For brain cells, what do aging, living in white matter, and hanging out with tumors have in common? If you’re a microglia, these all upset your homeostasis. On November 18 in Nature Neuroscience, researchers led by Marco Prinz of the University of Freiburg in Germany, Dominic Grün of the Max Planck Institute of Immunobiology and Epigenetics in Freiburg, and Josef Priller of the Charité Universitätsmedizin Berlin, reported 14 subtypes of microglial cells lurking within human brain biopsy samples. Based on their gene expression, the subtypes ranged from placid, homeostatic cells to reactive, proinflammatory ones. Aging, disease, and brain region all influenced where microglia landed along that spectrum.

  • Single-cell RNA and protein analysis of brain biopsy, a bevy of microglial subtypes.
  • Expression states range from homeostatic to reactive.
  • Microglia nudged closer to reactive end of spectrum with age, in white matter, and near glioblastomas.

“This is a very significant study that makes progress on understanding the diversity of microglial phenotypes in the human brain,” commented Bahareh Ajami of Oregon Health Sciences University, Portland. The data are also a crucial resource for the field, she said.

Microglia are central players in neurodegenerative disease, and researchers are hustling to understand how these dynamic cells respond to various insults in the brain. In mouse models of disease, they have been reported to transform from homeostatic to reactive and proinflammatory (Jun 2017 news; Oct 2017 newsSep 2017 newsDec 2018 news). 

Microglia from Biopsies. From human brain biopsies (top), microglia were isolated and their transcriptomes and proteomes analyzed cell by cell. Expression signatures revealed a cadre of different microglial subtypes (bottom). [Courtesy of Sankowski et al., Nature Neuroscience, 2019.]

But what about human microglia? Researchers have faced logistical and technical challenges trying to characterize them (Feb 2018 newsJul 2018 news; May 2019 news). For one, how is one to come by live microglia from the brain of a living person? Scientists rely on precious few samples extracted during biopsies or brain surgeries in people with epilepsy or cancer. Few labs have access to that kind of tissue. Others turn to archived frozen samples, but those come with potential quality issues.

So far, regardless of the methods used, researchers are discovering that, even more than their murine counterparts, human microglia exist in a diversity of states, ranging from those that express a homeostatic core signature to others that appear to be in the midst of one inflammatory response or another. What dictates these microglial fate changes, and which ones help or harm the brain, are crucial questions in neurodegenerative disease research.

Previously, Prinz used single-cell RNA-sequencing to identify homeostatic microglia expressing a signature of core genes, as well as transcriptionally unique subsets mingling within active multiple sclerosis lesions, both from biopsied brain tissue taken from MS and epilepsy patients (Feb 2019 news). This time around, co-first authors Roman Sankowski and Chotima Böttcher included samples from glioblastoma patients. They added time-of-flight mass cytometry (CyTOF)—a single-cell proteomic technique that detects cell-surface and intracellular proteins—to corroborate some of their prior single-cell transcriptomic data (Ajami et al., 2018). 

Cloud of Clusters. Based on their gene-expression profiles, adult human microglia fell into eight clusters. [Courtesy of Sankowski et al., Nature Neuroscience, 2019.]

For their initial survey, the researchers collected non-diseased brain tissue from 15 people. Ten were undergoing brain surgery to remove epileptic foci, four to remove malignant gliomas, and one to remove a metastatic carcinoma. The microglia extracted from cancer patients came from tissue at least 2 cm away from the tumor. All told, the researchers obtained 4,396 microglia. Using single-cell RNA sequencing, they defined nine major microglial subtypes by their gene-expression profiles (see image at left). The number of individual microglial cells assigned to each transcriptional cluster varied from just 119 to 1,327. The clusters expressed by the most cells—C2 and C3—comprised a core set of homeostatic genes including TMEM119 and CX3CR1, and were found in all 15 donors. C2 also contains MHC Class II and antiviral immunity genes. Two other clusters—C6 and C7—featured integrin receptor binding protein and metabolism genes, including SPP1, ApoE, and LPL. The smaller C1, C5, C8, and C9 clusters included chemokine and cytokine genes, as in a proinflammatory identity. The researchers omitted cluster C4 from further analysis because its gene-expression data was of poor quality.

