More than 700 participants hailing from 35 countries gathered in Keystone, Colorado, June 17–21 for a joint Keystone symposium: Advances in Neurodegenerative Disease Research and Therapy; and New Frontiers in Neuroinflammation: What Happens When CNS and Periphery Meet? Researchers presented fresh findings on how microglial meddling steers neurodegeneration, on tau’s latest antics, on ApoE’s actions independent of Aβ, and how aging transmits from the blood into the brain. Participants jumped back and forth between the two side-by-side symposia, and packed poster sessions abuzz with intense discussions capped off each data-filled day. Check out Alzforum’s coverage for the highlights.
A Delicate Frontier: Human Microglia Focus of Attention at Keystone
For a pesky cell type that comprises a small fraction of cells in the brain, microglia have an outsize impact on neurodegeneration. Many of the Alzheimer’s risk loci identified in genome-wide association studies harbor genes expressed in microglia, which seem to put their stamp on every stage of disease. Alas, their limited numbers and hot temper make microglia difficult to work with, especially the human kind, which must be plucked from postmortem brain samples. Despite these challenges, researchers presented new data on human microglia at the joint Keystone symposia—Advances in Neurodegenerative Disease Research and Therapy; and New Frontiers in Neuroinflammation: What Happens When CNS and Periphery Meet?— held June 17–21 in Keystone, Colorado. They tied the origin of AD risk more closely to microglia, wielding single-cell transcriptomics to uncover a striking degree of heterogeneity among microglia, and charting webs of microglial genetic networks that correlate with disease.
GWAS Hits: Pin the Blame on Microglia
At Keystone, Alison Goate of the Icahn School of Medicine at Mount Sinai, New York, aptly illustrated both the central role of microglia in disease, and the challenges of working with them. Goate had previously identified a polymorphism that reduced expression of PU.1, a microglial transcription factor that controls expression of many AD risk genes, including TREM2, TYROBP, CD33, the MS4A cluster, and ABCA7. The polymorphism delayed onset of AD, cementing the prominence of microglia in the disease process (Jun 2017 news). However, pinning the GWAS signal to SPI1, the gene encoding PU.1, was not a straightforward affair. As with the vast majority of GWAS hits, the polymorphism near the SPI1 gene resides in a noncoding region, meaning it could theoretically control expression of any nearby gene. To nab SPI1 as the gene responsible for the disease onset association, Goate and colleagues looked for expression quantitative trait loci (eQTL), i.e., genes whose expression was affected due to each polymorphism. Because microglial samples are too scarce to do a proper eQTL analysis, Goate resorted to the next best thing: She looked for eQTLs within the Cardiogenics Consortium, a set of gene-expression data from hundreds of monocyte and macrophage samples. This confirmed SPI1 as the most likely gene affected by the disease-onset-linked polymorphisms, at least in those related myeloid cells.
Gene Clusters. Single-cell RNA sequencing of postmortem and biopsy samples revealed 23 distinct gene-expression clusters among the cells, 14 of which expressed microglial markers. Circle size denotes proportion of the sample in each cluster. [Courtesy of Olah et al., bioRχiv, 2018.]
Goate then repeated this monocyte/macrophage eQTL analysis for more than 20 other GWAS loci. She added in a batch of new hits unveiled by a recent GWAS that counted people with “AD-by-proxy,” or family history of AD, as cases to beef up numbers (Apr 2018 news). For some of the GWAS hits, this analysis produced an obvious winner: Expression of MS4A4A, MS4A6A, PILRA, and FCR1g, for example, all correlated with disease-associated polymorphisms much more than expression of any other surrounding gene.
Other analyses produced a less obvious leaning toward one gene or another, and expression of at least one gene—Ptk2b—correlated with the nearby disease variant in monocytes, but not in macrophages. Whether monocytes or macrophages are a better proxy for microglia is unclear, Goate said. There is also the distinct possibility that some of the GWAS hits influence disease through expression in myeloid cells outside of the brain.
Christopher Glass of the University of California, San Diego, commended Goate’s work. He told Alzforum that the lack of microglial samples for eQTL analysis is an ongoing problem in the field. Glass said that his lab has access to only 25 such precious samples. Not only are microglia sparse in the brain, the cells are notorious for transforming their gene-expression patterns at the flip of a switch, making their careful isolation of utmost importance, Glass told Alzforum. He would know this as well as anyone, as he recently described how microglial transcriptomes shift rapidly in response to a change of scene, such as being plated into a culture dish (Jul 2016 conference news; Jun 2017 news). At Keystone, the Goate and Glass labs hatched plans to join forces in collecting microglial samples to run eQTL analyses.
In the meantime, Glass used a different approach to address the problem. He teamed up with UCSD colleague Bing Ren, who pioneered a technique called PLAC-seq (proximity ligation-assisted CHIP-seq) to link genetic regulatory regions with the genes they control (Fang et al., 2016). In a nutshell, the researchers sorted microglia from postmortem samples, isolated their nuclei, cut the DNA inside them with restriction enzymes, then performed a ligation reaction to fuse fragments of DNA that were stuck together, i.e., that were presumably in the midst of a regulatory liaison. The researchers then sequenced the DNA couples to discern which gene a given enhancer was controlling. PLAC-seq is not the only technique to decipher long-range chromatin connections—others include Hi-C and CHIA-PET—but Ren designed PLAC-seq to improve sensitivity and detect longer-range connections than other techniques; this made it suitable for rare cells like microglia.
At Keystone, Glass reported that in microglia sorted from postmortem samples, PLAC-seq captured more than 44,000 DNA interactions, 70 percent of which contained promoter sequences. One example was Sall1. Glass had previously reported that this gene’s expression became drastically downregulated when microglia were moved from the brain to culture. Using other chromatin activity mapping methods, Glass had proposed a region some 400kb away from the gene as an enhancer of its expression. PLAC-seq confirmed this suspicion, as this enhancer region was found associated with the SALL1 promoter. This served as confirmation that the technique works.
What about AD GWAS hits in noncoding regulatory regions? In Glass’ lab, postdoc Inge Holtman and colleagues next used PLAC-seq to ask which genes, if any, were connected by polymorphism-containing regulatory regions. They focused on newly identified AD risk loci from a recently expanded GWAS by the International Genomics of Alzheimer’s Project (IGAP) (Apr 2018 news on Kunkle et al., 2018). Specifically, Glass reported that one significant polymorphism near the BIN1 gene indeed connected strongly with the BIN1 promoter. This corroborated other chromatin analyses, such as ATAC-seq, which suggested the region was an active enhancer in microglia.
Glass also used PLAC-seq to reveal that yet another GWAS hit, nearest the ECHD3 gene, did not primarily associate with ECHD3 but rather with a different nearby gene, USP6NL. Notably, while no function is known for ECHD3, USP6NL has a role in endocytosis. Glass said he plans to apply the PLAC-seq approach to more fully investigate other GWAS risk loci.
Glass also reported results of his lab’s efforts to further determine which cells harbor the disease risk associated with each GWAS hit. Lab member Alexi Nott sorted neuronal (NeuN+) or microglial (PU.1+) nuclei from freshly frozen postmortem brain tissue, and performed an ATAC-seq analysis to identify active enhancer regions. In the case of BIN1, Nott found that the enhancer region containing the risk SNP was active in nuclei from microglia, but not neurons. This indicated that the Alzheimer’s-associated variant likely controls BIN1 expression in microglia, Glass said.
During questions following his talk, Glass told the audience that he plans to compare neuronal versus microglial epigenetic landscapes in cells derived from people with neurodegenerative disease with those without. He will also compare enhancer activation in iPSC-derived neurons and microglia, he later told Alzforum. Oleg Butovsky of Brigham and Women’s Hospital in Boston asked whether Glass could isolate nuclei specifically from microglia that surround plaques. Glass said that currently, there are no reliable markers that would distinguish nuclei from distant and plaque-associated microglia.
In a data-packed talk, Philip De Jager of Columbia University in New York described his lab’s efforts to link microglial gene expression profiles to disease phenotypes. De Jager aims to identify networks of co-expressed genes that could drive disease, and pick out biological pathways ripe for therapeutic targeting. He uses postmortem brain samples from the Religious Orders Study and the Rush Memory and Aging Project (ROSMAP). To date, ROSMAP has amassed more than 600 samples. At Keystone, De Jager reported that for 225 of them, the researchers took stock of the proportion of activated microglia (PAM) as per the cells’ morphological characteristics. They found that in cortical, but not subcortical, regions, PAM correlated with AD phenotypes, such as Aβ plaque burden, tau tangle pathology, and rate of cognitive decline. Using a statistical analysis to estimate causal relationships between these factors, De Jager proposed that together, activated microglia and Aβ pathology lead to tau pathology, which leads to cognitive decline. The findings are posted on bioRχiv (Felsky et al., 2018).
