Background
Brain aging is accompanied by T cell accumulation [
1‐
3]. The infiltrated T cells and their secreted cytokines lead to a loss of functional neural stem cells in the subventricular zone (SVZ) [
4,
5]. They also attenuate neurogenesis and neuroplasticity of aging brain, manifested by progressive cognitive decline [
6,
7]. However, the mechanisms underlying T cell infiltration in the aging SVZ remain elusive, which is fundamentally important for preventing brain aging.
The blood–brain barrier (BBB), which is composed of and regulated by endothelial cells, the basement membrane, pericytes, the glia limitans, and microglia, serves as a relay station between the circulation and the central nervous system (CNS) [
8]. Leukocyte migration across the BBB is a complex process, which is triggered by inflammation and chemotactic signals released from the CNS [
9,
10]. Once the signals are received, the T cells that express corresponding receptors are arrested by brain endothelial cells (BECs) and then cross the BBB [
11,
12]. Age-dependent changes in this process have partially been documented. BBB breakdown has been proven as a consistent feature in aging humans and rodents [
13] and manifests as a loss of tight junction integrity and altered transport properties [
14‐
16]. Accumulating evidence supports that the aging BECs show a zonation-dependent, rather than a consistent, change across the vascular bed [
14]. However, it is still unclear which part of the aging BECs allow peripheral immune cells to infiltrate the brain. Besides, the mechanisms underlying the BEC changes in the aging brain remain elusive. Our previous study indicated that a unique type of highly-activated microglia could evoke brain inflammation in aged mice [
17]. We therefore hypothesized that these senescent microglia were responsible for T cells accumulation, and the inflammatory factors they released might lead to BECs activation, thereby promoting T cells infiltration.
In the present study, we established interaction networks among BECs, microglia, and T cells by analyzing single-cell transcriptional profiles of cells from aged and young mice. We identified a chronic inflammation phenotype with T cell infiltration in the aged SVZ. The circulating T cells were recruited by aged microglia and entered the brain though anchoring adhesion molecules (VCAM1 and ICAM1) on venous BECs. Our findings provide a possible cause for age-related brain inflammation and may help identify potential therapeutic targets.
Materials and methods
Animals
Young (8–10 weeks old) C57BL/6 males and females were obtained from SLAC Laboratory Animal Company Limited (Shanghai, China). 20 aged (18 months old) male C57BL/6 mice and 10 aged female (18 months old) were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China). The mice were housed in plastic cages with controlled temperature and humidity and a 12/12-h light/dark cycle. All animal experiment protocols were approved by the Institutional Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine and were in compliance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. For microglia depletion, PLX5622 was supplied to mice in the diet (Research Diets) at 1200 PPM (1200 mg/kg of chow) for consecutive 7 days.
Immunostaining of brain sections and image analysis
Mice were deeply anesthetized and perfused transcardially with 25 mL of ice-cold phosphate-buffered saline (PBS), followed by 20 mL of 4% paraformaldehyde solution in PBS. Brains were postfixed in 4% paraformaldehyde for 24 h and dehydrated in serial 15 and 30% sucrose solutions at 4 °C. Then the brain samples were sectioned into coronal slices (25 μm thick). Brain sections were stored in cryoprotectant (40% PBS, 30% glycerol, 30% ethylene glycol) and kept at − 20 °C until immunostaining. Sections were washed twice with PBS, followed by permeabilization in 0.5% Triton X-100 at room temperature. Then sections were blocked with 5% normal donkey serum in PBS for 1 h at room temperature and incubated overnight at 4 °C with the following primary antibodies: anti-CD31 (Santa Cruz, sc18916, 1:50), anti-CD8a (Abcam, ab217344, 1:250), anti-IBA1 (Abcam, ab5076, 1:250), anti-CCL3 (Abcam, ab179638, 1:200), anti-CCL4 (Abcam, ab45690, 1:200), anti-mouse IgG-Alexa Fluor 488 (CST, 4408S, 1:200), anti-VCAM1 (Abcam, ab134047, 1:200), anti-ICAM1 (Abcam, ab109361, 1:100), and anti-TNF-α (Abcam, ab183218, 1:100). Then, the sections were incubated in the dark with donkey secondary antibody conjugated with Alexa Fluor 488, 555, or 594 (Invitrogen, 1:500) at room temperature for 1 h. After washing with PBS three times, the sections were mounted on glass slides with mount-G containing DAPI (Yeasen Biotech). Sections were observed with a Leica TCS SP8 confocal microscope (Leica Microsystems).
