Comprehensive immune ageing reveals TREM2/TIM3 myeloid cells drive brain immune evasion

Abstract

Background: Ageing-dependent low-grade inflammation is a hallmark of central nervous system (CNS) diseases. Vascular and immune abnormalities are implicated in the progression of gliomas and occur in the early stages of Alzheimer's disease (AD); however, the mechanisms by which these alterations manifest in the brain parenchyma remain unclear.

Methods: Using RNAseq, scRNAseq, bioinformatics tools and a cohort of patients with glioma and Alzheimer's disease for validation of results, we have established an analysis of blood-brain barrier (BBB) dysfunction and neuron loss. A mouse model for glioblastoma pathology was also used that reversed BBB disruption and neuron loss, with the incorporation of the IDH mutation. Finally, we established a characterization of the relevant immune populations with an IHC analysis and transcriptional profile.

Findings: In this study, molecular analyses of the brain ecosystem revealed that blood-brain barrier dysfunction and neuronal synapse integrity exhibit significant threshold-dependent changes that correlate directly and inversely, respectively, with brain ageing (significant changes at 57 years) and the progression of AD and gliomas (survival of 1525 vs 4084 days for patients with High vs Low BBB dysfunction). Using human samples and mouse models, we identified immunoageing processes characterized by an imbalance between pro-inflammatory and anti-inflammatory signals. This dysregulation promotes the extravasation of monocyte-derived macrophages (85% increase of cells), particularly those with a suppressive phenotype, alongside an increase in inflammatory cytokine levels. Notably, our data show that vascular normalization in a glioma model can reverse neuronal loss and attenuate the aggressiveness of the tumours. Finally, tumour development can be prevented by reactivating the ageing immune system.

Interpretation: We propose that the ageing brain represents a common, BBB dysfunction-associated process driving chronic inflammation. This inflammation is regulated by TREM2+/TIM3+ suppressive myeloid cells, which play a central role in disease progression. Our findings suggest that targeting these pathways could offer therapeutic strategies to mitigate CNS pathologies linked to ageing, characterized by toxic neuroinflammation and myeloid dysfunction.

Funding: This study was funded by ISCIII and co-funded by the European Union.

Keywords: Ageing; Alzheimer's disease; Glioblastoma; Immune evasion; TIM3; TREM2.

Fig. 1

Ageing leads to disruption of the BBB, synapse loss and increased pro- and anti-inflammatory signals. a, b, Representative pie chart of the processes in which down-regulated (a) and up-regulated (b) genes are involved in ageing, using the results obtained from the GSEA analysis performed on the GTEx database (n = 367) using age as a template and the different Biocarta datasets. c, Heatmap of BBB dysfunction and Synapse signatures gene expression analysis in healthy brain tissue from GTEx dataset (n = 367). Red, highest expression. Blue, lowest expression. d, Fold change of BBB dysfunction gene expression according to patients' age from GTEX dataset (n = 367). Significance was determined by two-tailed unpaired Student's t-test. e, f, Analysis of BBB-dysfunction (e) and synapse (f) gene signatures expression by RNA-seq in healthy brain tissue (GTEx cohort, n = 367), divided according to the age: age <57 years old and age >57 years old patients. Statistical significance was determined by unpaired Student's t-test with Welch's correction. g, qRT-PCR analysis of BBB-dysfunction gene signature in 3- and 13-month-old C57/B16 mouse brains (n = 5/group). Actin was used for normalization. Statistical analysis was performed using the two-tailed unpaired Student's t-test. h, Quantification of the vascular density measured with endomucin staining and TER119 positive cells from IF (n = 6/group). Statistical analysis was performed by two-tailed unpaired Student's t-test. i, qRT-PCR analysis of synapse gene signature in 3- and 13-month-old C57/B16 mouse brains (n = 5/group). Actin was used for normalization. Statistical analysis was performed using the two-tailed unpaired Student's t-test. j, GSEA enrichment plot analysis using the age as template and different gene set from the Biocarta pathways database: Inflammatory response, TNF $\alpha$ signalling, IL6-JAK-STAT3 signalling and INF $\gamma$ response. k, Heatmaps of inflammatory signalling (left) and anti-inflammatory signalling (right) gene expression measured by qRT-PCR in healthy brain samples (healthy brain cohort, n = 59) divided by age in <57 Years and >57 Years. HPRT was used for normalization. Statistical significance was determined by unpaired Student's t-test with Welch's correction. l, m, Fold change of inflammatory signalling (l) and anti-inflammatory signalling (m) gene expression according to patients' age from GTEX dataset (n = 367). Significance was determined by two-tailed unpaired Student's t-test. ∗p < 0.05; ∗∗p < 0.01,: ∗∗∗p < 0.001: ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 2

