A core transcriptional signature of human microglia: derivation and utility in describing region-dependent alterations associated with Alzheimer\'s disease
Summary (3 min read)
Introduction
- Microglia are the most abundant myeloid cell type in the central nervous system (CNS), accounting for approximately 5-20% of the brain parenchyma depending on region (Lawson, Perry, Dri, & Gordon, 1990; Mittelbronn, Dietz, Schluesener, & Meyermann, 2001) .
- These cells are phenotypically plastic and exhibit a wide spectrum of activity influenced by local and systemic factors (Cunningham, 2013; Perry & Holmes, 2014) .
- As the primary immune sentinels of the CNS, microglia migrate towards lesions and sites of infection, where they attain an activated state that reflects their inflammatory environment (Leong & Ling, 1992) .
- Recent transcriptomic studies have sought to characterize the human microglial transcriptomic signature from the CNS of non-neuropathologic individuals using data derived from either cells or tissue isolated from different brain regions (Darmanis et al., 2015; Galatro et al., 2017; Hawrylycz et al., 2012; Oldham et al., 2008) .
Comparison of published microglial signatures
- Ten publications that defined microglial signatures, four in human and six in mouse, were identified (Table 1 ).
- To compare across studies, genes from each signature were converted to a common identifier i.e. HGNC (Povey et al., 2001) or MGI (Shaw, 2009) for human and mouse, respectively, using the online tool g:Profiler (Reimand et al., 2016) .
- CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
- The copyright holder for this preprint (which was not this version posted April 26, 2018.
Transcriptomics data acquisition and pre-processing
- Tissue and cell transcriptomic datasets derived from the CNS were acquired for the derivation of the human microglial signature.
- The ABA data, generated on the Agilent microarray platform, consisted of 3,702 tissue samples taken from six individuals with up to 411 unique anatomical regions of the brain.
- Data quality was assessed using the ArrayQualityMetrics package (Kauffmann, Gentleman, & Huber, 2008) in Bioconductor, and samples failing more than one of three metrics (between arrays comparison, array intensity distribution and variance mean difference) were removed.
- GTEX and ABA preprocessed RNA-Seq data was downloaded directly from the GTEx portal .
- CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
Gene annotation through coexpression networks analysis
- To define a core microglial gene signature, the tissue and cell transcriptomics datasets described above were analyzed using the coexpression network analysis tool Graphia Professional.
- For each dataset, Pearson correlations (r) were calculated between all genes to produce a gene-to-gene correlation matrix.
- From this matrix, a gene coexpression network (GCN) was generated, where nodes represented genes and genes correlating greater than a defined threshold were connected by edges.
- All Pearson threshold values used for individual datasets were above r ≥ 0.7 and thereby graphs included only correlations that were highly unlikely to occur by chance .
- For each dataset, the threshold for correlations was further adjusted to achieve a single microglial cluster containing the three canonical marker genes for microglia, CX3CR1, AIF1 and CSF1R (Elmore et al., 2014; Mittelbronn et al., 2001) .
Validation of the core human microglial signature
- Various lines of evidence were investigated to validate the conserved nature of the derived human microglial signature.
- The copyright holder for this preprint (which was not this version posted April 26, 2018.
- BioRxiv preprint mouse, for comparable regions (Lawson et al., 1990), also known as doi.
- Signature genes were then compared with other published mouse and human microglial signatures.
- A GCN constructed (r ≥ 0.7) from the GTEx RNA-Seq dataset revealed five gene clusters enriched in Galatro et al. signature genes, representing various region-specific expression profiles (Table S5 ).
Heterogeneity of existing microglial signatures from human and mouse
- To examine the human microglia gene signature, previous signatures from human brain tissue or cells were compared .
- Four such studies varied considerably in the number of genes they defined, ranging from 21 to 1,236 genes.
- Of the 1,464 unique genes identified in all these studies, only a fraction (15%, 214 genes) were present in two or more signatures, with only 10 genes reported by all four publications.
- To verify that these results were not purely attributed by the individual variation in humans, the six publications reporting mouse-microglial signatures were also compared.
- Altogether these listed 690 genes (ranging from 47 to 433 genes) with 300 orthologues common to studies in human.
Derivation of a conserved core human microglial signature
- Observing the variability across published studies, the authors set out to define a human microglia gene signature from human tissue and cell data using a GCN . CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
- This method exploits the inherent variability amongst samples due to variation in sampling, donors and cellular diversity across different CNS regions.
- For constructing GCNs, genes are represented by nodes, and connected by an edge based on the similarity between their expression profiles, as quantified by the Pearson correlation coefficient .
- The final high confidence microglia gene signature was defined by 249 genes, which were present in three or more dataset-derived clusters, so as to avoid biases towards individual datasets.
Validation and description of the core human microglial signature
- To validate the microglial signature genes, various lines of evidence were examined.
- CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
- The copyright holder for this preprint (which was not this version posted April 26, 2018.
- A majority of the core signature genes (64%, 142 genes) were identified in two or more human studies, whilst 99 genes overlapped solely with the Galatro et al. signature.
- To further validate the specificity of the current microglial signature, the coexpression of these genes was compared with that of the Galatro et al. signature (1,236 genes), which included the majority of genes in other signatures.
Discussion
- CC-BY-NC-ND 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
- The copyright holder for this preprint (which was not this version posted April 26, 2018.
- The authors initial investigations demonstrated that published microglia gene signatures vary considerably in their size and composition relative to one another.
- When comparing Alzheimer's to age-matched controls, a similar trend towards increased expression levels of signature genes was also observed.
Did you find this useful? Give us your feedback
Citations
29 citations
References
82 citations
82 citations
"A core transcriptional signature of..." refers background in this paper
...…these findings TYROBP knockout mice models have proven to suppress inflammation in neurodegenerative models including Alzheimer’s, thereby minimizing neuronal dystrophy, implicating a failure in the resolution of inflammation in Alzheimer’s (Bakker et al., 2000; Haure-Mirande et al., 2017)....
[...]
81 citations
"A core transcriptional signature of..." refers background in this paper
...Genes relating to TYROBP signaling, which is implicated in Alzheimer’s and together regulates phagocytosis, cell proliferation, activation and survival were significantly upregulated (Keren-Shaul et al.; Landreth & Reed-Geaghan, 2009; Ma et al., 2015)....
[...]
80 citations
"A core transcriptional signature of..." refers background in this paper
...Alternatively, whilst tissue gene expression can be influenced 414 by cell activation and cell numbers, certain genes found differentially upregulated in 415 both Alzheimer’s and aging, such as TSPO, MS4A6A and MHC class 2 genes, are 416 known contributors of microglial activation based on previous studies (Bergen, 417 Kaing, Jacoline, Gorgels, & Janssen, 2015; Hamelin et al., 2016; Hu et al., 2017)....
[...]
...…numbers, certain genes found differentially upregulated in both Alzheimer’s and aging, such as TSPO, MS4A6A and MHC class 2 genes, are known contributors of microglial activation based on previous studies (Bergen, Kaing, Jacoline, Gorgels, & Janssen, 2015; Hamelin et al., 2016; Hu et al., 2017)....
[...]
75 citations
"A core transcriptional signature of..." refers background in this paper
...These include TLR signaling (TLR1, TLR2), complement pathway (C3AR1, C1QA and C2), TYROBP signaling (TREM2, TYROBP), and cytoskeletal organization (AIF1, CAPG and WAS) (Figure 3C)(Hong et al., 2016; Marinelli et al., 2015; Yeh, Hansen, & Sheng, 2017)....
[...]