Peripheral T cells infiltrate the aged SVZ and remodel the brain microenvironment
Brain aging has been reported to be accompanied by chronic inflammation. To investigate the molecular and cellular alterations of the brain immune microenvironment during normal aging, we explored the large scale transcriptomic dataset (PRJNA450425) of single cells derived from three young (3 months old) and three aged (28–29 months old) mouse SVZs (Supplementary Fig. 1a). On the basis of quality control, 13,760 cells were used for unsupervised clustering analysis (Supplementary Fig. 1b), and 2800 CD45+ cells were extracted for downstream analysis (Supplementary Fig. 1c). Sub-clustering analysis revealed microglia, T cells, and CNS border-associated macrophages were the major immune cells in the SVZ (Fig. 1a, Supplementary Fig. 1d). Notably, the T cells were almost exclusively derived from aged mice, and the proportion of T cells was significantly larger in aged SVZ (Fig. 1b). We further identified the T cells as CD3+CD8+CD4− by exploring the expression of marker genes, including Cd3e, Runx, Cd4, and Cd8a (Supplementary Fig. 1e). Immunofluorescence staining of CD31 and CD8a confirmed the obvious age-related T cells infiltration in the SVZ in both male and female mice (Fig. 1c-d, Supplementary Fig. 1f).
The infiltrated T cells highly expressed genes related to T cell activation (Cd69, Itk), cytokine release (Ccl5, Xcl1), and cytotoxicity (Gzmb, Gzmk) (Fig. 1e). We further performed functional enrichment analysis to clarify the role of the infiltrated T cells. A total of 335 GO terms were enriched (z-score ≥ 2, adjusted P-value < 0.05) (Supplementary Table 1). Top functional terms were T cell receptor (TCR) signaling, T cell activation, and immunoregulatory interactions between a lymphoid and a non-lymphoid cell (Supplementary Fig. 1 g). These results suggested an activated phenotype of the infiltrated T cells, which might have impact on the brain microenvironment. We further applied the InterCellDB toolkit to determine the effect of T cells on resident cells in the brain. First, we focused on the genes related to the Reactome term immunoregulatory interactions between a lymphoid and a non-lymphoid cell (R-HSA-198933). A widespread effect of the infiltrated T cells was identified. Microglia (432gene pairs), macrophages (426 gene pairs) and endothelial cells (175 gene pairs) were the main resident cells in the SVZ in communications with the infiltrated T cells (Fig. 1f, Supplementary Table 2). Second, we performed intercellular network analysis based on the cytokines released by T cells. This analysis identified Ccl4, Ccl5, Ifng, Xcl1, and Fasl as key cytokine-coding genes affecting the brain microenvironment since they showed effects on almost all resident cells in the SVZ (Supplementary Fig. 1 h, Supplementary Table 3).
Aged microglia release chemokines to recruit circulating CD8+ memory T cells
Peripheral T cells undergo tremendous changes with age [28]. In order to determine whether these changes contribute to the T cell infiltration in the SVZ, we analyzed another published single-cell transcriptomic dataset (GSE132901) of splenic T cells [29] (Supplementary Fig. 2a). A total of 30,168 cells from four young (29–34-wk-old) and three aged (88–93-wk-old) mice were included in unsupervised clustering analysis, and a CD8+ T cell cluster containing 3318 cells was extracted for sub-clustering and further downstream analysis (Supplementary Fig. 2b). A volcano plot displayed all receptor genes whose expression changed significantly with age in CD8+ T cells. Three chemokine receptor genes, Cxcr3, Ccr2, and Ccr5, were significantly upregulated in aged CD8+ T cells, which were possibly associated with the recruitment of CD8+ T cells (Fig. 2a). We noted that Cxcr3 was robustly upregulated in both CD4+ and CD8+ T cells, while Ccr2 and Ccr5 were upregulated specifically in CD8+ T cells (Fig. 2b), which indicated that Ccr2 and Ccr5 may account for the selective CD8+ T cell infiltration in the aged SVZ. Interestingly, Ccr2 and Ccr5 showed a remarkable heterogeneity in gene expression in CD8+ T cells in both young and aged mice (Fig. 2c).
Unsupervised sub-clustering analysis was applied to explore the heterogeneity of CD8+ T cells. The analysis of 3318 CD8+ T cells with UMAP revealed four distinct cell types (Supplementary Fig. 2b). Characterization of markers in these clusters (Supplementary Fig. 2c) identified them as naive T cells, central memory T cells (Tcm), effector memory T cells (Tem), and interferon response T cells (Supplementary Fig. 2b). The Ccr2 and Ccr5 genes were exclusively expressed in memory T cells (Fig. 2c), indicating their propensity to be attracted to the brain. Flow cytometry experiments confirmed the remarkable increase of CCR2+ CD8 memory T cells in both male and female mice (Fig. 2d, Supplementary Fig. 2d-e).