Patients with glioma older than 57 years show a greater BBB dysfunction and immune cells infiltration. a, Age distribution of patients with glioma, divided into LGG and GBM, from the TCGA cohort (GBM + LGG) (n = 702). b, Analysis of age range of patients with glioma, divided into Astrocytoma IDHmut and GBM IDHwt, from our glioma cohort (n = 143). c, Percentage of dilated BV per field (left) (n = 47) and vascular density (right) (n = 39) of patients with glioma from our own cohort. Patients are divided according to age into <57 Years and >57 Years.d, Analysis of necrosis score (n = 45) established on immunohistochemical staining in glioma cohort divided into <57 Years and >57 Years. e, Analysis of CAIX gene expression (n = 37) in glioma cohort divided into <57 Years and >57 Years. f, Quantification of IgG extravasation area (%) measured on immunohistochemical stains of patients <57 Years and >57 Years from our own glioma cohort (n = 26). Significance was determined by Mann–Whitney test. g, qRT-PCR analysis of BBB dysfunction gene signature in tumours (n = 47) from our own glioma cohort divided into <57 Years and >57 Years. HPRT was used for normalization. Significance was determined by Mann–Whitney test. h, Quantification of CD45 (n = 16), CD68 (n = 15), CD8 (n = 44) positive cells from IHC of patients from glioma cohort. Tumours were classified previously into two groups according to age: <57 Years and >57 years. i, Quantification of IBA1 (n = 16) positive cells from IHC of patients from glioma cohort. Tumours were classified previously into two groups according to age: <57 Years and >57 years. j, Representative images of IF co-staining of CD34 and IgG (top) or NeuN (bottom) on sections (n = 25) from our own glioma cohort and quantification of the percentage of NeuN positive cells in tumour tissue. k, qRT-PCR analysis of synapse gene signature in tumours (n = 26) from our own glioma cohort divided into <57 Years and >57 Years. HPRT was used for normalization. Significance was determined by unpaired Student's t-test with Welch's correction. l, Analysis of BBB disruption using IgG extravasation and CD34 IF co-staining in glioma recurrence with progression and without progression. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001: ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 3

Association of age and BBB-dysfunction with the different glioma subtype: Mesenchymal, Proneural and Classic. a, Volcano plot of whole transcriptome differential expression analysis of glioma with Low (n = 5) vs high BBB-disfunction (n = 5). b, c, GSEA enrichment plot analysis on the overexpressed genes of the differential expression analysis of (a), where enrichment of the Mesenchymal glioma subtype signature is observed (b, left) and negative enrichment of the Proneural glioma subtype (b, right), neuronal system (c, left) and the membrane synapses (c, right) signatures. d, Analysis of BBB-disfunction signature expression by qRT-PCR in our own glioma cohort (n = 30), grouped according to tumour subtype: proneural (PN), classic (CL), or mesenchymal (MES). Significance was determined by one-way ANOVA test with a Benjamini–Hochberg multiple comparison correction (FDR). e, f, qRT-PCR analysis of MES (e)- and PN subtype (f)-related genes in patients (n = 26) from our own glioma cohort. Tumours were classified previously into two groups according to age: <57 Years and >57 years. HPRT was used for normalization. p values were determined by a two-tailed unpaired Student's t-test. g, Representative images of IF co-staining of endomucin and IgG (top) or NeuN (bottom), on sections from a panel of patient-derived xenografts (PDXs), classified previously into different subtypes: proneural (PN), classic (CL), or mesenchymal (MES). h, i, Quantification of IgG extravasation (h) and NeuN positive cells per field (i) from two PDXs of each subtype. p values were determined by unpaired Student's t-test. j, Kaplan–Meier overall survival curves of mice that were orthotopically injected with GBM1 (MES glioma subtype) and GBM2 (PN glioma subtype) cells (n = 6). Survival curves was determined by a two-sided log-rank (Mantel–Cox) test. k, l, qRT-PCR analysis of synapse (k) and BBB-dysfunction (l) gene signature in intracranial tumours from (g). HPRT was used for normalization. Significance was determined by unpaired Student's t-test with Welch's correction. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 4