The researchers next explored what might underlie this diversity. Standard immunohistochemistry revealed that microglia inhabiting white matter expressed more MHC-II and CD68 than their fellow microglia in gray matter. These genes function in antigen presentation and phagocytosis, respectively. Further gene-expression analysis backed this up. White matter was enriched in cells expressing MHC-II-high clusters C2, C5, C6, and C7, while gray matter was enriched in cells with MHC-II-low clusters C3, C8, and C7.

Via CyTOF, the researchers used more than 50 antibodies against myeloid-specific markers to take stock of protein expression in the microglia. Corroborating the RNA-Seq data, white-matter microglia expressed higher levels of proteins involved in antigen presentation, inflammation, and lipid metabolism, including MHC-II and ApoE, but lower levels of homeostatic proteins than did the gray-matter microglia. The findings suggested that in white matter—albeit from people with epilepsy or glioblastoma—microglia shifted out of a homeostatic comfort zone into more reactive or proinflammatory states.

Age seemed to have a similar effect. The scientists compared microglial profiles in four patients aged 14–30, three between 30 and 50, and three older than 50. Strikingly, the youngest brains were flush with microglia expressing the homeostatic cluster C3, sparse with C6 and C7, and completely devoid of microglia with the more proinflammatory phenotype. Compared with the young brains, those from 30- to 50-year-olds had more C2 microglia, which were largely homeostatic except for high expression of MHC-II. Scarily, perhaps, microglia from people older than 50 had more C6 and C7 microglia, marked by high expression of ApoE and the proinflammatory gene SPP1, which codes for osteopontin.

The authors suggested that the transcriptional changes in microglia reflect increased exposure to age and/or disease-related signals, including rising inflammation throughout the body. Ajami agreed. She speculated that, compared with microglia tucked away in gray matter, those in white matter could be more exposed to inflammatory signals such as cytokines.

In a joint comment to Alzforum, Anouk Benmamar-Badel, Agnieszka Wlodarczyk, and Trevor Owens of the University of Southern Denmark in Odense noted that the distinct profile of microglia in white matter jibes with their reported role in supporting myelination. They noted that the C6 and C7 expression clusters, which rose with age and in white matter, are similar to disease-associated microglia reported in AD mouse models, which also express SPP1, APOE, GPNMB, LPL, and ITGAX.

How would brushing shoulders with an amyloid plaque or a neurofibrillary tangle tinge the microglial mood? The researchers had no biopsy from AD or tauopathy patients. They were able to profile microglia responding to a different type of trauma, a brain tumor. They compared profiles of microglia extracted from resected glioblastomas of four patients with profiles of microglia from non-tumor tissue of four age-matched patients who had either glioma or epilepsy. Fourteen gene-expression clusters emerged from the 1,701 microglia isolated from these eight samples, and their expression profiles fell along a spectrum from homeostatic to proinflammatory. At one end of the spectrum, three clusters predominantly came from non-tumor tissue, and highly expressed the core homeostatic genes. The middle eight clusters were found in cells hailing from both control and tumor samples. Some of these contained genes involved in vascular development and antigen processing, and some highly expressed in microglia from old people or within white matter, such as SPP1. Two clusters at the other end of the spectrum only appeared near tumor tissue; they had a more reactive signature, with abundant transcripts encoding metabolic, inflammatory, and interferon-associated genes.

Together, the findings suggest that microglia exist in a continuum of transcriptional states, likely based on their exposure to different inflammatory stimuli, Prinz said. The researchers used computational algorithms to piece together the reported spectrum of states. Prinz cautioned that because microglia were only analyzed at a single moment in time from each person’s brain, he can’t know whether microglia actually transition between these states or if they are committed to a particular fate. Further complicating matters is the lack of healthy controls—all the microglia in this study came from people with either epilepsy or a tumor.

Ajami suspects that the cells likely go back and forth between homeostasis and reactivity. She also noted that while aging, disease, and exposure to inflammation-stoking damage or distress all draw microglia out of homeostasis, the cells seem to respond uniquely to each type of threat. She thinks it is critical to understand how microglia respond to numerous different stimuli, and understand what prevents them from returning to their healthy state. She investigates the latter in her lab, using CyTOF on human microglia.