Images were adjusted for brightness and contrast using Fiji 2.1.0/1.53c. All confocal images were represented as maximum intensity projections. For cells quantification, three to five randomly selected microscopic regions were captured in each section, then the images were loaded into imageJ and positively stained cells were electronically labeled with the software to avoid duplicated counting. For quantification of IgG extravasation, the parenchymal staining of IgG was slected based on the pre-set threshold petameters. The integrated density of IgG fluorescence of the randomly chosen regions with same size were recorded.
Flow cytometry
Spleen cells were prepared as described previously [
18]. Briefly, the spleen was mechanically dissociated and passed through a 70um filter. The final 4 ml suspension was layered onto 2 ml Ficoll-Paque (GE,17–1440-02) and the cells at the interface were collected after certification (500 g,20 min,4 °C). Single cell samples were incubated with antibodies to surface antigens for 30 minutes on ice at 4 °C in the dark. Fluorochrome compensation was performed with single-stained UltraComp eBeads. Flow cytometery was performed on the BD LSRFortessa flow cytometer (BD biosciences). Data analysis were performed using Flowjo software.
The antibodies used for profiling splenic T cells included anti-mouse CD45-Pacific Blue (1:200; BioLegend, 103,126); anti-mouse CD3e-FITC (1:200; BD Pharmingen, 553,062); anti-mouse CD4-APC-Cy7 (1:200; BD Pharmingen, 552,051); anti-mouse CD8-Percp (1:200; BioLegend, 100,732); anti-mouse CD44-V500 (1:200; BD Pharmingen, 560,781); anti-mouse CD62L-BUV395 (1:200; BD Pharmingen, 740,218); anti-mouse CCR2-PE (150609).
In vitro CD8+ T cell cultures and cocultures with brain slice
Spleen was harvested from sham mice to prepare single cell suspensions as we described above. CD8
+ T cells were isolated using mouse CD8a microbeads (Miltenyi Biotec, 130–117-044) according to the manufacturer’s instructions. Coculture System was established as described before [
18], isolated CD8
+ T cells in a transwell insert were incubated with brain slices in the lower chamber in culture media (RPMI 1640, 10% FBS, 1% penicillin/Strepromycin, 1 mM pyruvate sodium, 55 μm β-mercaptoethanol with the presence of soluble anti-CD3, anti-CD28 and IL-2) for 24 h. Then cells in the lower compartment were collected and stained with anti-mouse CD3e-FITC (1:200; BD Pharmingen, 553,062), anti-mouse CD8a-APC (1:200; BD Pharmingen, 553,035) on ice at 4 °C in the dark. CD3
+ CD8
+ T cells in the lower compartment were counted using Precision count beads (BioLegend, 424,902).
Basic processing and clustering analysis of single-cell transcriptome data
Two single-cell RNA sequencing (scRNA-seq) datasets were downloaded from the Gene Expression Omnibus (GEO) database, including transcriptomic data of the SVZ neural stem cell niche in young and aged mice (PRJNA450425) and expression matrices of the spleen in young and aged mice (GSE132901). Basic processing and visualization of the scRNA-seq data were performed with the Seurat [
19] package (v3.2.2) in R (v 3.6.3). Briefly, low-quality cells and doublets were filtered out based on the following criteria: (i) number of expressed genes was less than 200 or more than 2500 and (ii) the percentage of mitochondrial genes was more than 10%. The data were normalized to the total expression and log-transformed. The variable genes were detected using the FindVariableFeatures function with default parameters. Linear scaling was then applied and the mitochondrial contamination was removed using the ScaleData function. The batch effect was removed using the IntegrateData function. Principal component analysis was carried out on the scaled data, and the top 20 principal components were stored. Then, clusters were identified using the FindClusters function. Non-linear dimensional reduction methods including uniform manifold approximation and projection (UMAP) and t-distributed stochastic neighbor embedding (t-SNE) were used to visualize clustering results.