IDH mutation modulates the vascular properties of glioma. a, Analysis of non-silent somatic mutations in genes commonly modified in diffuse glioma ordered according to the expression of BBB-dysfunction and synapse gene signature using the (TCGA-LGG + GBM) (n = 661). b, Volcano plots showing highly methylated genes with differential distribution in glioma comparing tumours with presence or absence of IDH mutations. c, Kaplan–Meier overall survival curves of patients from the TCGA GBM cohort (n = 539). Patients were stratified into two groups based on G-CIMP or non-GCIMP phenotype. Survival curves were determined by a two-sided log-rank (Mantel–Cox) test. d, Kaplan–Meier overall survival curves of mice that were orthotopically injected with GBM1 control or GBM1 IDH1 R32H cells (IDH mut) (n = 6/group). Survival curves were determined by a two-sided log-rank (Mantel–Cox) test. e, Representative images of IF co-staining of endomucin and IgG (top) or NeuN (bottom) on sections from intracranial tumours from (d). f, g, Quantification of IgG extravasation (f) and the percentage of NeuN positive cells (g) in tumour tissue from (e) (n = 3/group). Statistics by unpaired Student's t-test. h, i, qRT-PCR analysis of BBB-dysfunction (h) and synapse (i) gene signature in tumours from (d). HPRT was used for normalization. Significance was determined by unpaired Student's t-test with Welch's correction. j, Comparison of performance on rotarod test between animal from (d). Mean latency to fall times (second) for each mouse (n = 5/group) were obtained from the mean of quadruplicates performed per day. Significance was determined by unpaired Student's t-test with Welch's correction. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001: ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 5

Ageing is responsible of brain inflammation in patients with glioma. a-d, Flow cytometry analysis of the immune infiltrate of gliomas. (a) Percentage of CD45+, lymphoid (CD45 + CD11b-SSClo) and myeloid (CD45 + CD11B + SCChi) cells on total tumour suspension. Percentage of macrophages (b), myeloid-derived suppressor cells (MDSCs) (c) and neutrophils (d) on total tumour suspensions. Glioma cohort stratified into two groups based on age >57 Years and <57 Years (n = 8/group) measured in Fig. S3e. Statistics by a two-tailed unpaired Student's t-test. e, Representative images of IF co-staining of CD68 and CD49d on sections from our own glioma cohort, stratified in <57 Years and >57 Years (n = 10/group). Quantification of the percentage of CD68/CD49d positive cells in tumour tissue on the right. Statistics by a two-tailed unpaired Student's t-test. f, Analysis of Monocytes-derived macrophages (MDM) gene signature determined by qRT-PCR analysis in our own glioma cohort, divided in <57 (n = 12) Years and >57 Years (n = 11). On the right a heatmap representation of the expression of each gene included in the MDM signature for each patient. HPRT was used for normalization. Significance was determined with Mann–Whitney test. g, WB analysis (left) and quantification (right) of CSF1-R, phospho- NF-kb and phospho-STAT3 in tissue extracts from patients with glioma stratified according to age into <57 Years and >57 Years groups (n = 13/group). GAPDH was used as a loading control. Statistics by a two-tailed unpaired Student's t-test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 6

Characterization of myeloid cell dysfunction in the brain tumour microenvironment. a, Representative image and quantification of TREM2 IHC staining in glioma of each group: <57 Years and >57 Years (n = 25). Statistics by unpaired Student's t-test. b, c, Representative image of the TREM2 and HAVCR2 (TIM3) expression in each cell cluster (Macrophage, Malignant cells, Oligodendrocyte and T-cells) by Single-cell RNA-seq (b) of 28 GBM tumours. Spearman's rho (b) was calculated by Spearman's rank correlation coefficient. d, Representative image (left) and quantification (right) by IF from <57 Years to >57 Years glioma tissue (n = 6/group) of CD68+/TREM2+/TIM3+ positive cells. Statistics were determined by unpaired Student's t-test. e, Dot plot expression of immunosuppressive related genes in each cell cluster. Colours (colour-scaled) indicate the average expression of selected marker genes, and the circle sizes indicate the proportion of cells expressing the selected gene. f, g, Analysis of inflammatory (f) and suppressive myeloid phenotype (g) genes signatures determined by qRT-PCR analysis in our own glioma cohort, grouped according to age in <57 Years and >57 Years (n = 22). HPRT was used for normalization. Significance was determined with a Mann–Whitney test. h, Representative IF image of CD8+/PD-1+ positives in association with TREM2+ cells from glioma tissue of <57 Years and >57 Years patients (n = 8/group). p values were determined by Mann–Whitney test. i, Correlation analysis of CD8+/PD-1+ positive cells with TREM2 positive cells in <57 Years and >57 Years gliomas. Spearman's rho was calculated by Spearman's rank correlation coefficient. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001: ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 7

