Not all microglia in the plaque-ridden brain are equal. Compared to those merely hanging out near plaques, mouse microglia that are actively consuming them express a distinct transcriptional signature, according to a study published May 21 in Nature Communications. Using a fluorescent dye to single out microglia with a belly full of fibrillary Aβ, researchers led by Enrico Petretto of Duke-National University of Singapore Medical School, and Jose Polo of Monash University in Clayton, Australia, reported that these plaque-eaters possessed a transcriptional profile akin to those identified previously in amyloid models, but with a wider array of genes involved. The profile was driven by the transcription factor HIF-1α, which also cropped up in some microglia in brain samples taken postmortem from people who had had Alzheimer’s disease. Curiously, microglia without Aβ in their innards aged faster in amyloid models, and were more likely to contain bits of synapses than were microglia gorging on Aβ.

  • Aβ-laden microglia have a unique gene-expression signature.
  • It includes TREM2, ApoE, other AD-associated genes.
  • Moved to plaque-free environs, the microglia reverted to a homeostatic state.
  • HIF-1α drove the transcriptional regime. It is up in microglia from AD postmortem brain.

“This work provides important new insight into our understanding of how interaction with amyloid plaques impacts microglial biology, and it has the potential to inform therapeutic strategies targeting this cell type,” wrote Joseph Lewcock and Pascal Sanchez of Denali Therapeutics in San Francisco.

As the resident immune cells in the brain, microglia, by definition, are poised to sense and rapidly respond to changes in their environment. In the case of amyloid, recent studies suggest that these cells not only surround plaques, but also build them by ingesting and regurgitating Aβ (Apr 2021 news). In mouse models of amyloidosis, the cells were found to ditch their homeostatic transcriptional signature for a disease-associated one, marked by more expression of TREM2, ApoE, and a handful of other genes (see Jun 2017 news; Sep 2017 news). The studies spotted microglia with this signature milling around plaques. However, these transcriptional studies did not determine whether active ingestion of plaques, as opposed to mere proximity to plaques, influenced the cells’ transcriptional profile.

To investigate, co-first authors Alexandra Grubman, Xin Yi Choo, and Gabriel Chew injected 5xFAD mice intraperitoneally with methoxy-XO4. A fluorescent dye, X04 binds Aβ fibrils. Grubman and colleagues extracted the animals’ brains two hours later, and used fluorescence-activated cell sorting to isolate microglia with and without the dye. They found that 13.5 and 15.8 percent of microglia were actively gobbling Aβ in the brains of 4- and 6-month-old mice, respectively. In the cerebellum, a region nearly devoid of plaques, only 4 percent of microglia did.

Caught in the Act. Two hours after injecting methoxy-XO4 (blue), the dye labeled plaques and co-localized with microglia (green) in the hippocampi of 5xFAD mice (right), but was absent from wild-type hippocampi (left). [Courtesy of Grubman et al., Nature Communications, 2021.]

Analyzing the transcriptomes of the microglia, the scientists identified 2,475 genes that were differentially expressed in fibrillar Aβ-positive versus fAβ-negative microglia. Featuring prominently in this profile were genes involved in ribosome function, oxidative phosphorylation, and phagolysosomal pathways. TREM2, ApoE, and interacting genes were among the most upregulated ones. The profile partly overlapped with the DAM and MGnD signatures previously reported in 5xFAD and APP/PS1 mice.

Compared with these prior signatures, however, the fibrillar Aβ (fAβ) profile encompassed more genes. They implicated a broader set of functions, including the HIF-1 signaling pathway, steroid biosynthesis, mitophagy, and endoplasmic reticulum protein processing. Genes linked to a range of neurodegenerative diseases, including Alzheimer’s, Parkinson’s, and Huntington’s, also cropped up among the plaque consumers. A principle component analysis indicated that plaque phagocytosis held sway over the microglial transcriptomes.

Lewcock and Sanchez noted that the fAβ trancriptome profile jibes with what they saw in XO4-labeled microglia in their APP knock-in mice, and partially confirms the “PIGs” signature defined by spatial transcriptomics (Feb 2021 newsJul 2020 news). In addition, the Denali scientists reported that the Aβ-stuffed microglia had profound lipid alterations. A profile similar to the Aβ-phagocytosing one emerged when the researchers dissected transcriptomes using single-cell RNA sequencing instead of bulk sequencing of sorted microglia.

The scientists also compared the microglial transcriptomes from 6- or 24-month-old mice to look for age differences. They found that wild-type microglia from 24-month-old mice resembled fAβ-negative microglia from 6-month-old 5xFAD mice. This suggested, perhaps surprisingly, that in the plaque-laden brain, microglia that do not actively partake in plaque phagocytosis appear to age faster than do microglia in normal brain.

