Brain Endothelial Cells Are Exquisite Sensors of Age-Related Circulatory Cues

Abstract

Brain endothelial cells (BECs) are key constituents of the blood-brain barrier (BBB), protecting the brain from pathogens and restricting access of circulatory factors. Yet, because circulatory proteins have prominent age-related effects on adult neurogenesis, neuroinflammation, and cognitive function in mice, we wondered whether BECs receive and potentially relay signals between the blood and brain. Using single-cell RNA sequencing of hippocampal BECs, we discover that capillary BECs-compared with arterial and venous BECs-undergo the greatest transcriptional changes in normal aging, upregulating innate immunity and oxidative stress response pathways. Short-term infusions of aged plasma into young mice recapitulate key aspects of this aging transcriptome, and remarkably, infusions of young plasma into aged mice exert rejuvenation effects on the capillary transcriptome. Together, these findings suggest that the transcriptional age of BECs is exquisitely sensitive to age-related circulatory cues and pinpoint the BBB itself as a promising therapeutic target to treat brain disease.

Keywords: aging; blood-brain barrier; brain endothelial cells; plasma proteome; rejuvenation; single-cell RNA sequencing.

Figure 1.. BECs Segment into Arterial, Capillary, and Venous Identities

(A) Experimental design for single-cell analyses of BEC transcriptomes. (B) (Top panel) tSNE of a subset (179) of young BECs collected in an unbiased manner (CD31⁺) overlaid with the expression pattern (log10CPM) of key segment identity genes Bmx (arterial), Slc16a1 (capillary), Nr2f2 (venous), and Vcam1 (arterial and venous). Note the low number of Vcam1⁺ cells and the absence of a clear venous population. (Bottom panel) Addition of a subset of VCAM1⁺ cells into principal-component analysis (PCA) (62, as collected by FACs enrichment) significantly improves A-C-V identification. (C) Identification of arterial, venous, and capillary populations after the addition of Vcam1⁺ cells. Pie charts of the proportion of A-C-V cells in unbiased CD31 sorts, in VCAM1⁺CD31⁺ sorts, and unbiased sorts with an added 25% of cells from VCAM1⁺ cell enrichment (final condition). (D) Heatmap of the top 25 most enriched genes per A-C-V population (which were identified by unbiased whole transcriptome clustering) in young BECs. (E) Expression bar plots of top genes that could serve as novel markers for arterial and venous identities. Genes encoding for cell surface receptors are indicated by *R. Expression levels in A-C-V segments are consistent across 981 young BECs (4 biological replicates). (F) Representative in situ RNA hybridization images confirming Cdh13 expression and enrichment in Vwf⁺ Acta2⁺ arterial cells (representative of 4 mice). Scale bar, 20 μm. (G) Representative in situ RNA hybridization images confirming Ilr1r1 expression and enrichment in Vwf⁺ Acta2⁻ venous cells (representative of 4 mice). Scale bar, 20 μm.

Figure 2.. Systemic LPS Exposure Activates Common Transcriptional Programs across Segment Identities

(A) Venn diagram showing the number of DEGs (FDR < 0.05) shared between each vessel segment. Heatmap showing the distribution of up- and downregulated genes per vessel segment. (B) Dotted heatmap of top 60 DEGs ranked by the signed-FDR value (product of log₂FC*⁻log₁₀(FDR)). Color indicates the average log₂FCof LPS/untreated, while the dot size represents degree of statistical significance. Only genes with FDR < 0.05 for at least one vessel segment is listed, and hierarchical clustering is performed (dot size = 0 indicates FDR > 0.05). (C) GO enrichment analysis of DEGs (FDR < 0.05) up- and downregulated in LPS treated A, C, and V cells. Left-hand side (red) indicates pathways over-represented in DEGs upregulated with LPS, right-hand side (blue) likewise in DEGs downregulated with LPS exposure. Exemplary genes contributing to pathway enrichments are listed on the side. (D) Representative images and quantification of MFSD2A expression (green) in the hippocampus of PBS- and LPS-exposed young (3 months) mice (n = 5, two-sided t test; mean ± SEM). Scale bar, 40 μm.

Figure 3.. Normal Aging Induces Transcriptomic Changes Specific to Each Segment Identity

(A) t-Distributed stochastic neighbor embedding (tSNE) of normally aged and young BECs after canonical correlation analysis (CCA), using the top-9 correlation components. Aged and young cells show comparable distributions of A, C, and V identities along the A-C-V axis. Note that segmental identity largely drives cluster formation, rather than age. (B) Distribution of key A-C-V segment identity markers in tSNE-space. (C) Venn diagram showing the numbers of DEGs (FDR < 0.1) shared between different vessel segments. Heatmap of the union of all DEGs up- and downregulated in aged A, C, or V cells, illustrating the degree of overlap of DEGs between each segment. (D) Dotted heatmap of the top 80 DEGs ranked by the signed-FDR value (product of log₂FC*⁻log₁₀(FDR)). Color indicates the average log₂FCof aged/young, while the dot size represents degree of statistical significance. DEGs with FDR < 0.1 for at least one vessel segment are listed and ordered by hierarchically clustering (dot size = 0 indicates FDR > 0.1). (E) GO analysis of DEGs (FDR < 0.1) up- and downregulated in aged A, C, and V cells. Left-hand side (red) indicates pathways over-represented in DEGs upregulated with aging, right-hand side (blue) likewise in DEGs downregulated with aging. Exemplary genes contributing to pathway enrichments are listed on the side. (F) Representative images and quantification of B2m and Itm2a expression in young and aged hippocampal capillaries (Slc16a1⁺Pecam1⁺) following RNA in situ hybridization (B2m: n = 205 cells (young), 295 cells (aged); Itm2a: n = 429 cells (young), 356 cells (aged), two-sided t test; mean ± SEM). Scale bar, 5 μm. (G) Representative images and quantification of VWF and ALPL expression (green) in CD31⁺ or collagen IV⁺ vasculature (red) in the hippocampus of young (3 months) and aged (20 months) mice (n = 6, two-sided t test; mean ± SEM). Scale bar, 40 μm.