Differentially expressed gene calculation
The Seurat FindAllMarkers and FindMarkers functions were employed to identify differentially expressed genes (DEGs) using the Wilcoxon rank sum test. Only genes with Bonferroni adjusted P-value < 0.05 and |log2(fold change) | > 0.1 were considered DEGs.
Gene set enrichment analysis
Gene set enrichment analysis (GSEA) was carried out with the GSEA toolkit (v4.0.3, Broad Institute) following the published protocol [
20]. Briefly, we generated a preranked gene list according to their log (fold change) and significance values. Five datasets (Hallmark pathways, Gene Ontology (GO) biological processes, KEGG, BioCarta, and Reactome) with in total 4495 gene sets were used as a reference. Then 1000 random permutations were performed to calculate the
P-values. Significantly enriched gene sets were defined as gene sets with false discovery rate q-value < 0.05. Then, we used Cytoscape software (v3.7.2) and the AutoAnnotate app (v1.3.2) to cluster and annotate the gene sets.
GO enrichment analysis
The online tool Metascape (
http://metascape.org) was used for GO enrichment analysis [
21]. All genes in the mouse genome were used as the enrichment background. Firstly, a set of DEGs was submitted. Metascape would return a set of statistically enriched terms. Then, the activation z-score of each term was calculated by the R package GOplot [
22]. We defined significantly changed terms as those fulfilling
P-value < 0.05 and an absolute z-score exceeding 2. Finally, the results were visualized with a chord chart or a dot plot using GOplot or prism software (v8.2.1, GraphPad Software Inc.).
Pseudotime analysis
The Monocle R package (v2.14.0) was applied for pseudotime analysis [
23‐
25]. Briefly, the top 250 highly variable genes were defined as the ordering genes. We reduced the dimensionality of data using the discriminative dimensionality reduction with trees algorithm with the reduceDimension function in Monocle. Then the cellular trajectory was constructed and the two-dimensional diffusion map was generated using plot_cell_trajectory and plot_genes_in_pseudotime functions.
Cell–cell interaction analysis
We have developed the InterCellDB package for cell–cell interaction analysis of scRNA-seq data [
26]. Briefly, the significant DEGs in each cluster were calculated as described before, and these genes were mapped to the mouse gene reference database, which was generated by collecting data from a variety of sources, including STRING, NCBI-gene, COMPARTMENTS, GO, and UniProt. Finally, matched gene pairs were visualized with the sankeyplot or dotplot function using networkD3 or InterCellDB. InterCellDB is publicly accessible as a R package in GitHub (
https://github.com/ZJUDBlab/InterCellDB).
Human tissue specimens
The radiologically healthy brain tissue sample was derived from patients undergoing brain surgery for epilepsy. Cerebrospinal fluid (CSF) was collected during preoperative lumbar puncture for patients older than 65 years with idiopathic normal intracranial pressure hydrocephalus (iNPH). After centrifugation (1500 g; 4 °C; 15 min), the supernatant was collected and stored at − 80 °C until further use. Sample collection and data analysis were approved by Institutional board of the Second Hospital affiliated to Zhejiang University (Protocol 2020–997).
Primary human microglia cultures
The brain tissue was dissociated using the Adult Brain Dissociation Kit (Miltenyi Biotec) according to the manufacturer’s protocol. Microglia were isolated using the CD11b MicroBeads (Miltenyi Biotec) and then cultured (37 °C, 5% CO2) in Microglia Medium (ScienCell) with 5% fetal bovine serum,1% penicillin, 1% streptomycin and Microglia growth supplement for 7 days. At day 8, cultures were exposed to each individual CSF sample for 24 h. The exposure medium contained 25% CSF or PBS in microglia medium.