Based on RNA-sequencing data from 541 samples of prefrontal cortices, De Jager also identified modules of co-expressed microglial genes that correlated with specific AD features. Of primary interest, De Jager reported at Keystone and on bioRχiv, was that expression of genes in one module, Module 5, correlated with the extent of tau pathology (Patrick et al., 2017). The module contained genes involved in activation of the immune response, and looking back at the PAM analysis, he also found that this same module tracked with how many microglia were activated. Overall, De Jager concluded that these findings strongly implicated microglial activation as a pivotal step in the pathogenesis of AD.
While most of De Jager’s analyses used unsorted tissue samples to construct genetic networks, he also presented the fruits of initial efforts to investigate individual microglia (Olah et al., 2018). The researchers sorted live microglia from seven fresh autopsy samples of dorsolateral prefrontal cortices, as well as two fresh samples of hippocampi and six samples of temporal neocortices from epilepsy surgeries. In all, the researchers sequenced RNA from nearly 16,000 individual cells, 97 percent of which were microglia. They found significant levels of more than 1,200 RNA transcripts, and used them to cluster microglia into subtypes based on their gene-expression patterns (see image above).
This generated 14 unique clusters of microglia. A majority of microglia across samples fell into clusters 1 or 2; the other 12 clusters were more variable, and some were only found in samples from a couple of people. This suggests that some of these microglial clusters might be specific to certain diseases or functions. While the researchers had too few samples to correlate microglial clusters with disease phenotypes, they did find that some clusters were enriched for AD risk factors.
When the scientists compared human microglial clusters to clusters identified in mouse studies, they found the human repertoire to be more nuanced. For example, regarding the disease-associated microglia (DAMs) reported by Ido Amit of the Weizmann Institute in Israel, De Jager found that disease-associated genes were spread out among multiple clusters of human microglia (Jun 2017 news). However, he did find that expression levels of P2YR12 and ApoE were diametrically opposed in the human cells, as reported in mouse studies (Sep 2017 news).
De Jager also uncovered commonalities with disease-stage-specific clusters of microglia reported by Li-Huei Tsai at the Massachusetts Institute of Technology in the p25 mouse model of inducible AD. Specifically, human microglial cluster 10 expressed genes involved in cell proliferation, as did a cluster of microglia that appeared early in disease in Tsai’s mouse model. In addition, a group of microglia expressing interferon response genes—found at a late stage of the AD mouse model—also appeared in a subset of the human samples.
A growing number of scientists believe that microglial research has to be done using human cells. Tsai jumped the mouse ship, as well. When asked how the different types of mouse microglia identified in her studies compared with those of others, she replied that at the end of the day, what matters is characterizing microglia in the human brain. To that end, her group performed single-cell RNA sequencing on a subset of ROSMAP samples—24 from controls, and 24 from people with high Aβ burden. Unlike De Jager, Tsai did not sort out microglia first, but rather performed a co-expression cluster analysis on nuclei derived from all cell types in each sample. This was by design, Tsai told Alzforum, as her goal was to investigate how all cell types in the brain changed in the face of disease. In all, she identified 20 distinct expression clusters among cells, assigning each to a specific cell type based on the complement of genes. Tsai reported that the number of clusters identified for each cell type roughly correlated with the abundance of that cell type in the brain. For example, she identified nine expression clusters among excitatory neurons, five clusters in inhibitory interneurons, and two clusters among oligodendrocytes, which made up an average of 50, 12.6, and 26 percent, respectively, of cells in the 48 samples. Only one cluster for astrocytes and one for microglia, which made up 5 and 2.8 percent of cells in the brain, respectively, emerged from the analysis, Tsai reported.
Tsai correlated expression clusters of excitatory neurons and oligodendrocytes with AD phenotypes. One cluster of each—ex1 and oli1—were highly represented in people with low or no Aβ burden, while another cluster of each cell type—ex2 and oli2—correlated strongly with high Aβ burden. Relative to ex1, ex2 highly expressed genes for synaptic signaling and excitability, and poorly expressed genes for mitochondrial function. For the oligodendrocyte clusters, oli2 cells expressed higher levels of genes involved in protein folding, stability, and translation, and lower levels of intracellular trafficking genes than did oli1. Notably, Tsai also found that across all samples, men had higher numbers of the “good” cells (ex1 and oli1) than women, while the women had an overrepresentation of “bad” ones (ex2 and oli2). Tsai wondered whether the proportions of these cells could somehow reflect a greater susceptibility to AD among women.
Tsai told Alzforum that she plans to more closely examine how different gene-expression clusters of interneurons—the third most abundant cell type that Tsai identified after excitatory neurons and oligodendrocytes—correlate with disease. To look specifically at microglia or astrocytes, different techniques, such as presorting the cells, might need to be used. Glass, who opts to sort out microglia prior to analysis, told Alzforum that the field is still on the steep part of the learning curve when it comes to single-cell transcriptomics and other gene-expression analyses, especially when used on sparse cells like human microglia. Despite the limitations of working with postmortem samples, the only way forward is to keep trying, he said.—Jessica Shugart
- Microglial Master Regulator Tunes AD Risk Gene Expression, Age of Onset
- When a Microglia Is No Longer a Microglia
- What Makes a Microglia? Tales from the Transcriptome
- Hot DAM: Specific Microglia Engulf Plaques
- ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
- Fang R, Yu M, Li G, Chee S, Liu T, Schmitt AD, Ren B. Mapping of long-range chromatin interactions by proximity ligation-assisted ChIP-seq. Cell Res. 2016 Dec;26(12):1345-1348. Epub 2016 Nov 25 PubMed.
- Kunkle BW, Grenier-Boley B, Sims R, Bis JC, Naj AC, Boland A, Vronskaya M, van der Lee SJ, Amlie-Wolf A, Bellenguez C, Frizatti A, Chouraki V, Alzheimer's Disease Genetics Consortium (ADGC), European Alzheimer's Disease Initiative (EADI), Cohorts for Heart and Aging Research in Genomic Epidemiology Consortium (CHARGE), Genetic and Environmental Risk in Alzheimer's Disease Consortium (GERAD/PERADES), Schmidt H, Hakonarson H, Munger R, Schmidt, Farrer LA, Broeckhoven CV, O'Donovan MC, Destefano AL, Jones L, Haines JL, Deleuze JF, Owen MJ, Gudnason V, Mayeux RP, Escott-Price V, Psaty BM, Ruiz A, Ramirez A, Wang LS, van Duijn CM, Holmans PA, Seshadri S, Williams J, Amouyel P, Schellenberg GD, Lambert JC, Pericak-Vance MA. Meta-analysis of genetic association with diagnosed Alzheimer's disease identifies novel risk loci and implicates Abeta, Tau, immunity and lipid processing. bioRχiv 294629; April 5, 2018
- Felsky D, Roostaei T, Nho K, Risacher SL, Bradshaw EM, Petyuk V, Schneider JA, Saykin A, Bennett DA, De Jager PL. Neuropathological correlates and genetic architecture of microglial activation in elderly human brain. Nat Commun. 2019 Jan 24;10(1):409. PubMed.
- Patrick E, Olah M, Taga M, Klein HU, Xu J, White CC, Felsky D, Gaiteri C, Chibnik LB, Mostafavi S, Schneider JA, Bennett DA, Bradshaw EM, De Jager PL. A cortical immune network map identifies a subset of human microglia involved in Tau pathology. bioRχiv. December 14, 2017
- Olah M, Menon V, Habib N, Taga M, Yung C, Cimpean M, Khairalla A, Dionne D, Hopp S, Frosch MP, Hyman BT, Beach TG, Sarkis R, Cosgrove GR, Helgager J, Golden JA, Pennell PB, Schneider JA, Bennett DA, Regev A, Elyaman W, Bradshaw EM, De Jager PL. A single cell-based atlas of human microglial states reveals associations with neurological disorders and histopathological features of the aging brain. bioRχiv. June 11, 2018
No Available Further Reading
TREM2: Diehard Microglial Supporter, Consequences Be DAMed
Like a dedicated mother doling out heaps of unconditional love, TREM2 supports microglial cells in all their pursuits, whether they help or hurt the brain. Trash clean-up, Aβ plaque containment, calling in blood cells for reinforcement in ALS, or gorging on tau-stricken neurons—those are but a few of the many TREM2-supported microglial functions researchers described at the joint Keystone symposia Advances in Neurodegenerative Disease Research and Therapy/New Frontiers in Neuroinflammation, held June 17–21 in Keystone, Colorado.