Figure 4.. Brain Capillaries Sense and Adopt Cues in the Circulatory Milieu, with Aged Plasma Exposure Recapitulating the Normal Aging Transcriptomic Signature

(A) Schematic of the aged mouse plasma (AMP) acute infusion paradigm in young mice. (B) tSNE of AMP- and PBS-treated BECs from young (3 months) mice after CCA, using the top-9 correlation components. AMP- and PBS-treated cells show comparable distributions of A, C, and V identities, suggesting plasma infusions do not significantly alter native segmental identities. (C) Expression bar plots of canonical A-C-V marker genes in AMP- and PBS-treated BECs, with segmental identities defined by unbiased clustering in (B). (D) Volcano plot depicting up- and downregulated genes with AMP treatment in capillaries (compared to PBS control). Genes marked in red are significant (FDR < 0.1). FDR values are calculated only with genes showing an average log₂FC > 0.1. Genes labeled red are FDR <0.1. (E) Dotted heatmap of top 60 DEGs ranked by the signed-FDR value (product of log₂FC*⁻log₁₀(FDR)). Color indicates the average log₂FCof AMP/PBS, while the dot size represents degree of statistical significance. DEGs with FDR < 0.1 for at least one vessel segment are listed and ordered by hierarchically clustering (dot size = 0 indicates FDR > 0.1). (F) Scatterplot of genes and their log₂FCin both aged/young and AMP/PBS treatment in capillaries. The 153 genes commonly up- (blue) or downregulated (green) in both groups (FDR <0.1 in both) are labeled. These DEGs may be modulated by plasma factors upregulated in normal aging. Inset shows the same genes (red) on a scatterplot of the signed-FDR value (product of log₂FC*⁻log10(FDR)) for normal aging and AMP. (G) Top pathways over-represented by DEGs upregulated in both normal aging and with AMP treatment (149 intersecting DEGs).

Figure 5.. Young Plasma Exposure Partially Reverses the Transcriptomic Signature of Normal Brain Capillary Aging

(A) Schematic of the young mouse plasma (YMP) acute infusion paradigm in aged mice. (B) tSNE of YMP-treated and PBS-treated BECs from aged (20 months) mice after CCA, using the top-9 correlation components. AMP- and PBS-treated cells show comparable distributions of A, C, and V identities, suggesting plasma infusions do not significantly alter native segmental identities. (C) Expression bar plots of canonical A-C-V marker genes in YMP- and PBS-treated BECs, with segmental identities defined by unbiased clustering in (B). (D) Volcano plot depicting up- and downregulated genes with YMP treatment in capillaries (compared to PBS control). Genes marked in red are significant (FDR < 0.1). FDR values are calculated only with genes showing an average log₂FC > 0.1. Genes labeled red are FDR < 0.1. (E) Scatterplot of genes and their log₂FCin both aged/young and YMP/PBS treatment in capillaries. The 89 genes that are upregulated with age (aged/young) but decreased with YMP (YMP/PBS)—and vice-versa (FDR < 0.1 in aging and FDR < 0.1 in YMP)—are labeled. These age-upregulated DEGs are likely modulated and/or reversed via exposure to YMP. Inset shows the same genes (red) on a scatterplot of the signed-FDR value (product of log₂FC*⁻log₁₀(FDR)) for normal aging and YMP. (F) Top pathways over-represented by DEGs upregulated in normal aging and downregulated with YMP treatment (89 intersecting DEGs).

Figure 6.. The Circulatory Environment Induces Cell Non-autonomous Brain Endothelial Aging

(A) Venn diagram depicting the number of DEGs shared between each treatment condition (aged/young, AMP/PBS, and YMP/PBS). 42 genes are differentially expressed in all three treatment groups (i.e., increasing with normal aging and AMP but decreasing with YMP). (B) Bar plot of top biological pathways over-represented in DEGs shared between aging and AMP (149 DEGs) or shared between aging and YMP (89 DEGs). Score is derived from GeneAnalytics software. The presence of common pathways suggests YMP may reverse some transcriptional consequences of AMP treatment and normal aging. Select genes in each pathway are listed, with DEGs shared across all three treatments labeled in red. (C) Bar plot of the log₂FC of top DEGs that intersect in all three treatments (aged/young, AMP/PBS, and YMP/PBS). These genes are likely modulated by plasma factors upregulated in a normal aged milieu, with expression changes reversible via exposure to young plasma. (D) Sankey plot depicting relationships between DEGs encoding BEC surface receptor or membrane proteins, and their corresponding plasma ligands. Directionality of expression changes in each condition (normal aging, AMP, and YMP) are denoted with arrows. Corresponding ligands found significantly up- (red) or downregulated (blue) with age in mouse plasma (SOMALogic) are highlighted. (E) Representative images and quantification of Sod1 and Ifnar1 expression in AMP-exposed hippocampal capillaries (Slc16a1⁺Pecam1⁺) following RNA in situ hybridization (Sod1: n = 595 cells [PBS], 576 cells [AMP]; Ifnar1: n = 746 cells [PBS], 916 cells [AMP]; two-sided t test; mean ± SEM). Scale bar, 5 μm.