Prinz said that as studies continue to parse how microglia behave in different scenarios, researchers will be better able to target problem cells while leaving the homeostatic ones alone. He also hopes this work will lead to PET tracers that can track different subsets of microglia in people with neurological diseases.—Jessica Shugart



  1. Several studies have unveiled an important phenotypic heterogeneity of mouse microglia. Yet our understanding of the human microglia is limited, especially because of the technical difficulties to obtaining fresh brain tissues from human donors. Nonetheless, a few recent studies have indicated that, like the mouse, the human brain harbors a variety of microglial phenotypes that possibly reflect different topological locations and/or different functions. In this regard, this recent work by Sankowski et al. provides a detailed characterization of human microglia using a combination of single-cell RNA-Seq, CyTOF, and immunohistochemistry. Overall, this study generates a large amount of transcriptomic and proteomic information, defining microglial phenotype in both gray and white matter from individuals with either epilepsy or cancer, but no obvious brain abnormalities, including older individuals. Additionally, this work presents the first transcriptomic description of microglia, at single-cell resolution, within the human glioblastoma.

    Sankowski and colleagues perform scRNA-Seq on microglia from 15 subjects, ranging from 14 to 74 years old. Sorting CD45+ cells, the authors could successfully identify a major microglia population, but also a minor fraction of monocytes, T cells, and oligodendrocytes. Microglia from biopsied brain tissue formed eight distinct clusters, highlighting the presence of multiple phenotypic subsets under homeostasis. The majority of microglia were enriched for homeostatic signature genes (Cx3cr1, Tmem119, Csf1r, P2ry12, Selplg). Another main population expressed high levels of antigen-presentation genes (HLA-DR and CD74), indicating an immune-activated state. However, other minor subsets were enriched for inflammatory genes (Spp1, ApoE, LPL, Ccl2, Il1b and IFN-associated genes). Such a phenotypic diversity may reflect a different distribution within the brain parenchyma. Indeed, authors showed that white-matter microglia expressed high levels of HLA-DR, ApoE, CD68, and F4/80, whereas gray matter microglia exhibited a relatively low expression of these genes.

    The authors also show that the microglia phenotype changes during aging. Indeed, microglia from young individuals were mostly characterized by a homeostatic signature. By contrast, microglia from senile subjects exhibited an increased expression of inflammatory genes (like Spp1), especially in the white matter. These data show that microglial phenotype in the human brain is instructed by the local environment (i.e., gray vs. white matter) and becomes more proinflammatory during aging.

    Lastly, Sankowski and colleagues show that glioblastoma-associated microglia are remarkably heterogeneous. For example, one cluster matched the microglial phenotype of the healthy subjects. Conversely, a distinct population exhibited a significant upregulation of antigen-presentation (HLA-DR), inflammatory (Spp1, ApoE, Trem2), and interferon-signature genes, resembling more closely the phenotype of age-associated microglia. Additional studies are needed to determine whether this population is capable of presenting tumor antigens to T cells, thus eliciting an anti-tumor response. Additionally, another population was enriched for hypoxia-induced genes such as Hif1a and VEGFA. This population may encompass those microglial cells located in the most inner part of the tumor mass, which is notoriously highly hypoxic. Alternatively, this subset may define a pro-angiogenic phenotype, which may then represent a suitable target for therapy.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Hot DAM: Specific Microglia Engulf Plaques
  2. Changing With the Times: Disease Stage Alters TREM2 Effect on Tau
  3. ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
  4. Microglia Reveal Formidable Complexity, Deep Culpability in AD
  5. Microglial Transcriptome Hints at Shortcomings of AD Model
  6. A Delicate Frontier: Human Microglia Focus of Attention at Keystone
  7. When it Comes to Alzheimer’s Disease, Do Human Microglia Even Give a DAM?
  8. Single-Cell Profiling Maps Human Microglial Diversity, Flexibility

Paper Citations

  1. . Single-cell mass cytometry reveals distinct populations of brain myeloid cells in mouse neuroinflammation and neurodegeneration models. Nat Neurosci. 2018 Apr;21(4):541-551. Epub 2018 Mar 5 PubMed.

Further Reading

Primary Papers

  1. . Mapping microglia states in the human brain through the integration of high-dimensional techniques. Nat Neurosci. 2019 Nov 18; PubMed.