The moment TREM2 was pinned as an AD risk factor (Nov 2012 news), researchers started hustling to understand how this receptor, which is expressed primarily on microglia in the brain, exerts its sway. The discovery that variants linked to neurodegenerative disease appear to sap the protein’s function strongly suggests that TREM2 plays a broadly beneficial role, either in staving off disease onset or slowing progression. However, animal models of disease have painted a complex picture, revealing that TREM2 can evoke both helpful and harmful responses from microglia. Which type of response predominates varies across both disease models and disease stages.
Initial studies, which focused on TREM2 in microglial clearance of Aβ plaques, showed that the receptor’s role could be dramatically different from one mouse model to the next, and now scientists are seeing equally variable behavior among tauopathy models. At Keystone, Cheryl Leyns from David Holtzman’s lab at Washington University in St. Louis presented a poster implicating TREM2 in exacerbating neurodegeneration in the PS19 tauopathy model. Published last year, her data blamed TREM2 for inciting harmful microgliosis and neuroinflammation in these mice (Leyns et al., 2017). However, Bruce Lamb’s lab at Indiana University School of Medicine in Indianapolis reached the opposite conclusion using hTau mice, in which tauopathy progresses more slowly. In those animals, TREM2 reduced accumulation of hyperphosphorylated tau and prevented neuroinflammation (Bemiller et al., 2017; Trem2 KO (KOMP) x htau research model).
TREM2 discoverer Marco Colonna of Washington University in St. Louis, who collaborated with Leyns and Holtzman, told Alzforum that together, the findings support the idea that microglia respond in proportion to the aggressiveness of threats they encounter (Oct 2017 news). In severe disease models, such as PS19, microglia react strongly and may harm neurons. In milder models, such as hTau, microglial meddling may prove beneficial. Either way, TREM2 supports microglia in their endeavors, Colonna said.
Harking back to TREM2’s role in Aβ plaque formation and clearance, Christian Haass of the German Center for Neurodegenerative Diseases in Munich presented data on the earliest stages of Aβ accumulation. To tightly control the onset of deposition, his lab adapted a method developed by Melanie Meyer-Luehmann of the University of Freiburg in Germany. In it, extracts from plaque-laden APP/PS1 mouse brain were injected into the hippocampi of three lines of mice: wild-type, TREM2 knockout, and those expressing T66M loss-of-function TREM2. Six weeks later, nascent plaques appeared in all the mice, but were more abundant and larger in TREM2 knockout or T66M mice. Furthermore, microglia in mice lacking functional TREM2 did not crowd around plaques, and they expressed less of the lysosomal activity marker CD68. But while TREM2 appeared to stave off the initial seeding of plaques, it only did so for a while. By 12 months of age, the plaque burden in wild-type mice caught up to that in mice lacking functional TREM2.
The findings suggested that TREM2 helps microglia slow the seeding of new plaques, but how? Haass reported that plaques in seeded wild-type mice contained about double the ApoE as those from seeded TREM2 KO or T66M-TREM2 mice. In line with what previous studies have reported, Haass also found that TREM2 facilitated increased ApoE expression in microglia in seeded wild-types, but not in animals lacking functional TREM2. He proposed that microglia clustering around plaques might infuse ApoE directly into the deposits. In support of this idea, Haass found that depleting microglia from the mouse brain dramatically reduced the amount of ApoE in plaques. Perhaps ApoE aids in the compaction of plaques, which could explain the striking observation made by several labs that plaques in TREM2-deficient mice are larger and more diffuse. Such "fluffy" deposits might be more apt to leach toxic Aβ oligomers than highly compacted versions, some researchers at Keystone speculated.
Holtzman collaborated with Haass on this work. He favors the idea that another cell, astrocytes—the prime source of ApoE in the brain—are the major contributors of plaque-associated ApoE. Astrocytic ApoE likely facilitates the earliest stages of plaque seeding, which occurs long before microglia arrive at the scene and start pumping out ApoE themselves, Holtzman said. Colonna commented that while it is possible microglial-derived ApoE enters plaques and helps compact them, other TREM2-mediated microglial functions, such as barrier formation, might achieve a similar effect (May 2016 news). Either way, the data strengthen the prevailing view that TREM2 helps microglia limit Aβ plaques, he said.
In ALS, TREM2 Helps Microglia Call for Backup
In his talk, Oleg Butovsky, Brigham and Women’s Hospital in Boston, set the stage for a flash of provocative new data. In recent years, Butovsky uncovered a homeostatic gene-expression signature in microglia under healthy conditions. In the face of neurodegeneration, TREM2 helps transform homeostatic microglia into a so-called microglia neurodegenerative phenotype, a.k.a. MGnD. These activated cells upregulate expression of ApoE, among other genes (Feb 2015 news; Sep 2017 news). Whether this switch is helpful or harmful likely depends on the type of pathology, as well as disease stage, Butovsky contends.
At Keystone, he reported that, in the SOD1-G93A model of amyotrophic lateral sclerosis, TREM2 plays a beneficial role, though in a surprising way. In SOD1-G93A mice, Butovsky spotted numerous cells expressing CD169—a myeloid cell marker not known to be expressed on cells in the brain—in the spinal cord at later stages of disease. These cells seemed distinct from those expressing the microglial marker TMEM119. In TREM2-deficient animals, CD169+ cells did not infiltrate the spinal cord, and these mice died 70 days earlier than their TREM2-replete counterparts. Butovsky proposed that the CD169+ cells are peripheral macrophages that somehow stave off neurodegeneration.
Butovsky suggested that in the SOD1-G93A ALS model, TREM2 somehow bestows MGnDs with the ability to cry out for help—in other words, to secrete cytokines that recruit peripheral macrophages. Strikingly, the presence of these “infiltrating” cells correlated with longer survival. The data are still preliminary, Butovsky said. If true, they would suggest that recruitment of peripheral cells is a critical function of late-stage MGnD.
Similar phenomena could play out in neurodegenerative diseases besides ALS, such as AD, Butovsky said. At first glance, this might seem to fly in the face of findings from parabiosis experiments led by Holtzman and Colonna, in which peripheral cells did not infiltrate the brain in AD mouse models of amyloidosis (May 2016 news on Wang et al., 2016). However, Colonna acknowledged that while the results of those parabiosis experiments were crystal clear, they were also limited. For example, the researchers could only keep animals conjoined for two months, leaving open the possibility that peripheral cells infiltrate later in disease. Colonna has not tried parabiosis in SOD1-G93A mice.
When it comes to subtypes of microglia found in brains stricken with neurodegeneration, MGnDs aren’t the only game in town. Consider the disease-associated microglia, or DAMs, reported by Michal Schwartz and Ido Amit of the Weizmann Institute in Israel, as well as several other subtypes described by Li-Huei Tsai of Massachusetts Institute of Technology (Jun 2017 news on Keren-Shaul et al., 2017; Oct 2017 news on Mathys et al., 2017). At Keystone, both Amit and Tsai reviewed published findings on this issue. Both researchers used single-cell RNA sequencing to monitor transcriptionally distinct clusters of microglia as their gene-expression patterns shifted during disease progression. Overall, Amit and Tsai drew roughly similar conclusions as Butovsky in regard to gene-expression changes incited by neurodegeneration, such as microglial upregulation of ApoE, but they also reported that these changes occurred in distinct temporal waves as disease progressed.
Notably, in the case of DAMs, the transformation from homeostasis to disease-associated phenotype occurred in two stages, only the second of which was TREM2-dependent. Collaborating with Colonna, Amit reported that in TREM2 KO 5xFAD mice, microglia did transform into Stage I DAMs—characterized by reduced expression of homeostatic genes and upregulation of ApoE. But from there, the cells did not progress further into Stage II DAMs, which additionally express genes involved in phagocytosis. If TREM2 doesn’t kick off the whole DAM cascade, what does? Fielding this question from the audience, Amit hinted that he was closing in on a culprit, but shared no more. Researchers at Keystone also left the familiar territory of mouse models behind to study human microglia, finding a more nuanced repertoire of cells in the human brain (see Part 1 of this series).