This aging signature was marked by an uptick in expression of α-defensin genes. These encode anti-microbial peptides with unknown functions in the brain (Selsted and Ouellette, 2005). Aβ also has antimicrobial properties (Mar 2010 news; May 2016 news). 

Does something about the plaque environment trigger microglia to switch on the fAβ signature even before they take their first bite of a plaque, or does gobbling Aβ itself flip the switch? To address this, the researchers ran a series of ex vivo crossover experiments, in which they added microglia from 5xFAD or wild-type mice to hippocampal slice cultures from mice of either genotype. In a nutshell, they found that the fAβ-positive signature was induced by consumption of Aβ. Wild-type microglia transferred to a 5xFAD slice culture only turned on the transcriptional program once they engulfed Aβ, while those WT microglia that did not imbibe the peptide remained in homeostasis mode.

Further, the fAβ-positive signature was reversible. Fibril-containing microglia all reverted to a homeostatic state when they were added to wild-type cultures. Together, the ex vivo data suggest that internalized Aβ fibrils set a transcriptional program in motion and, when Aβ is no longer around, the cells revert back to normal.

Lewcock and Sanchez found this fascinating. “State reversibility of microglia may have interesting implications for effective anti-amyloid therapies, since microglia in the vicinity of those plaques may revert back to a functional homeostatic state when the amyloid aggregates are cleared out,” they wrote.

Microglia can also acquire a taste for synapses, and have indeed been found to prune them rather zealously in mouse models of amyloidosis (Apr 2016 news). Do the plaque eaters also prune synapses? Perhaps; however, Grubman et al. found more synaptic material within fAβ-negative microglia than within fAβ-positive cells, at least in 6-month-old 5xFAD mice. Still, when they isolated microglia from the 5xFAD mouse brain and fed them synaptosomes, the fAβ-positive cells engulfed more hungrily, in keeping with their more phagocytic transcriptional profile.

Why did fAβ-positive microglia nosh on synaptosomes fed to them in culture, but not on synapses within the brain? The answer is unclear. One possibility is that in the brain, fAβ-positive microglia simply contain less synaptic material than do fibril-negative microglia because synapses are sparser around plaques.

On that note, the authors found that the HIF-1α transcription factor appeared to drive a large proportion of the fAβ-positive signature. It also promoted phagocytosis of synaptosomes. The researchers went on to demonstrate that HIF-1α signaling could be partially induced by Pam3csk, a toll-like receptor 2 ligand, and repressed by rapamycin, an inducer of autophagy.

But How About People?
Do any microglia in the human brain adopt an Aβ phagocytosis transcriptional signature? The researchers tackled this important question by integrating single-nucleus transcriptomic data from four independent postmortem cohorts (Zhou et al., 2020; Mathys et al., 2019; Grubman et al., 2019Leng et al., 2021). Combined, the dataset included nearly 12,000 microglial nuclei from entorhinal and prefrontal cortex samples taken postmortem from 102 people, some of whom had died of Alzheimer’s. These nuclei split into 21 clusters based on their transcriptomes; some clusters only appeared in one dataset.

Two transcriptional clusters, dubbed 10 and 11, had profiles that resembled the mouse fibrillar Aβ signature. These clusters were identified in every dataset, but not in every sample. Among the samples that contained cluster 10 microglia, those that came from people with AD had a higher proportion of these cells than did samples from people without AD. The microglial genes in cluster 10 overlapped with roughly 20 percent of genes in the mouse Aβ signature, compared with 10 percent of genes in the previously reported DAM signature. Notably, genes controlled by HIF-1α were among the genes differentially expressed in cluster 10 microglia.

Cluster 10 was not found in all AD samples. This intrigued Colm Cunningham of Trinity College Dublin. “This could represent one level of differential susceptibility among individuals, perhaps affecting progression or patient-specific manifestations,” he wrote (comment below).

All told, the researchers propose that as microglia age, their transcriptomes change in a way that is accelerated in the AD brain, as exemplified by the fAβ-negative signature in the 6-month-old 5xFAD mice. Upon Aβ plaque internalization, the HIF-1α program switches on, sparking a feed-forward loop that promotes even more gorging on Aβ. The authors suspect this same transcriptional program could also exacerbate the destruction of synapses by microglia, highlighting the delicate balance of microglial functions that can both help and hurt the brain.

Polo and Grubman said the Aβ signature is likely beneficial in the early stages of amyloidosis. How the signature changes throughout disease, and in the face of other stressors including tau pathology and age-related comorbidities, remains to be tested.