We then screened the expression levels of the genes encoding their corresponding ligands in the mouse brain. Ccl3 and Ccl4 were expressed specifically in microglia, and Ccl3 was significantly upregulated in aged microglia (Fig. 2e). In order to clarify whether CCL3 plays a role in trafficking of CD8+ T cells to the aged brain, we downloaded the Aging Plasma Proteome data from a previous study [30]. This dataset contains data on 2925 plasma proteins from 4263 subjects ranging from young adults to nonagenarians. As expected, the plasma CCL3 level increased significantly during normal aging (Fig. 2f). Immunofluorescence staining was performed to verify the expression of CCL3 and CCL4 in microglia. We observed strongly increased staining of CCL3 and CCL4 in old microglia compared with their younger counterparts (Fig. 2g, Supplementary Fig. 2f-g).
Various age-related transcriptomic changes of brain endothelial cells from different arteriovenous zones
BECs serve as relay stations between the circulation and the brain. Immunofluorescence staining showed CD31 and CCL3 were partially co-localized in the aged SVZ (Supplementary Fig. 3a), indicating that aged BECs may also contribute to the T cell recruitment. Then, we extracted and re-clustered the BECs of the SVZ for further downstream analysis (Supplementary Fig. 3b–c). Unsupervised sub-clustering analysis revealed seven distinct clusters (clusters E0–E6, Supplementary Fig. 3c). Notably, the trajectory constructed by Monocle pseudotime analysis was strongly associated with the arteriovenous axis. Known arterial (Fbln5) and venous (Nr2f2) markers peaked at opposite ends of the trajectory and the capillary (Rgcc) markers summited in the middle of the trajectory (Supplementary Fig. 3d). Then, we visualized each cluster of BECs on the trajectory (Supplementary Fig. 3e), and redefined them as large artery (E6), artery (E3), arterial-capillary (E2), capillary (E0, E1), vein (E4), and interferon-EC (E5) based on their transcriptomic profiles (Supplementary Fig. 3e, Fig. 3a). In addition, we identified 154 artery-specific markers, 22 capillary-specific markers, and 57 vein-specific markers (Supplementary Fig. 3f; Supplementary Table 4).
The cellular composition was similar across young and aged BECs (Fig. 3a). Nonetheless, we observed generally downregulation of tight junction component-encoding genes across the vascular network (Fig. 3b). Previous studies have reported that aged BECs downregulate the expression of Cldn5 (encoding claudin 5) and Ocln (encoding Occludin) [31, 32]. We observed a significant downregulation of Cldn5 and Ocln in capillary BECs, but not in arterial or venous BECs (Fig. 3b, Supplementary Fig. 3 g). In addition, we found obvious extravasation of IgG in the aged mouse brain (Fig. 3c-d). Together, these data indicated that tight junctions were dysfunctional in brain microvascular endothelial cells (BMECs).
To investigate how aging may commonly and differentially affect gene expression of BECs from different vascular beds, GSEA analysis was performed to assay aging-related pathways in each BEC subtype (Fig. 3e). Consistent with recent reports [14, 31, 33], we found an upregulation of immune/cytokine signaling (adaptive immune response, interferon signaling, and antigen presentation) and ribosome biogenesis for all vascular segments (Fig. 3e). In addition, a significant downregulation of Wnt–β-catenin signaling was observed in the arterial segment, which had been identified as a key regulator of BBB maintenance [34]. Aged capillaries downregulated “ATP metabolic process”, implicating dysfunctional energy metabolism in aged BMECs. Notably, upregulation of “T cell migration-related process” was specifically observed in aged venous BECs (Fig. 3e), which implicated the vascular bed dependence of T cell infiltration in the aged SVZ.
Venous BECs upregulate adhesion molecules and promote T cell infiltration in the aged SVZ
To further clarify the mechanism by which T cells infiltrate the aged SVZ and the exact location, we deeply examined all biological processes associated with leukocyte migration among different vascular segments (Fig. 4a). This functional cluster was predicted to be strongly activated in venous BECs in the aged SVZ. Subsequently, we extracted all upregulated DEGs involved in lymphocyte migration and depicted a dot plot to display the relationship between these genes and the vascular bed (Fig. 4b). Numerous key genes were significantly upregulated in aged venous BECs, including Vcam1, Podxl, Icam1, Ch25h, Cd200, and Apod (Fig. 4b). Notably, two genes encoding critical adhesion molecules of endothelial cells, Vcam1 and Icam1, were specifically upregulated in venous BECs (Fig. 4b). These two molecules play an important role in mediating the firm adhesion of leukocytes to endothelial cells. Then, we applied the InterCellDB toolkit to explore cell–cell interactions between T cells and BECs (Fig. 4c). Strong interactions between venous BECs and Tcm were observed (Fig. 4c). We further explored gene pairs that mediate cell migration between BECs and Tcm. The results showed that the Vcam1–Itgb2, Vcam1–Itgb1 and Icam1–Itgb2 pairs were relatively specific for Vein–Tcm communication (Fig. 4d). Notably, Itgb1, Itgb2, and Ccr2 were selectively expressed in aged memory T cells (Fig. 4e–f, Fig. 2c). Finally, expression of VCAM1 and ICAM1 in cells immunoreactive for CD31, a marker of brain endothelial cells, was significantly upregulated in the aged SVZ (Fig. 4g–h).
