A transcriptomic atlas of aged human microglia

Abstract

With a rapidly aging global human population, finding a cure for late onset neurodegenerative diseases has become an urgent enterprise. However, these efforts are hindered by the lack of understanding of what constitutes the phenotype of aged human microglia-the cell type that has been strongly implicated by genetic studies in the pathogenesis of age-related neurodegenerative disease. Here, we establish the set of genes that is preferentially expressed by microglia in the aged human brain. This HuMi_Aged gene set captures a unique phenotype, which we confirm at the protein level. Furthermore, we find this gene set to be enriched in susceptibility genes for Alzheimer's disease and multiple sclerosis, to be increased with advancing age, and to be reduced by the protective APOEε2 haplotype. APOEε4 has no effect. These findings confirm the existence of an aging-related microglial phenotype in the aged human brain and its involvement in the pathological processes associated with brain aging.

Fig. 1

The HuMi_Aged and its relationship to genetic risk of AD. a Scatter plot depicting the distribution of gene expression values between the bulk DLPFC (N = 540) and the isolated microglia (N = 10). Each dot represents a gene. The X and the Y axes show normalized expression values. One thousand fifty four genes have been identified to be preferentially expressed by microglia (based on the differences in gene expression between the isolated microglia and the bulk cortical tissue, FC > 4) in the aged human brain (dark blue dots). This set of genes has been coined the HuMi_Aged gene set. The HuMi_Aged gene set contained many of the previously established microglia markers in the brain (black dots), such as CD74, CX3CR1, P2RY12, TREM2 or GPR34. b Scatter plot showing the distribution of AD risk genes (black dots) in the transcriptomic universe defined by the bulk cortical and the microglial RNA-Sequencing datasets. Each dot represents a gene. The X and the Y axes show normalized expression values. By using an overrepresentation test, the HuMi_Aged gene set (blue dots) was found to be significantly enriched in Alzheimer’s disease risk genes (black dots; enrichment p-value = 4.1e-05). The HuMi_Aged included AD risk genes, such as CD33, TREM2, INPP5D, APOC1 or SCIMP, while other AD risk genes (e.g., BIN1 or TREML2) were found to be not specific to microglia in the aged brain. c APOE ε2 was associated with reduced expression of HuMi_Aged in the bulk tissue level data (N = 540). DLPFC dorsolateral prefrontal cortex, FC fold change

Fig. 2

The phenotype of aged human microglia. Functional annotation reveals that human microglia aging manifests itself as loss of function as well as gain of function changes in phenotype. a Volcano plot depicting the results of differential gene expression analysis between the Zhang et al. (middle aged microglia; N = 3) and our dataset (aged microglia; N = 10). One thousand sixty genes were found to be upregulated while 1174 genes were downregulated with aging. b, c GSEA enrichment plots of representative gene sets that were significantly enriched in either downregulated (b) or upregulated (c) genes. The TGFβ signaling pathway gene set was enriched in downregulated genes (loss of function with aging, (b)) while the Amyloid fiber formation pathway was enriched in upregulated genes (gain of function with aging, (c)). d Pathway analysis of the genes that were significantly upregulated with aging in microglia in the two RNA-Seq datasets. e Pathway analysis of the proteins detected with a shotgun proteomic approach in aged human microglia. Please note the significant overlap between the functional phenotypes revealed by the two approaches (overlapping REACTOME entities highlighted in orange). Legends in d and e: orange diamonds (–log2(FDR)), columns (Percentage of entities). GSEA gene set enrichment analysis, FDR false discovery rate