In his talk, Colonna interpreted the single-cell RNA-Seq findings as evidence that TREM2’s key role lies not in initiating the switch of microglia into DAMs, but rather in supporting the cells along their path of transformation. Analogous, perhaps, to the role of co-stimulatory receptors on T cells, TREM2 signals might muster a microglia’s metabolic strength required for energy-taxing functions—including proliferation, migration, and phagocytosis—rather than instigating the transformation in the first place. As to the question of what flips the initial switch, Colonna offered up a host of possibilities, including toll-like, Nod-like, and scavenger receptors, all of which sense and respond to threats.
Last but not least, TREM2 is also a receptor in its own right. Its long list of ligands includes ApoE, Aβ, and lipids (Wang et al., 2015; Mar 2018 news). At Keystone, Colonna surveyed recently published findings implicating soluble TREM2—the extracellular portion of TREM2 shed by ADAM proteases—as a ligand for neurons. In 5xFAD mice expressing human TREM2, he found sTREM2 blanketing the cells in the vicinity of Aβ plaques. Surprisingly, however, sTREM2 coated neurons whether or not they appeared healthy. This contradicts the prevailing view in the field, namely that, at least for full-length TREM2, the receptor specifically latches on to plasma membranes of apoptotic cells. Colonna reported that sTREM2’s coating of neurons was greatly reduced in mice expressing the R47H AD risk variant of TREM2 (Jan 2018 news).
Because sTREM2 latched onto healthy neurons, Colonna proposed that sTREM2 might serve as a sensor of lipid density more broadly. Around plaques, lipid density would be quite high, as lipids from plaques, healthy and dying neurons, and glial cells comingle. Exactly how sTREM2’s binding to these elements steers microglial responses is unclear, but Colonna intends to find out. In keeping with this general theme of lipid sensing, a new study reported that AD risk variants near the gene coding the fatty acid sensor family MS4A correlate with levels of sTREM2 in the CSF (Jul 2018 news).—Jessica Shugart
Mutation Interactive Images Citations
- Enter the New Alzheimer’s Gene: TREM2 Variant Triples Risk
- Changing With the Times: Disease Stage Alters TREM2 Effect on Tau
- Barrier Function: TREM2 Helps Microglia to Compact Amyloid Plaques
- Microglia in Disease: Innocent Bystanders, or Agents of Destruction?
- ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
- Hot DAM: Specific Microglia Engulf Plaques
- A Delicate Frontier: Human Microglia Focus of Attention at Keystone
- TREM2 Binds Aβ, Reprograms Microglia to Curb Plaques
- New Mouse Models Reveal Unexpected Property of TREM2
- MS4A Alzheimer’s Risk Gene Linked to TREM2 Signaling
Research Models Citations
- Leyns CE, Ulrich JD, Finn MB, Stewart FR, Koscal LJ, Remolina Serrano J, Robinson GO, Anderson E, Colonna M, Holtzman DM. TREM2 deficiency attenuates neuroinflammation and protects against neurodegeneration in a mouse model of tauopathy. Proc Natl Acad Sci U S A. 2017 Oct 24;114(43):11524-11529. Epub 2017 Oct 9 PubMed.
- Bemiller SM, McCray TJ, Allan K, Formica SV, Xu G, Wilson G, Kokiko-Cochran ON, Crish SD, Lasagna-Reeves CA, Ransohoff RM, Landreth GE, Lamb BT. TREM2 deficiency exacerbates tau pathology through dysregulated kinase signaling in a mouse model of tauopathy. Mol Neurodegener. 2017 Oct 16;12(1):74. PubMed.
- Wang Y, Ulland TK, Ulrich JD, Song W, Tzaferis JA, Hole JT, Yuan P, Mahan TE, Shi Y, Gilfillan S, Cella M, Grutzendler J, DeMattos RB, Cirrito JR, Holtzman DM, Colonna M. TREM2-mediated early microglial response limits diffusion and toxicity of amyloid plaques. J Exp Med. 2016 May 2;213(5):667-75. Epub 2016 Apr 18 PubMed.
- Keren-Shaul H, Spinrad A, Weiner A, Matcovitch-Natan O, Dvir-Szternfeld R, Ulland TK, David E, Baruch K, Lara-Astaiso D, Toth B, Itzkovitz S, Colonna M, Schwartz M, Amit I. A Unique Microglia Type Associated with Restricting Development of Alzheimer's Disease. Cell. 2017 Jun 15;169(7):1276-1290.e17. Epub 2017 Jun 8 PubMed.
- Mathys H, Adaikkan C, Gao F, Young JZ, Manet E, Hemberg M, De Jager PL, Ransohoff RM, Regev A, Tsai LH. Temporal Tracking of Microglia Activation in Neurodegeneration at Single-Cell Resolution. Cell Rep. 2017 Oct 10;21(2):366-380. PubMed.
- Wang Y, Cella M, Mallinson K, Ulrich JD, Young KL, Robinette ML, Gilfillan S, Krishnan GM, Sudhakar S, Zinselmeyer BH, Holtzman DM, Cirrito JR, Colonna M. TREM2 lipid sensing sustains the microglial response in an Alzheimer's disease model. Cell. 2015 Mar 12;160(6):1061-71. Epub 2015 Feb 26 PubMed.
ApoE Has Hand in Alzheimer’s Beyond Aβ, Beyond the Brain
As the strongest genetic risk factor for Alzheimer’s disease, ApoE, and particularly its relationship with Aβ, has been studied six ways to Sunday. More recently, scientists have turned to what the apolipoprotein does beyond Aβ. At a joint Keystone symposia—Advances in Neurodegenerative Disease Research and Therapy; and New Frontiers in Neuroinflammation—held June 17–21 in Keystone, Colorado, researchers implicated ApoE in harmful glial responses to tau pathology. In tau models, the E4 allele drove up expression of genes involved in cholesterol biosynthesis in microglia and astrocytes, while at the same time lowering expression of genes needed to export cholesterol from those cells. Other researchers reported that E4 wrought havoc in the brain even when expressed only in the periphery. In all, the meeting revealed that despite a quarter century in the research limelight, ApoE still has some secrets to give up.
In a talk, David Holtzman of Washington University in St. Louis followed up on his published data that human ApoE exacerbates tau pathology in mice (Apr 2017 conference news; Sep 2017 news). In that study, Holtzman and collaborators showed that, in the P301S tauopathy model, human ApoE intensified tau accumulation, neuronal damage, and neuroinflammation, with the E4 allele being more potent than E3 or E2. Conversely, P301S mice on an ApoE knockout background aged without neurodegeneration.
At Keystone, Holtzman attributed these effects to the secreted form of ApoE. Besides existing in different lipidation states, ApoE can also reside in different locales: Most is secreted, some is retained inside the cell. To understand the effects of extracellular ApoE, Holtzman crossed P301S mice to a strain overexpressing low-density lipoprotein receptor (LDLR), which binds to and internalizes ApoE. Holtzman had previously reported that overexpressing this receptor 10-fold in normal mice removed 90 percent of the brain’s extracellular ApoE (Dec 2009 news). The same was true when those LDLR-overexpressing animals were crossed with P301S mice, he reported. Notably, the LDLR-OE P301S animals fared similarly to P301S mice on an ApoE knockout background: They had less phospho-tau accumulation and no neurodegeneration. This suggests that it is extracellular ApoE that worsens tau pathology.
Holtzman also reported data in keeping with the idea that extracellular ApoE incites microglia. According to single-cell transcriptomics on P301S brains, clusters of cells expressing microglial activation markers expanded dramatically when the animals were between six and nine months of age. This did not happen in P301S ApoE knockouts or in P301S mice that overexpressed LDLR, suggesting again that secreted ApoE was required to rile up microglia. Overall, Holtzman interpreted his findings as more evidence that ApoE aids and abets neurodegeneration, once again casting ApoE as a therapeutic target (Apr 2018 news).