Aged microglia promote expression of adhesion molecules on venous BECs
To further investigate the cause of transcriptome changes in venous BECs, we analyzed cell–cell interactions between brain resident cells and venous BECs (Fig. 5a). There is a strong impact of aged microglia on the function of “leukocyte migration” in venous BECs (Fig. 5a). Thus, we compared the transcriptomic landscape of microglia between young and aged mice (Supplementary Fig. 4a–b). Five distinct clusters of microglia were identified in both young and aged mice (C0–C5, Fig. 5b). Interestingly, microglia from aged mice were most enriched in clusters 1, 2 and 4, whereas those from young mice were more frequently found in clusters 0 and 3 (Fig. 5b). There were strong communications between aged microglia (C1 and C2) and venous BECs (Fig. 5c). Thus, we performed differential expression analysis between cluster 2 and cluster 0 to determine the transcriptomic changes induced by aging in homeostatic microglia (Supplementary Fig. 4c). We identified 405 upregulated DEGs, the functional implications of which were further explored by GO enrichment analysis. Five functional clusters were significantly overrepresented in aged microglia, among which the largest was associated with the inflammatory response and cytokine production (Supplementary Fig. 4d). We further annotated the relationship between specific genes and biological processes in the inflammatory response and cytokine production functional cluster (Supplementary Fig. 4e). Numerous key genes upregulated in aged microglia, including Apoe, Spp1, Cd74, Ccl3, Mif, Ccl4, and Tnf, were involved in at least two subcategories of the functional cluster (Supplementary Fig. 4e).
We further explored whether cytokines secreted by microglia could activate venous BECs in the aged SVZ. We applied the InterCellDB toolkit and found several gene pairs between aged microglia and venous BECs (Fig. 5d). Gene pairs including Tnf–Vcam1, Tnf–Icam1, and Spp1–Icam1 were specific with positive effects (Fig. 5e). TNF-α is a crucial pro-inflammatory cytokine with elevated expression during normal aging [35]. Then We performed double-label immunofluorescence staining and observed a significantly higher number of TNF-α+ microglia in both male and female aged mice (Fig. 5f, Supplementary Fig. 4f).
Together, these results indicated that aged microglia might shift towards a pro-inflammation and chemotactic state and recruit peripheral T cells into aged brain. To confirm these observations obtained by transcriptomic analysis, we established a coculture system to further dissect the impact of aged microglia on CD8+ T cells. CD8+ T cells were isolated from mouse spleens using magnetic-activated cell sorting and then cocultured with brain slices in the lower compartment for 24 h. Cells in the lower compartment were then collected for flow cytometry analysis (Supplementary Fig. 4 g). The results revealed a significantly higher number of CD8+ T cells entering the lower compartment when co-cultured with aged brain slices compared to co-cultured with young brain slices (Fig. 5g). However, depleting microglia with PLX5622 or blocking CCL3 could partially reduce the number of CD8+ T cells recruited by aged brain slice (Fig. 5g). These results support a critical contribution of aged microglia to CD8+ T cells recruitment in aged mouse brain.
CSF from the elderly promotes the transition of human microglia to a chemotactic phenotype in vitro
Next, we investigated the similarity in expression profiles and associated biological properties between human and mouse primary microglia during normal aging. To obtain primary human microglia, we collected radiologically healthy CNS specimens from patients undergoing brain surgery for removal of epileptic foci. Then the tissues were dissociated and sorted using magnetic-activated cell sorting for cells positive for CD11b, which were then transferred into 6-well plates. After exposed to each individual CSF sample or PBS for 24 h, the cells were collected for RNA sequencing (Fig. 6a). The sorted human microglia expressed known microglial genes such as AIF1, TMEM119 (Fig. 6b). In contrast, the well-established markers of neurons (MAP 2, NCAM1), astrocytes (GFAP, ALDH1L1), oligodendrocytes (APC, PLP1), endothelial cells (CLDN5, ITM2A), T cells (CD3E), and B cells (CD79A) were all expressed at low levels (Fig. 6b). Next, to determine the transcriptomic changes of microglia induced by the CSF from the elderly, we applied the R package DEseq2 to calculate the DEGs, which contains 329 upregulated genes and 501 downregulated genes (Fig. 6c, Supplementary Table 5). To elucidate the functional alterations in the microglia treated with CSF from the elderly, GO enrichment analysis was performed on the upregulated DEGs by an online tool Metascape. Immune inflammatory response, cytokine production, cell activation and chemotaxis accounted for the majority of enriched GO terms (Fig. 6d), which showed remarkable similarity with the results of aging mouse microglia (Supplementary Fig. 4d). Since we identified chemotaxis as an important functional alternation of aging mouse microglia, we further explored the pathways related to leukocyte migration and chemotaxis. There were many significantly upregulated chemokines including CCL3 and CCL4 which were involved in three or more GO functional terms (Fig. 6e), indicating their critical roles in recruiting peripheral immune cells.