People often associate Schrödinger with the observer effect, namely that you cannot measure something without perturbing it. Are researchers studying microglia facing an observer paradox of their own? Just how fickle these immune cells truly are dominated conversations at “Microglia in the Brain,” a Keystone symposium held June 12-16 in Keystone, Colorado. Recently researchers have begun using transcriptomics and genomics to characterize these cells from human and mouse brain and figure out how they respond to environmental cues. A take-home message from this symposium was that once yanked out of their normal environs, microglia can change so radically that they no longer appear to be microglia.

In his keynote address, Christopher Glass, University of California, San Diego, emphasized the de-differentiation that microglia undergo. Once you remove microglia from the brain, Glass said, their transcriptome changes fast and the transcripts that fade the most are those that made the cells unique in the first place. He reported that the expression of microglial-specific genes drops fourfold within six hours of culturing the cells in vitro. The change is usually complete by eight hours. “That’s on the same time scale as how long they take to sense their local environment in vivo,” said Glass, suggesting that microglia need constant signals from their surroundings to maintain their phenotype.

Glass attributed these expression changes to a dramatic reorganization of gene enhancers. His group had previously reported that the local environment influences the complement of active enhancers that drive gene expression in microglia and macrophages (see Feb 2015 conference news on Gosselin et al., 2014). Enhancers are regions of DNA that bind transcription factors determining cell lineage, and they are subject to epigenetic regulation. Any given cell contains 20,000 to 30,000 active enhancers, of which up to 300 are deemed “super enhancers.” “These are DNA regions where the cell has collected a lot of resources to regulate expression, so we think they are particularly important,” Glass said. The function of about 25 of the microglial super enhancers is known; they regulate expression of genes such as the fractalkine receptor and the microglial transcription factor PU.1. “We don’t know much about the other genes covered by these super enhancers,” said Glass.

Glass noted that when microglia are removed from their natural habitat, half of the super enhancers fade away. The significance of this genetic reorganization worried researchers at Keystone, given that many studies use isolated microglial cell lines. Researchers seemed to agree that those cells may be useful in addressing certain questions, but that they poorly reflect the overall behavior of microglia in vivo. To come to grips with this, researchers have begun to profile the transcriptomes of microglia from the brain to determine what makes these cells unique and how they respond to different environmental challenges, such as inflammation or Aβ pathology.

Researchers from Joe El Khoury’s lab at Massachusetts General Hospital proposed a mouse microglia “sensome.” This would be a cadre of transcripts encoding proteins and receptors involved in detecting ligands and microbes (see Mar 2013 conference news and Hickman et al., 2013). Oleg Butovsky and colleagues at Brigham and Women’s Hospital, Boston, identified a purely microglial transcriptome that depends on TGF-β signaling (see Butovsky et al., 2014). Other groups have analyzed microglial transcriptomes in various mouse models of disease, such as neurodegeneration or glioblastomas.

Signature Transcriptomes
How do these transcriptomes compare, and what can they tell us about pathology? Brad Friedman, a computational biologist at Genentech, South San Francisco, addressed this by analyzing 18 different data sets looking for microglial genes that are co-regulated, i.e., that tend to be perturbed in the same way. In Keystone, Friedman reported on large clusters of genes that are uniquely regulated in different disease models, perhaps reflecting distinct ways in which microglia are activated. For example, he found one cluster that responds to interferon and another that drives cell proliferation. He also found a “neurodegenerative disease cluster” of about 100 genes that are common to PS2APP and 5xFAD models of AD, and to an SOD1 model of ALS. This cluster showed no response to infection or when mice are challenged with lipopolysaccharide, which elicits robust immune responses in mice. The implication is that this cluster reflects a unique neurodegenerative signal, but whether it represents a toxic or neuroprotective response is not known. Friedman wants to study this and also what makes this signature turn on and off. “If we could modulate it, perhaps we could alter the course of disease,” he said. The neurodegenerative disease expression signature seemed to increase in older mice compared to younger, and in the cerebellum over other brain regions. 

Intrigued by these signatures, scientists at the meeting peppered Friedman with questions. Has he looked at tau models? (He has not.) Are there disease-specific signatures? (He had no specific examples.) Could he identify some curiously unlabeled clusters on his slides? When pressed, Friedman revealed that one of them was from a knockout mouse that might model AD, but this cluster poorly overlapped with the neurodegenerative disease cluster.

While Friedman didn’t elaborate on individual genes in this neurodegenerative cluster in his talk, he told Alzforum that ApoE was one of them. “I found that interesting because people think ApoE is mostly expressed in astrocytes,” he said. The cluster was highly enriched in transcripts for lipoprotein receptors and transmembrane and secreted proteins, said Friedman. This suggests the microglia may be engaging more with their environment. IGF-1, thought to be neuroprotective in ALS, turned up in the cluster as well.

Some researchers queried Friedman about batch effects. How valid is it to compare expression levels across different data sets? Friedman said he used z-score normalization, a statistical method that allowed him to compare relative expression levels across different studies, with each study serving as its own control.