RNA sequencing and data analysis
Total RNA from primary human microglia was isolated using TRIzol (Invitrogen, CA, United States) according to the manufacturer’s protocol. The RNA amount and purity of each sample was quantified using Nano Drop ND-1000 (NanoDrop, Wilmington, DE, USA) and the RNA integrity was assessed using the Agilent 2100 bioanalyzer. Sequence libraries were constructed according to the standard RNA-seq protocol, and 2 × 150 bp paired-end sequencing was performed with Illumina Novaseq 6000 (LC Bio) following the vendor’s recommended protocol. Cutadapt software was used to remove the reads that contained adaptor contamination. HISAT2 was used to align and map reads to the hg38 human reference genome. The mapped reads of each sample were assembled using StringTie. Then, all transcriptomes from the samples were merged to reconstruct a comprehensive transcriptome using perl scripts. After the final transcriptome was generated, StringTie and DEseq2 were used to estimate the expression levels of all transcripts. StringTie was used to perform expression level for mRNAs by calculating Fragment per Kilobase of transcript per Million mapped reads (FPKM). The differentially expressed genes (DEGs) were selected with fold change > 0.5 or fold change < − 0.5 and with statistical significance (Benjamini– Hochberg adjusted
p-value < 0.05) by DESeq2 package [
27].
Statistical analysis
The scRNA-seq data were statistically analyzed using the Wilcoxon rank sum test. Statistical comparison of the means between the two groups was performed by using the student’s t-test or the Mann–Whitney U test. Multiple comparisons were analyzed by one-way analysis of variance (ANOVA) followed by the Bonferroni multiple comparison test. All statistical analyses were performed with R (v.3.6.3) or GraphPad Prism (v.8.2.1). Statistical significance was defined as P ≤ 0.05.
Discussion
In this study, we applied intercellular network analysis of single-cell transcriptomic data from young and aged mice. Major findings of our study include that (i) peripheral T cells infiltrate the neurogenic niche during normal aging and extensively affect the brain microenvironment; (ii) aged microglia release CCL3 to recruit CD8+ memory T cells; and (iii) aged microglia shift towards a pro-inflammatory state and release TNF-α to activate venous BECs, which specifically upregulate VCAM1 and ICAM1 and promote the transendothelial migration of T cells.
Emerging evidence indicates that the immune privilege in the aging CNS is compromised [
2,
36]. CD8
+ T cells and natural killer cells were reported to accumulate in the aging brain [
1,
5,
37]. In our study, we identified a considerable increase of CD8
+ T cells in the aged SVZ using single-cell transcriptomic analysis and immunofluorescence. Several studies have emphasized the harmful role of the infiltrated T cells. T cells in the aged brain were reported to be detrimental for neural stem cells function by inducing interferon-γresponse [
5]. Accumulation of CD8
+ T cells drives axon degeneration in the normal aging mouse CNS and contributes to age-related cognitive and motor decline [
1]. Similarly, we found that the infiltrated T cells express markers associated with TCR activation and interactions with non-lymphoid cells. They were involved in a complex intercellular network and showed significant interactions with microglia, macrophages, endothelial cells, and oligodendrocytes. Furthermore, our data showed that several cytokines they release, such as C
cl4, C
cl5, X
cl1, I
fng, F
lt3l, and L
tb, widely affect various resident cells in the aged brain.
The major objective of this study was to explore transcriptomic alterations in the brain microenvironment when hematogenous T cells enter the brain through intercellular analysis, as there are only few leukocytes in the brain participating in immune surveillance in a healthy state. A critical issue that needs to be addressed is whether T cells enter the aged brain passively or actively. Our results suggest that the age-related T cell infiltration is primed by microglia. We found that hematogenous CD8
+ T cells undergo tremendous changes during normal aging, which corroborated evidence from previous studies [
28,
38]. While the upregulation of C
cr2 and C
cr5 was specifically found in CD8
+ memory T cells, their corresponding ligand CCL3, an important chemokine in the migration of effector T cells during CNS infection [
39], showed an age-related upregulation in human plasma and in microglia. These results implicated that microglia actively recruit CD8
+ memory T cells during normal aging. These findings are in line with previous reports, in which the TCR repertoire of aged brain T cells were identified to be clonally expanded and different from that of aged blood, supporting the antigen-driven infiltration hypothesis [
5].