Where did the extracellular ApoE come from? One possibility is the microglia themselves. While astrocytes are generally considered the primary source of ApoE, Holtzman reported a 100-fold spike in microglial ApoE expression in six- to nine-month-old P301S mice. In astrocytes, ApoE expression only doubled or tripled during this time period. However, Holtzman noted that astrocytes are far more abundant than microglia, and express much higher levels of the apolipoprotein in earlier disease stages than do microglia. He also cautioned that while ApoE transcripts eventually skyrocketed in microglia, to what extent that translated into ApoE protein was uncertain. In all, Holtzman still favors the idea that astrocytes are the prime source of ApoE in the brain throughout the course of disease.
There was also much chatter about ApoE’s lipidation status at Keystone. In his talk, Guojun Bu of the Mayo Clinic in Jacksonville, Florida, reported that the amount of lipids strapped to the protein varies depending on which cell type secretes it. Bu extracted astrocytes or microglia from the brains of young mice, cultured them separately, and compared the size of the ApoE particles they spat into the medium. He found that both microglia and astrocytes secreted lipidated particles, but that microglial ones were larger. Bu cautioned that the process of extracting microglia from the brain and placing them in culture is known to activate the cells, and it is unclear how closely the cultured cells resemble those in the brain (see also Part 1 of this series). Bu said he is performing a thorough lipidomics analysis of ApoE particles, comparing the lipid content of ones secreted from different cell types, and of different ApoE isoforms. In his talk, he proposed that ApoE2 particles were larger and more lipidated than ApoE4 particles, suggesting that ApoE2 might be more adept than ApoE4 at removing cholesterol from circulation, and possibly help repair axonal injuries more effectively as well.
Holtzman was skeptical that ApoE lipidation differed depending on isoform, pointing out that particles derived from human CSF look similar across genotypes. However, he was intrigued by the idea that different cell types might produce differentially lipidated particles. “The possibility that a microglia particle is different than an astrocyte particle is very high,” he said. “But it remains to be seen what those differences are, and how they matter.”
ApoE: From the Outside In?
Bu’s lab is venturing beyond the borders of the brain to investigate the effects of ApoE expression in the periphery. Unlike the AD risk factor TREM2—a receptor expressed primarily on microglia—ApoE is made throughout the body, especially in the liver. Peripheral ApoE is excluded from the blood-brain barrier, so it cannot efficiently pass into the brain. Nevertheless, Bu hypothesized that the peripheral protein might somehow influence processes in the brain.
To investigate this idea, Bu generated mice expressing either ApoE3 or ApoE4 driven by the albumin promoter to restrict expression to hepatocytes in the liver. The animals were on an ApoE knockout background and, in keeping with the apolipoprotein’s exclusion by the blood-brain barrier, Bu detected no ApoE in the brain. However, cognitive differences emerged, as albumin-E4 mice performed worse on tests of contextual fear memory than ApoE knockouts, while albumin-E3 mice performed better. When Bu crossed these animals to APP/PS1 mice, he found that compared to the ApoE knockouts, which had only diffuse plaques, albumin-E3 mice had fewer diffuse plaques, while albumin-E4 mice had more.
How could peripheral ApoE possibly influence these processes in the brain? Bu believes that peripherally expressed E4 somehow compromises the integrity of the blood-brain barrier. More dextran leaked across it in the albumin-E4 mice than in the other mice, and blood flowed more slowly through their brain arterioles, he reported at Keystone. How these vascular problems influenced myriad processes in the brain remains to be seen, though Bu speculates that neuroinflammatory responses triggered by injured vessels might play a role. The findings also exemplify the beneficial role of ApoE3, Bu told Alzforum, and argue against the idea of targeting ApoE in people who do not carry the E4 allele.
Could ApoE have crossed the barrier and seeded plaques? Bu said that so far, he has not found any ApoE in these animals’ brains. However, he did not rule out the possibility that a small amount did get in.
Some of Bu’s findings mesh with those of a previous study led by Joachim Herz of the University of Texas Southwestern Medical Center in Dallas, who reported that expression of human ApoE3 in the periphery prevented memory problems in ApoE knockout mice (Oct 2016 news). However, peripheral ApoE3 did not rescue some synaptic deficits the knockouts had. Herz also detected small amounts of human ApoE3 in the interstitial fluid, suggesting that some level of peripheral ApoE sneaks into the brain.
Bu told the audience that while he still views ApoE’s relationship with Aβ as a pivotal driver of AD risk, he thinks ApoE’s functions in multiple cell types and organs together influence the onset and course of the disease.
ApoE4: Oil Slick in a Dish?
Other researchers at Keystone presented data on ApoE’s specific effects on different cell types. In particular, several groups investigated how different ApoE isoforms swayed gene expression in iPSC-derived neurons, astrocytes, and microglia. As presented in a talk and on a poster, Alison Goate and postdoc Julia TCW of the Icahn School of Medicine at Mount Sinai, New York, generated iPSC-derived neurons, astrocytes, microglia, and microvesicular endothelial cells from 13 donors, including seven homozygous E3 carriers and six homozygous E4 carriers. They also used CRISPR gene editing to create isogenic lines from several of the donors, changing E3/E3 to E4/E4 and vice versa. At Keystone, Goate and TCW presented some of the fruits of this massive cell culture effort, reporting that compared with E3/E3 cells, E4/E4 microglia and astrocytes had dramatically elevated expression of genes involved in cholesterol biosynthesis, but reduced expression of lysosomal processing and lipid efflux genes. In other words, cholesterol synthesis was up, while the machinery needed to break it down and get rid of it was down. TCW found no elevation of cholesterol biosynthesis in iPSC-derived neurons, indicating that the pathway was only affected in glial cells.
Goate interpreted this as evidence of an uncoupling of lipid production and degradation pathways in E4 microglia and astrocytes. The findings mesh with Goate’s broader hypothesis about what drives AD, namely a defect in efferocytosis, a.k.a. the removal of dead cells and debris. Goate’s and others’ GWAS findings (see Part 1 of this series) implicate failing microglial functions such as phagocytosis in AD, and E4 appears to compound these problems by causing a lipid pileup inside cells.
John Hardy of University College London put forth a similar overarching hypothesis in his talk at Keystone. He suggested central processes that drive each neurodegenerative disease. For AD, he proposed faulty microglial clean-up of damaged membranes as a pivotal vulnerability, pointing out that many AD risk factors—such as TREM2, ABCA7, MS4A genes, and ApoE—sense and bind to lipids. Hardy added that even in the case of autosomal-dominant AD mutations in PS1, PS2, or APP, increased production of Aβ damages neuronal membranes, and ultimately microglia cannot keep up with the demand for clean-up.
In their talks at Keystone, Yadong Huang of the Gladstone Institute of Neurological Disease in San Francisco and Li-Huei Tsai of the Massachusetts Institute of Technology in Boston also described effects of ApoE isoforms on iPSC-derived human cells. Both researchers recently published their findings. In a nutshell, Huang generated iPSC-derived neurons from three E3/E3 donors and three E4/E4 donors, reporting that ApoE4 boosted both Aβ production and tau phosphorylation. Tsai, who made iPSC-derived cells and isogenic lines from a single donor, reported the same. She also found, similar to Goate and TCW, that at least in astrocytes, ApoE4 elevated expression of lipid synthesis genes, but downregulated lipid transport genes. In microglia, E4 stoked inflammatory genes, and downregulated genes involved in cell migration (see Jun 2018 news for Alzforum coverage of both papers).
While some parallels exist between Goate and Tsai’s findings, TCW commented that samples from multiple donors are needed to generate sufficient statistical power to draw conclusions. She is currently generating even more isogenic lines, in hopes they will be a useful resource for drug discovery efforts in the field.
Exemplifying an unbiased approach to exploring the brain biology of ApoE, Tal Nuriel in Karen Duff’s lab at Columbia University, New York, presented multi-omics data implicating ApoE4 in a range of cellular processes. He extracted RNA, lipids, small molecules, and proteins from aged ApoE4 or ApoE3 transgenic mouse brain, and compared isoform-dependent differences across myriad processes in AD-vulnerable regions such as entorhinal cortex, to differences in AD-resistant regions such as visual cortex. Previously, the scientists had found dysregulation of the endolysosomal pathway and an uptick in neuronal hyperactivity in AD-vulnerable brain regions in ApoE4 mice (see Nuriel et al., 2017 and Nuriel et al., 2017).