Ageing is responsible of brain inflammation in patients with AD. a, b, Heatmap representing colour-coded expression levels of BBB-dysfunction (a) and synapse (b) gene signature in our own Alzheimer cohort (n = 29). In both cases, cerebral cortex/hippocampus normal tissue was used as control. The expression was determined by qRT-PCR analysis and HPRT was used for normalization. c, Histogram representing the distribution of synapse (left) and BBB-dysfunction (right) gene signatures between normal tissue (NT), AD Braak I-III and AD Braak IV-VI (n = 29). Expression was determined by qRT-PCR analysis and HPRT was used for normalization. Significance was determined with Mann–Whitney test. d, ROC curve plotted for diagnostic potential and discriminatory accuracy of BBB-dysfunction and synapse gene signature to distinguish AD disease from healthy patients. The corresponding AUC (Area) and p value are reported. e, Analysis of Monocytes-derived macrophages (MDM) gene signature determined by qRT-PCR analysis in normal tissue (n = 8) and patients from our own AD cohort (n = 19). HPRT was used for normalization. Significance was determined with a Mann–Whitney test. f, Analysis of inflammatory (left) and suppressive myeloid phenotype (right) gene signatures determined by qRT-PCR analysis in normal tissue (n = 8) and patients from our own AD cohort (n = 19). HPRT was used for normalization. Significance was determined with a Mann–Whitney test. g, Analysis of TREM2 expression by microarray in patients with Alzheimer disease compared to normal tissue using Berchtold (n = 80) cohort (GSE48350). h, Single-cell RNA-sequencing of TREM2 expression in Alzheimer disease model. Sequencing data are available on Single Cell Portal.i, j, Correlation between TREM2 and TIM3 expression in our own AD cohort (n = 29), divided into low Braak (I-III) (i) and high Braak (IV-V) (j). Pearsons's correlation test was used. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s., not significant.

Fig. 8

Impact of ageing in BBB dysfunction and neuroinflammatory response. a, Fold change of inflammatory (left) and myeloid suppressor phenotype (right) gene expression according to patients' age from cerebral cortex/hippocampus GTEX dataset (n = 367). Significance was determined by two-tailed unpaired Student's t-test. b, Analysis of inflammatory (left) and suppressive myeloid phenotype (right) gene signatures determined by RNA-seq in healthy brain tissue (cerebral cortex/hippocampus; GTEx cohort, n = 367), divided according to the age: age <57 years old and age >57 years old patients. Statistical significance was determined by unpaired Student's t-test with Welch's correction. c, qRT-PCR analysis of Inflammatory (left) and suppressive Myeloid phenotype (right) gene signature in brain of young (3 months) and old (13 months) mice (n = 5/group). Actin was used for normalization. Significance was determined with a Mann–Whitney test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s., not significant. d-f, Representative images (d) and quantification (e, f) of IF co-staining of TREM2 and TIM3 (e) or CD8+/PD-1+ (f) on sections from young and old C57/B16 mice brain (n = 5/group). Quantification is measured as the percentage of positive cells for the corresponding markers in tumour tissue. Statistics by a two-tailed unpaired Student's t-test. g, Heatmap comparing fold-changes of different immunological processes between young (3 months) and old (13 months) mice: Inflammatory response (TAM M2), MDSCs gene signature, Immunocheckpoint markers, INF$\gamma$ response, IL6-STA pathway and TNF-α signalling via NFkβ. (n = 5/group). h, Representative scheme of the experiment performed on young (3 months) and old (13 months) C57/BL6 mice, in which a group of old C57/BL6 mice were treated with immunotherapy (anti-PD1 and anti-TIM3 antibodies) intraperitoneally prior to tumour implantation. i, Kaplan–Meier overall survival curves of nude mice and C57/BL6 mice of different age: 3 months and 13 months, that were orthotopically injected with SVZ-EGFRvIII cells. In addition, a group of C57/BL6 old were treated with intraperitoneal injections of anti-PD1 (5 mg/kg/day) and anti-TIM3 (5 mg/kg/day) n = 6 nude; n = 20 C57/young; n = 20 C57/old; n = 30 C57/old plus anti-PD1/TIM3. Survival curves were determined by a two-sided log-rank (Mantel–Cox) test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s., not significant.