“This study is a bioinformatics tour de force that highlights the HIF-1α signaling pathway as a central mediator of the microglial response to Aβ plaque phagocytosis,” commented Jonas Neher of the German Center for Neurodegenerative Diseases in Tübingen. Neher added that the results align well with previously published work from his lab, which highlighted HIF-1α signaling as a critical microglial response to Aβ pathology (Apr 2018 news).—Jessica Shugart


  1. In this study, Alexandra Grubman and colleagues leveraged an impressive set of experimental approaches and technologies to characterize a subpopulation of microglia with high phagocytic activities around amyloid plaques from the 5xFAD transgenic mouse model. To do so, the authors used an innovative flow cytometry approach to isolate two populations of microglia based on their labeled-fibrillar Aβ content; an approach that we have also successfully used to characterize phagocytic microglia in a novel APP KI mouse model (Xia et al., 2021). Grubman et al. provide important new insight into our understanding of how interaction with amyloid plaques impacts microglial biology and has the potential to inform therapeutic strategies targeting this cell type.

    The authors demonstrate that methoxy-XO4-positive (XO4-positive) microglia sorted from brains have a different gene expression signature compared to XO4-negative microglia in 5xFAD mice. A subset of the differentially expressed genes overlaps with those measured in XO4-negative microglia, but the magnitude of alteration is exacerbated in XO4-positive microglia. The upregulation of genes belonging to pathways involved in metabolic and phagolysosome functions and the increased phagocytic capacity of XO4-positive microglia supports the notion that plaque-associated microglia transition to a responsive state that may help them reduce Aβ aggregates and/or associated pathology (e.g., neuritic dystrophy).

    Those findings further add to our understanding of microglia phenotype at the site of amyloid plaques and partially confirm the “PIGs” signature defined by spatial transcriptomics (Jul 2020 news). Interestingly, we recently reported that XO4-positive microglia in the AppSAA knock-in mouse model (Xia et al., 2021), show not only exacerbated gene expression changes, notably in genes controlling lysosomal function, but also profound lipid alterations, which could indicate that those phagocytic microglia may struggle to regulate lipid metabolism as previously discussed (Lewcock et al., 2020). 

    Another interesting observation from this work is the discovery that XO4-negative microglia, potentially distal from amyloid plaques, show a relatively high level of non-fibrillar Aβ. This observation would suggest that not only the quantity of Aβ but its conformation may determine the transcriptomic and functional changes in microglia. This may have implications for amyloid-targeting antibodies that may bind to certain types of Aβ and drive internalization of the immunocomplex in microglia. However, it is unclear if this observation might be caused by the artificially high levels of Aβ production in the brain of this transgenic model. One alternative explanation is that XO4-positive microglia may become overwhelmed by the large amount of phagocytosed fibrillar Aβ and struggle to degrade those peptides. If this is the case, it would be important to determine if the deficit in lysosomal degradation could cause lysosomal lipid accumulation and eventually cellular dysfunction.

    This manuscript reports many other fascinating observations, including the ability of XO4-positive microglia to revert back to a homeostatic state when exposed to an environment deprived of amyloid plaques. State reversibility of microglia may have interesting implications for effective anti-amyloid therapies, since microglia in the vicinity of those plaques may revert back to a functional homeostatic state when the amyloid aggregates are cleared out.

    Overall, this study adds to our understanding of the complex and heterogeneous roles of microglia in the context of neurodegeneration. Continued work to examine whether alteration of the pathways described in this work is able to impact disease progression will be an exciting area for future study.


    . Fibrillar Aβ causes profound microglial metabolic perturbations in a novel APP knock-in mouse model. bioRxiv. January 20, 2021.

    . Emerging Microglia Biology Defines Novel Therapeutic Approaches for Alzheimer's Disease. Neuron. 2020 Dec 9;108(5):801-821. Epub 2020 Oct 22 PubMed.

  2. The study, by sorting microglia on the basis of which ones had taken up fluorescently labeled Aβ (i.e., the cells become methoxy-XO4-positive), revealed a phagocytic population, which reflected some of the differentially expressed genes we have come to expect from other single-cell RNA-sequencing studies in the AD field (Trem2, Tyrobp, ApoE, Lpl and lysosomal proteases and protease inhibitors such as cathepsins and Cst7 respectively). However, a number of other pathways appeared to be induced, including HIF-1α-induced genes, indicating an activation of anaerobic glycolysis (Pkm, Ldha). This perhaps denies the citric-acid cycle and oxidative phosphorylation of the pyruvate that initiates those bioenergetic pathways. This is interesting in the context of the immunometabolic status of microglia-driving microglial phenotype.