Endothelial cells are major participants in and regulators of the inflammatory response [
40]. Their interaction with circulating leukocytes is a critical step in pathogenesis of inflammation reactions. A plethora of studies showed that adhesion molecules and chemokines play critical roles in T cell migration across the BBB [
41,
42]. However, no studies have thoroughly investigated the mechanisms underlying age-related T cell infiltration across the BBB. Single-cell transcriptomic analysis allows us to identify endothelial cell subtypes along the arteriovenous axis [
43,
44]. According to the reported transcriptomic profiles of different vascular zones, we classified the BEC transcriptome into six subclusters. Consistent with a previous study [
14], we found that aging induced distinct transcriptomic alterations across the vascular bed. T cell immune surveillance mainly occurs in the perivascular space of postcapillary venules [
8]; correspondingly, we found that biological processes related to T cell migration were specifically upregulated in venous BECs with age. Using cell–cell interaction analysis, we identified
Vcam1 and
Icam1 as key genes mediating age-related T cells transendothelial migration by firm adhesion and spreading of leukocytes. Although studies focused on the normal aging brain remain limited, Yousef et al. have shown that blocking VCAM1 could reverse microglial reactivity and cognitive deficits in the brain of aged mice [
45]. Lastly, our data revealed that aged microglia shift towards a pro-inflammatory state and significantly upregulate TNF-α, which could evoke an inflammatory response of BECs and upregulate the expression of VCAM1 and ICAM1 [
46]. All of these findings from our present study indicate that the pro-inflammatory cytokines released by aged microglia are major contributing factors affecting the upregulation of adhesion molecules on venous BECs.
Several limitations of the present study should be noted. First, our findings are mainly based on the analysis of single-cell transcriptomic data from small number of mice, while changes in the transcriptome do not always dictate molecular alterations at the protein or functional level. In this study, we have verified several changes at the protein and functional levels including the infiltration of CD8+ T cells, IgG extravasation, and CCL3, CCL4, VCAM1, ICAM1, and TNF-α protein expression. Nevertheless, further studies are needed to verify the functional implication of transcriptional alterations derived from this report. Second, the present research only focused on the establishment of firm adhesion between T cells and BECs. The mechanisms by which T cells cross the endothelium and the glia limitans during normal aging still need to be further studied.
The absence of transcriptomic analysis in aged female mice is an important flaw in this paper, given the salient sex differences in the aging process [
47]. To reduce the gender bias on our conclusions, we validated our main results in female mice. First, we found an obvious age-related T cells infiltration in the SVZ in female mice. And a remarkable increase of CCR2
+ CD8 memory T cells was also identified in female mice. Besides, we found aged microglia in female mice also transit to a chemotactic and pro-inflammation state with elevated expression of CCL3 and TNF-α. These results suggest that chronic inflammatory process occurs in both male and female mouse brain during normal aging.
In this study, an in vitro experiment of human microglia exposed to CSF from the elderly or PBS was employed to explore the changes of human microglia in aged environment. Microglia extracted from an epilepsy patient showed obvious pro-inflammatory transformation in vitro. A recent study showed that young CSF restores oligodendrogenesis and memory functions in aged mice via Fgf17 [
48], stressing the significance of identifying the changes of aged CSF and the mediators that promote CNS degeneration. Given the complex composition in CSF, studies with large sample size are warranted to explicit the mediators of age-related microglia response.
Although the exact link between microglia, endothelial cells, and T cell migration remains to be established, we have provided transcriptome evidence of the mechanism by which T cells infiltrate the aged brain and suggest avenues through which to maintain brain homeostasis during normal aging.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.