At Keystone, Nuriel added metabolic data. Differences in small-molecule metabolites in the E3 versus E4 mice hinted at alterations in mitochondrial function. To investigate this with a functional assay, Nuriel compared the oxygen consumption rate—a measure of mitochondrial function—in different brain regions of 18-20-month old E3 and E4 mice. Relative to E3 mice, E4 mice had flagging mitochondria in the cortex and hippocampus. However, the E4 entorhinal cortex, known for its vulnerability to AD, consumed more oxygen than in E3 mice. It is unclear what this means, but Nuriel suspects that neurons in the entorhinal cortex might ramp up mitochondrial function to compensate for cell stress and/or hyperactivity triggered by ApoE4.
TCW commended the novelty and comprehensive nature of this work. Taken together, Nuriel’s, TCW’s, and other scientists’ Keystone presentations suggest that when it comes to ApoE in the brain, all cards are still on the table.——Jessica Shugart
- ApoE and Tau: Unholy Alliance Spawns Neurodegeneration
- ApoE4 Makes All Things Tau Worse, From Beginning to End
- Mind Over Heart—LDL Receptors Crimp ApoE, Aβ Accumulation
- Human ApoE Antibody Nips Mouse Amyloid in the Bud
- A Delicate Frontier: Human Microglia Focus of Attention at Keystone
- Distinct Roles for Brain and Blood ApoE in Neuron Health?
- In Human Neurons, ApoE4 Promotes Aβ Production and Tau Phosphorylation
- Nuriel T, Peng KY, Ashok A, Dillman AA, Figueroa HY, Apuzzo J, Ambat J, Levy E, Cookson MR, Mathews PM, Duff KE. The Endosomal-Lysosomal Pathway Is Dysregulated by APOE4 Expression in Vivo. Front Neurosci. 2017;11:702. Epub 2017 Dec 12 PubMed.
- Nuriel T, Angulo SL, Khan U, Ashok A, Chen Q, Figueroa HY, Emrani S, Liu L, Herman M, Barrett G, Savage V, Buitrago L, Cepeda-Prado E, Fung C, Goldberg E, Gross SS, Hussaini SA, Moreno H, Small SA, Duff KE. Neuronal hyperactivity due to loss of inhibitory tone in APOE4 mice lacking Alzheimer's disease-like pathology. Nat Commun. 2017 Nov 13;8(1):1464. PubMed.
No Available Further Reading
Synaptic Tau Clangs the Dinner Bell for Hungry Microglia
When tau strays into neuronal synapses, microglia see food. That is one possible interpretation from a handful of presentations at the joint Keystone symposia—Advances in Neurodegenerative Disease Research and Therapy / New Frontiers in Neuroinflammation—held June 17–21 in Keystone, Colorado. Researchers reported that neuronal hyperactivity drives tau from microtubules into presynaptic terminals, where it latches onto vesicles. Others saw tau congregate on both the pre- and postsynaptic sides, along with complement proteins that bait microglia. What’s worse, microglia not only engulf neurons afflicted by tau pathology, they may also help tau spread across the brain, claimed scientists. Collectively, the findings suggest that microglia gorge—for better or worse—on neurons harboring misplaced forms of tau.
Synaptic Tau: Which Side Are You On?
As a protein that stabilizes microtubules, tau primarily resides inside cells. However, because mounting evidence points to tau’s travels between neurons, researchers have started investigating how the cytosolic protein winds up beyond the boundaries of the plasma membrane. Previous studies have reported that neuronal activity stokes tau secretion (Feb 2014 news). At Keystone, meeting organizer Li Gan of the Gladstone Institute of Neurological Disease in San Francisco presented news from her investigation of how neuronal firing shifts tau’s locale inside the cell. Gan is in the process of moving her lab to Weill Cornell Medical College in New York.
Gan’s postdoc Tara Tracy investigated which proteins buddy up with tau in the cell, and whether these liaisons change when neurons fire. The researchers ran an APEX2 interaction screen in i3 neurons. Made to resemble cortical neurons, i3 neurons were generated from human induced pluripotent stem cells (iPSCs) using a protocol previously developed by Gan and colleagues (Wang et al., 2017; Fernanadopulle et al., 2018). In these proximity screens, proteins that touch or closely mingle with an APEX2-tagged protein, in this case tau, become biotinylated and can then be identified by mass spectrometry (for review of method, see Lam et al., 2015).
Appetite for Tau. Cultured microglia (red: Iba1, green: tau fibrils] ingest tau fibrils (yellow inside cell). [Courtesy of Marcus Chin, Gan lab.]
Tracy reported that under basal conditions, tau mingled primarily with microtubule and cytoskeletal proteins, as expected. But 30 minutes after stimulating the cells with a high concentration of potassium chloride to depolarize the membrane, tau had ditched these microtubule partners in favor of synaptic proteins, including multiple SNAREs, which facilitate the fusion of synaptic vesicles with the synaptic membrane. Of the 207 tau interactors the scientists identified, 52 associated with tau in stimulated neurons only. By zeroing in on the exact residues that touched tau from each protein, the researchers further concluded that tau hooked up with vesicular proteins from the cytosolic side, as opposed to the vesicle lumen. They concluded that when neurons are triggered, tau rapidly relocates into presynaptic terminals, where it associates with synaptic vesicles but does not enter them. Gan proposed that tau’s attachment to outgoing synaptic cargo could help tau exit the cell.
The findings mesh with a recent study led by Patrik Verstreken of KU Leuven in Belgium, which found that tau mislocalized to the presynapses and latched onto vesicles via an interaction with synaptogyrin-3. This essentially clumped the vesicles and prevented their efficient release (Feb 2018 news). Synaptogyrin-3 was among the proteins identified in Gan’s screen. At Keystone, Joseph McInnes of Baylor College of Medicine in Houston, first author on Verstreken’s study, told Alzforum that Gan’s findings beautifully demonstrated tau’s dramatic change of locale in the face of stimulation. Furthermore, McInnes proposed that KCl stimulation, a relatively harsh treatment that activates numerous kinases, might trigger tau phosphorylation. Thus, KCl stimulation could serve as a model for tauopathy.
Bearing news from the other side of the synapse, Morgan Sheng of Genentech in South San Francisco reported that in a mouse model of tauopathy, tau is in cahoots with postsynaptic proteins. Numerous studies have reported that tau’s movement from axons into the dendritic compartment, which is speckled with postsynaptic proteins and structures, bodes poorly for neurons. To explore tau’s movement into the postsynapse, Sheng analyzed the postsynaptic density (PSD) proteome in nine-month-old P301S mice. At this age, the mice have no overt neurodegeneration, but their neurons contain plenty of phospho-tau and show inklings of synaptic deficits.
A mass spec analysis of the proteins in the PSD revealed a striking amount of phospho-tau there. Sheng noted that due to the imperfect purification process, a fair amount of presynaptic proteins also popped up in the proteomic analysis, making it difficult to say for certain which side of the synapse phospho-tau came from. Similarly, immuno-gold labeling of P301S neurons revealed clusters of phospho-tau on both sides of the synapse.
What’s more, Sheng told the audience that the PSD preps contained complement C1q, a key component of the complement cascade. Sheng claimed that all three C1q subunits could be found in the PSDs of nine-month-old P301S mice. He noted that neurons do not produce C1q, and suggested it came from glia. Using super-resolution microscopy and immune-electron microscopy, Sheng found that the C1q sat on the extracellular side of the neuronal membrane, immediately adjacent to, but not within, the synapse. C1q appeared sandwiched between pre- and postsynaptic markers. Previous work from Beth Stevens and Cynthia Lemere’s labs have reported that overzealous pruning of complement-tagged synapses leads to synapse loss in AD mouse models (Aug 2013 conference news; Nov 2015 conference news). Now, Sheng’s work implicates tau pathology as an “eat-me” signal as well.
Finally, Sheng investigated whether microglia truly engulf synapses in P301S mice. He found that by nine, but not six, months of age, microglia contained both the pre- and postsynaptic proteins synapsin and PSD95, respectively. The amount of synaptic proteins microglia contained correlated with levels of phospho-tau. Treating the mice with an anti-C1q antibody prevented this synaptic engulfment and blocked synapse loss. In all, the findings support the idea that microglia munching on tau-laden synapses promotes synaptic loss.
Marco Colonna of Washington University in St. Louis told Alzforum that he found Sheng’s proteomics approach convincing. He speculated that when tau accumulates in the synapse, it could trigger changes in the conformation of membranes there. This by itself could signal microglia to pump out complement. Whether TREM2, the lipid-sensing microglial receptor that Colonna discovered, plays any role in that process is an open question, he said.