Whether these mouse transcriptome signatures reflect what goes on in the human brain remains to be seen. Researchers are only just beginning to profile microglia from human tissue, and they face many hurdles, not least being access to fresh tissue. Using a dataset generated at Genentech of gene expression in the fusiform gyrus of people with Alzheimer’s disease, Friedman looked to see if the microglial neurodegenerative signature turns up. The researchers had chosen this area of the brain because samples are easier to obtain than the more highly sought-after hippocampal specimens and the gyrus is also one of the earliest affected regions in the disease. While acknowledging that the data are from bulk tissue, and so include all cell types, Friedman reported that two-thirds of the 100 genes in the mouse neurodegenerative disease signature were differentially expressed in the AD samples, as well.

Several other groups are trying to analyze purified microglia from human brain. Erik Boddeke, from the University of Groningen, The Netherlands, takes fresh postmortem samples from the Dutch brain bank and also through a collaboration with the University of São Paulo, Brazil. “The major bottleneck in the field is ready availability of sufficient postmortem tissue samples,” Boddeke told Alzforum. Because he wants to examine healthy tissue and extract sufficient numbers of viable microglia for study, only about a quarter of the samples he obtains turn out to be suitable.

Even so, Boddeke has made some progress comparing transcriptome profiles of human and mouse microglia. While many genes are similarly expressed, he said there are also substantial numbers of human microglia-specific genes that turn out to be not specific to microglia in rodents. Researchers at Keystone thought that was troubling for a field that relies so heavily on mouse data.

Boddeke further reported that aging changes mouse and human microglial transcriptomes in different ways. Microglia from older mice ramp up expression of genes involved in inflammatory responses, whereas the older human microglia lean toward genes involved in neuronal support. Boddeke found that microglia from mouse models of AD and ALS express aging-like inflammatory genes. Glass considered Boddeke’s study very interesting. Working with Fred Gage at The Salk Institute, La Jolla, California, Glass studies biopsy samples taken from children who have brain tumors or epilepsy. While he cautioned that it is extremely challenging to obtain viable cells from postmortem tissue, he noted that Boddeke’s core transcriptome looked like it might turn out to be very similar to the one he identified in the samples from children. “Given that the demographics of the samples are completely different, the similarities in the data may be pointing to genes that are most highly expressed in microglia and that make them most different from other macrophages,” Glass said.

Going forward, Boddeke plans to compare transcriptomes of microglia from healthy people with those at various stages of AD. He admits the major problem will be getting sufficient numbers of high-quality, carefully staged samples. Even so, the transcriptome analysis may identify new microglial markers, since some of the genes in the microglial-specific transcriptome encode cell surface proteins.

Glass uses human microglial transcriptomes to assess a different problem, namely heterogeneity in gene expression between one person and the next. This, he said, presents a challenge for the field, but also an opportunity. Glass told Alzforum that so far, he has studied samples from about 20 different children, and while the expression levels of some microglial genes are similar, others vary considerably. For example, Glass noted a fourfold difference in TREM2 and dramatic differences in ApoE expression. “It’s possible we might be able to predict such quantitative changes based on genotype,” said Glass. “That could ultimately help us predict what is going on in the brain in response to environmental stress.”

Heterogeneity in microglial gene expression could make it difficult for researchers to interpret the effects of hits from AD genome-wide association studies (GWAS). Many variants uncovered by those studies lie in non-coding regions in microglial genes and are thought to weakly influence gene activity. How to distinguish these subtle effects against background noise? Here again, Glass thinks expression analyses could help. The key point, he said, is that regulatory regions in microglia are very different from regulatory regions in neurons or astrocytes, therefore a cell-specific approach is needed to understand the effect of a given genetic variant on these cells. “If we can overlay microglial enhancer atlases on GWAS loci, then we may obtain very useful information.” 

It became clear at the meeting that scientists need to work out how to keep track of the reams of transcriptome data that are beginning to flow from different labs. Boddeke has set up a microglial gene expression database (see Holtman et al., 2015) where researchers can deposit and access transcriptomics data. While some individual labs maintain their own databases, Boddeke said it will be immensely valuable for the field to have access to data in a standard format that is as open as possible. He plans to take an expanded version live in October.—Tom Fagan


No Available Comments

Make a Comment

To make a comment you must login or register.


News Citations

  1. Nature Versus Nurture: What Gives Microglia Their Identity?
  2. Microglia Activation—Venusberg Meeting Questions M1, M2 Designations

Research Models Citations

  1. PS2APP
  2. 5xFAD (B6SJL)

Brain Bank Citations

  1. Netherlands Brain Bank

Paper Citations

  1. . Environment drives selection and function of enhancers controlling tissue-specific macrophage identities. Cell. 2014 Dec 4;159(6):1327-40. PubMed.
  2. . The microglial sensome revealed by direct RNA sequencing. Nat Neurosci. 2013 Dec;16(12):1896-905. Epub 2013 Oct 27 PubMed.
  3. . Identification of a unique TGF-β-dependent molecular and functional signature in microglia. Nat Neurosci. 2014 Jan;17(1):131-43. Epub 2013 Dec 8 PubMed.
  4. . Glia Open Access Database (GOAD): A comprehensive gene expression encyclopedia of glia cells in health and disease. Glia. 2015 Sep;63(9):1495-506. Epub 2015 Mar 25 PubMed.

Further Reading

No Available Further Reading