within the specified network (y-axis) comparing females (teal) with males (red). (E) Heatmap comparing differential number (left) and strength of interactions (right) between specific cell-populations in females versus males. Top bar in both heatmaps=column sum. Right bar=row sum. Red=increased female signaling, Blue=decreased female signaling. (F) Heatmap showing strength of incoming and outgoing signaling in males (left) and females (right) demonstrating differences based on cell-type and signaling-pathway. Rows=signaling pathways. Columns=cell populations. Darker red=greater signaling strength. Top bar in both heatmaps=column sum. Right bar=row sum. (G) Scatter plots of individual cell-types showing changes in signaling pathways based on sex and incoming/outgoing signals. Red=male-specific, Teal=female specific, Black=shared by both sexes, Squares=Incoming, Triangles=outgoing, Diamonds= incoming and outgoing.

Figure 6:

Brain-region based transcriptomic differences post-CCI. (B) Heatmap of significant DEGs in contralateral brain-regions post-CCI (ipsilateral brain-region=reference) across microglia (left), peripheral immune-cells (center) and neuro-glial-vascular and ependymal cells (right). Log₂FC threshold≥0.5, p-value_adjusted<0.05. (B-E): Chord diagrams of gene set enrichment analyses highlighting key biological processes and associated genes altered in astrocytes (B), NK-cells (C), T-memory (D) and endothelial-1 (E) contralaterally post-CCI. Reference=ipsilateral brain-region. (F) Significant cell-communication signaling pathways ranked based on differences in the relative information flow (top-graph) within the specified network (y-axis) comparing contralateral (teal) ipsilateral/pericontusional (red) cortex post-CCI. Bar graph (bottom) showing differences in number and strength of cell-signaling pathways in the pericontusion versus contralaterally post-CCI. (G) Heatmap comparing differential number of interactions (left) and strength of interactions (right) between cell populations contralaterally versus pericontusion. Top bar in both heatmaps=column sum. Right bar=row sum. Red=increased signaling contralaterally. Blue=decreased signaling contralaterally. (H) Heatmap showing strength of incoming and outgoing signaling in pericontusion (left) and contralateral cortex (right) demonstrating differences based on cell-type and signaling pathway. Rows=signaling pathways. Columns=cell populations. Darker red= greater signaling strength. Top-bar in both heatmaps=column sum. Right-bar=row sum. (I) Scatter plots of select cell-types showing changes in signaling pathways based on brain-region and incoming/outgoing signals. Red=pericontusion. Teal=mirror-image contralateral cortex. Black=shared by both regions. Squares=incoming. Triangles=outgoing. Diamond=incoming and outgoing.