McInnes, whose work supports the idea that tau exerts its toxic effects from the presynapse, said Sheng’s approach could not definitively nail down tau’s whereabouts, nor the source of its synaptoxicity, to either the pre- or postsynaptic side. McInnes proposed that, rather than in response to a shift in membrane structure, microglia might douse synapses with complement in response to waning synaptic activity brought about by tau’s corralling of vesicles.
In his talk, Naruhiko Sahara of the National Institute of Radiological Sciences in Chiba, Japan, reported that microglia engulfed tau-laden neurons in rTg4510 mouse models of tauopathy. Previously, Sahara had used tau PET and TSPO PET imaging in these mice to track neurofibrillary pathology and microglial activation, respectively. Both rose together, but the latter continued unabated even after tau pathology had plateaued (Ishikawa et al., 2018). At Keystone, Sahara described the results of two-photon confocal microscopy through cranial windows in rTg4510 mice. He injected adeno-associated viruses (AAVs) equipped with fluorescent tags driven by microglial or neuron-specific promoters. This enabled him to visualize microglia and neurons, respectively. He also injected PBB3, a fluorogenic dye that labels tau tangles and serves as a tau PET tracer. Daily tracking of the interactions between these cells in five- to seven-month-old mice revealed that microglia not only nibbled on PBB3-labeled neurons, but ultimately killed them. Microglia also killed neurons without tau pathology, yet seemed to have a preference for those with it.
Finally, Sahara reported a dramatic increase in complement proteins C1q and C3 in the brains of mice as microglia reached their peak of neuron consumption. Sahara’s findings meshed with Sheng’s, as both implicated a complement-driven microglial response in the demise of neurons burdened with tau pathology.
In the latter part of her talk, Gan investigated yet another potential consequence of microglia’s appetite for tau: The cells might serve as unwitting transporters of tau across the brain. Others have reported that in addition to trans-synaptic travels between connected neurons, tau may spread via microglia (Oct 2015 news). To investigate, Chao Wang from Gan’s lab used a model developed by Virginia Lee at the University of Pennsylvania in Philadelphia, in which injection of tau fibrils into aged PS19 tauopathy mice triggers the propagation of tau pathology throughout the brain (Iba et al., 2013). Wang found that depleting microglia, using the CSF1R agonist PLX3397, cut tau propagation by half, suggesting the cells play a role in spreading tau.
Wang went on to detail how cultured primary microglia that take up tau fibrils enlist the transcription factor NFκB to switch on inflammatory genes. The researchers generated mice in which NFκB was constitutively active, or repressed, only in microglia. While all microglia gobbled up tau (see image above), those with active NFκB retained less of it than their NFκB-inactive counterparts.
Gan wondered what would happen to the tau fibrils ingested by microglia. The researchers crossed the NFκB /microglia animals to PS19 mice. Those that repressed microglial NFκB had far less spread of tau inclusions and they performed better on spatial memory tests. Mice with constituitively active NFκB in their microglia had more tau spread. Incidentally, forcing NFκB activation in microglia triggered cognitive deficits irrespective of transgenic tau, suggesting that overzealous microglia might harm healthy neurons, perhaps by vigorous synaptic pruning, Wang told Alzforum at his poster.
Overall, Gan proposed that activated microglia act not merely as trash cans for tau, but rather process and release it again. On his poster, Wang also reported that microglia acetylate tau before releasing it, which could imply that tau emerges from microglia in a more toxic form than when it went in. He is investigating this. Wang told Alzforum that his findings do not contradict the idea that tau spreads between neurons independently of microglia, but instead implicate activated microglia as one possible route of tau processing and dissemination.—Jessica Shugart
- Neurons Release Tau in Response to Excitation
- Tau Uses Synaptogyrin-3 to Clump Synaptic Vesicles
- Curbing Innate Immunity Boosts Synapses, Cognition
- Microglia Control Synapse Number in Multiple Disease States
- Deadly Delivery: Microglia May Traffic Tau Via Exosomes
Research Models Citations
- Wang C, Ward ME, Chen R, Liu K, Tracy TE, Chen X, Xie M, Sohn PD, Ludwig C, Meyer-Franke A, Karch CM, Ding S, Gan L. Scalable Production of iPSC-Derived Human Neurons to Identify Tau-Lowering Compounds by High-Content Screening. Stem Cell Reports. 2017 Oct 10;9(4):1221-1233. Epub 2017 Sep 28 PubMed.
- Fernandopulle MS, Prestil R, Grunseich C, Wang C, Gan L, Ward ME. Transcription Factor-Mediated Differentiation of Human iPSCs into Neurons. Curr Protoc Cell Biol. 2018 Jun;79(1):e51. Epub 2018 May 18 PubMed.
- Lam SS, Martell JD, Kamer KJ, Deerinck TJ, Ellisman MH, Mootha VK, Ting AY. Directed evolution of APEX2 for electron microscopy and proximity labeling. Nat Methods. 2015 Jan;12(1):51-4. Epub 2014 Nov 24 PubMed.
- Ishikawa A, Tokunaga M, Maeda J, Minamihisamatsu T, Shimojo M, Takuwa H, Ono M, Ni R, Hirano S, Kuwabara S, Ji B, Zhang MR, Aoki I, Suhara T, Higuchi M, Sahara N. In Vivo Visualization of Tau Accumulation, Microglial Activation, and Brain Atrophy in a Mouse Model of Tauopathy rTg4510. J Alzheimers Dis. 2018;61(3):1037-1052. PubMed.
- Iba M, Guo JL, McBride JD, Zhang B, Trojanowski JQ, Lee VM. Synthetic tau fibrils mediate transmission of neurofibrillary tangles in a transgenic mouse model of Alzheimer's-like tauopathy. J Neurosci. 2013 Jan 16;33(3):1024-37. PubMed.
No Available Further Reading
VCAM1: Gateway to the Aging Brain?
Remember those eerie research findings evoking “The Twilight Zone”? Mysterious ingredients in the blood of young mice rejuvenated the brains of their elders, while the blood of old mice sped up aging in young’uns. The identity of blood-borne agents of aging remains a mystery, but researchers believe they have at least zeroed in on the agents’ gateway to the brain. At the joint Keystone symposia Advances in Neurodegenerative Disease Research and Therapy / New Frontiers in Neuroinflammation, held June 17–21 in Keystone, Colorado, Tony Wyss-Coray of Stanford University in Palo Alto reported that expression of vascular cell adhesion molecule 1 (VCAM1) on brain endothelial cells was required for old blood to accelerate aging in young brains. In fact, blocking this receptor even slowed classic symptoms of normal brain aging, such as less neurogenesis and more neuroinflammation, in control mice that never received blood transfusions.
The findings cement the role of systemic factors in brain aging and suggest that blocking their lines of communication with the brain could have restorative effects. The therapeutic prospect is especially appealing because treatments targeting VCAM1 would not need to cross the blood-brain barrier, which remains a formidable hurdle for CNS drug development. The researchers uploaded their findings to bioRχiv, where the pdf has been viewed nearly 2,200 times since January (Yousef et al., 2018).
Aging is the strongest risk factor for Alzheimer’s and most other neurodegenerative diseases. Circulating factors play a key role, according to findings over the years by Wyss-Coray and colleagues. Not only did they find that blood from young mice halted synaptic deficits, loss of neurogenesis, and even memory loss in AD mouse models, but blood from old mice quickened these processes in young mice (Aug 2011 news; May 2014 conference news; Sep 2016 conference news).
At Keystone, Wyss-Coray presented new results of efforts to understand how outside factors manage to drive aging inside the brain. Reasoning that the blood-brain barrier must be an active participant, Wyss-Coray and colleagues sifted through plasma proteomics data the lab had generated in years prior. They searched for proteins whose concentration shot up with age, and which were known to be expressed by the brain endothelial cells (BECs) that line the BBB. Of the 31 proteins that went up with age, eight were expressed in BECs, and five of those were implicated in vascular function. Of those five, soluble VCAM1 stood out as the protein that increased most in the plasma with age, Wyss-Coray reported at Keystone.
VCAM1 Opens Door to Aging. In the young brain, low levels of VCAM1 on brain endothelial cells allow microglia to remain ramified and calm, and neurogenesis to proceed. In the old brain, these cells increase VCAM1 expression, which tethers more leukocytes to their luminal side. This activates microglia, suppresses neurogenesis. [Courtesy of Yousef et al., bioRχiv, 2018.]
VCAM1 is part of the immunoglobulin receptor family, whose expression rises on many cell types in response to injury. The receptor binds to the integrin VLA4, expressed on leukocytes. The soluble version of VCAM1—sVCAM1—is continually shed by the ADAM17 protease, and its concentration in the blood correlates with cognitive impairment and even mortality, Wyss-Coray said.
To investigate VCAM1 expression on BECs, the scientists infused fluorescently tagged anti-VCAM1 antibodies into mice; this labeled cells expressing the receptor on the luminal side of the cerebrovasculature. They found a patchy, sparse expression pattern. In young mice, at most 4 percent of BECs in hippocampal blood vessels expressed VCAM1, and this rose to about 5 percent as the animals aged. Injecting young mice with lipopolysaccharide boosted VCAM1 expression to 15 percent of BECs, and even more so in older mice. Single-cell RNA sequencing of BECs revealed that VCAM1 expression was limited to BECs in arteries/arterioles and veins/venules; it was not expressed in capillaries. A single-cell transcriptomic study of the brain vasculature, by Christer Betsholtz at Sweden’s Uppsala University, also reported expression of VCAM1 on arteries and veins, but not capillaries (Vanlandewijck et al., 2018).
At Keystone, Wyss-Coray reported that vein/venule BECs expressing VCAM1 also expressed a slew of pro-inflammatory cytokine receptor genes, as well as receptors known to interact with leukocytes.
The researchers next investigated how young versus old plasma affected BEC VCAM1 expression. Exposing young mice to plasma from aged mice—whether via parabiosis or intravenous injection—boosted VCAM1 expression on BECs. Similarly, cultured BECs ramped up VCAM1 expression when treated with plasma from old, but not young, mice. Notably, the researchers found the same result when treating young mice or cultured BECs from young mice with aged human plasma. This suggested that shared factors in aged mouse and human plasma ramped up VCAM1 expression on BECs.
Stop VCAM1, Stop Aging? Anti-VCAM1 antibodies increase the number of proliferating (EdU+) neural progenitor cells (Sox2+) in the subventricular zone (white lines) of the dentate gyri of young mice treated with aged human plasma (left two panels). Anti-VCAM1 antibodies decrease the number of activated (CD68+) microglia (Iba1+) (right two panels). [Courtesy of Yousef et al., bioRχiv, 2018.]
So far so good, but does VCAM1 facilitate what aged plasma does to the brain? To find out, the researchers injected an anti-VCAM1 antibody into young animals treated with plasma from old mice or humans. As seen previously, in young brains, aged plasma reduced proliferation of neural progenitor cells, a marker of neurogenesis, and boosted activation of microglia; however, treatment with the VCAM1 antibody completely blocked these effects. The same was true in the context of normal aging, as 16-month-old mice injected with the antibody every three days for three weeks had higher rates of neurogenesis and fewer reactive microglia than untreated mice.
Does the VCAM1 need to be on BECs? To find out, the researchers generated mice in which VCAM1 expression can be shut off only in BECs. Doing so between the age of two and 16 months resulted in old mice having more neurogenesis and less neuroinflammation than did mice that had been expressing VCAM1 on BECs all along.
The findings suggest that VCAM1 expression on BECs somehow mediates the detrimental effects of aged plasma on the brain, Wyss-Coray told the audience. It is unclear how this works, but Wyss-Coray hypothesized that systemic factors ramp up VCAM1 expression on the brain endothelium, and possibly also VLA-4 expression on circulating leukocytes. This would lead to leukocyte tethering along the luminal side of the BBB, where they might inflame the endothelium and transmit inflammatory mediators into the brain. Either directly or via activation of microglia, these mediators could suppress neurogenesis. The exact sequence of events remains to be discovered.
Fielding questions from the audience after his talk, Wyss-Coray added nuance to his findings. He said that using an anti-VLA4 antibody instead of an anti-VCAM1 antibody prevented the age-related increase in microglial activation but not the decline in neurogenesis. When asked whether a VCAM1 antibody might work as an anti-aging therapeutic, Wyss-Coray cautioned that anti-VLA4 antibodies, which are a standard treatment for people with multiple sclerosis and Crohn’s disease, cause cancer in a small proportion of patients. Given the severity of those diseases, the cancer risk is considered acceptable. “VCAM1 antibodies would probably be seen in a similar light: If they are efficacious against AD or a similar neurodegenerative disease, a black label might be accepted, and the potential cancer risk would be carefully monitored,” he told Alzforum later.
Jonas Neher of the German Center for Neurodegenerative Diseases in Tübingen was intrigued by the work. Neher recently reported that systemic inflammation steers the course of future microglial responses to neurodegenerative disease in the brain (Apr 2018 news). However, Neher contends that soluble factors, such as cytokines, are highly likely to play a role in signaling into the brain. Wyss-Coray agreed, adding that soluble factors could still influence the brain via leukocytes, by upregulating both VCAM1 on the endothelium and VLA-4 on leukocytes.
‘Fit’ Infusions, Youthful Fractions
While the new data on VCAM1 speak to how old blood ages the brain, the Keystone meeting also featured findings about how young blood, perhaps even that of athletes, rejuvenates. Zurine de Miguel, a senior scientist in Wyss-Coray’s lab, reported that not just youth but physical fitness might bestow neurogenesis powers onto blood. De Miguel found that plasma collected from mice with access to a running wheel and infused into sedentary mice boosted the birth of newborn neurons in them. Previous studies have linked exercise to neurogenesis, and De Miguel’s findings now suggest that factors in the plasma facilitate the benefit. De Miguel also reported that so-called “runner’s plasma” boosted expression of neuroplasticity genes in recipients, and dampened their inflammatory response to lipopolysaccharide. De Miguel demurred when asked to opine whether regular infusions with the blood of exercisers could substitute for actually working out.
Eva Czirr of Alkahest, a biotech company in San Carlos, California, presented clinical findings based on Wyss-Coray’s research. He co-founded Alkahest with Karoly Nikolich in hopes of parlaying plasma products into treatments for neurodegenerative and other age-related diseases. The company completed a small Phase 1 study testing the safety of “young plasma” derived from 18- to 22-year-old donors in recipients with mild to moderate AD (Dec 2017 conference news).
As that trial was ongoing, Czirr and colleagues zeroed in on a plasma fraction—dubbed GRF6019—that they speculate will contain the beneficial properties of whole plasma without its downsides. Comprising about 400 proteins, GRF6019 is devoid of immunoglobulins and clotting factors; this eliminates the need to match recipients to donors and reduces the risk of stroke. Plasma fractions are made from pooled plasma of many donors, facilitating production of large, standardized batches. Alkahest acquires its plasma fractions from Barcelona-based Grifols, the world’s largest producer of blood products. Czirr said that Grifols collects an average of 25,000 plasma donations per day around the world.
At Keystone, Czirr reported that GRF6019 had lasting benefits in old mice. Three months after receiving intravenous infusions, the mice still had elevated rates of neurogenesis, lower markers of neuroinflammation, and performed better on tests of memory than their untreated counterparts.
In addition, Czirr reported that Alkahest’s Phase 2 clinical trial is officially underway (clinicaltrials.gov). The trial tests GRF6019 in 40 participants with clinically diagnosed mild to moderate AD. They will be equally randomized to receive a low or high dose of GRF6019 intravenously for five consecutive days on the first week of the trial, then again for five days during week 13. The primary outcome measure is safety; cognitive, functional, and neuropsychological tests make up secondary outcomes. Participants will be monitored for a total of six months during the trial.—Jessica Shugart
- Paper Alert: Do Blood-Borne Factors Control Brain Aging?
- In Revival of Parabiosis, Young Blood Rejuvenates Aging Microglia, Cognition
- Young Blood a Boon for APP Mice
- Stuck in the Past? Microglial Memories Dictate Response to Aβ
- Blood, the Secret Sauce? Focus on Plasma Promises AD Treatment
- Vanlandewijck M, He L, Mäe MA, Andrae J, Ando K, Del Gaudio F, Nahar K, Lebouvier T, Laviña B, Gouveia L, Sun Y, Raschperger E, Räsänen M, Zarb Y, Mochizuki N, Keller A, Lendahl U, Betsholtz C. A molecular atlas of cell types and zonation in the brain vasculature. Nature. 2018 Feb 14; PubMed.
No Available Further Reading