    This signature appeared to be induced even in WT microglia following amyloid phagocytosis. This was shown in experiments using transplantation of Aβ-positive microglia onto organotypic hippocampal slice cultures prepared from WT or 5XFAD mice. However, these are complex experiments with many variables and are complex to interpret. For example, it did appear that some of the microglia that were not phagocytic (i.e., NIAD4-negative) still acquired the phagocytic gene signature and one wonders: Would simply presenting WT animals with fibrillar Aβ, in vivo, actually induce the phagocytic (XO4-positve) signature described here?

    There are fascinating data, suggesting that even though the XO4-positive microglia are clearly more phagocytic in ex vivo experiments (taking up more E. coli, PSD95 and fibrillar Aβ), it is actually the XO4-negative microglia that contain more PSD95 in the 5XFAD brain. Are the XO4-negative microglia, therefore, better equipped to phagocytose synaptic elements? Well, there are several interpretations of those data, but it may simply be a case of exposure: If XO4-positive microglia are most proximal to plaques and fully engaged in amyloid phagocytosis, it may be that those microglia that are a little more distal from the plaque are exposed to the synaptic loss that occurs in the vicinity of plaques.

    The authors do not dwell on one important finding: The XO4-positive signature, although present in microglia isolated from humans, was not enriched in microglia from brains of patients with Alzheimer’s disease.

    However, this is mitigated somewhat by the observation that patients who did have some XO4-positive microglial signatures tended to have upregulated genes from one particular cluster that, once again, expressed these HIF-1α-dependent genes. Possible upstream drivers of this HIF-1α-dependent signature are proinflammatory stimuli, including Pam3csk, which is an activator of Toll-like 1/2 receptors. TLR2 has been shown several years ago to recognize amyloids, including Aβ (Tükel et al., 2009). This could indicate that amyloid may activate pro-inflammatory pathways via TLR2, driving an HIF-1α response that is key in the induction of a phagocytic phenotype. It is intriguing that this is present in some AD cases but is not strongly associated with microglia from AD patients more generally. This could represent one level of differential susceptibility among individuals, perhaps affecting progression or patient-specific manifestations.


    . Responses to amyloids of microbial and host origin are mediated through toll-like receptor 2. Cell Host Microbe. 2009 Jul 23;6(1):45-53. PubMed.

  3. This study by Grubman et al. is a bioinformatics tour de force that highlights the HIF-1α signaling pathway as a central mediator of the microglial response to Aβ plaque phagocytosis. These results are very interesting and match very well with our own previously published work, which highlighted HIF-1α signaling as an important microglial response to Aβ pathology (Wendeln et al., 2018). Moreover, in unpublished data on single-cell microglia from APP23 mice, we have also found that methoxy-XO4 (MX04)-positive microglia are characterized by increased HIF-1α signaling, fully in line with the current findings by Grubman et al., who demonstrate very nicely that this response is directly triggered by microglial interaction/uptake of aggregated Aβ.

    I am not completely convinced by their single-cell RNA-Seq analysis demonstrating that MX04-positive cells may be on a premature aging trajectory, because this dataset contains relatively few cells (sometimes from single animals per group or of different sex) and it does not demonstrate microglial heterogeneity within the same age/experimental groups, which is something that has now been shown repeatedly (e.g., Sala Frigerio et al., 2019; Hammond et al., 2019). Moreover, while the joint analysis of the different human datasets is commendable and important, I am a little concerned by the (sometimes almost complete) separation of the different datasets. One would at least expect to see an overlap of homeostatic microglial subtypes from all datasets, and this lack may indicate that there are still significant effects of batches/different methodologies that were not sufficiently compensated during data integration.

    The analyses performed by Grubman et al. regarding the role of HIF-1α in altering microglial effector functions rely heavily on in vitro assays and (as pointed out by the authors themselves) a chronic microglial-plaque interaction in vivo may result in very different responses. In this regard, a recent study has shed more light on the role of microglial HIF-1α activation in response to plaque pathology (March-Diaz et al., 2021). It not only corroborates the crucial role of HIF-1α in the microglial response to Aβ plaque deposition, but also shows that, in vivo, stabilization of HIF-1α (leading to enhanced HIF-1α signaling) induces “microglial quiescence,” reducing microglial association with plaques and increasing neuritic damage. This is reminiscent of the microglial phenotype in Trem2 knockout animals and TREM2 mutation carriers as described by the Colonna group, and it raises the question of how these pathways are interrelated.

    Grubman et al. speculate that MX04-positive microglia may represent a subset of DAM microglia and it will be interesting to see whether this holds up in future work, or if HIF-1α may actually be required for inducing the DAM phenotype. 


    . Innate immune memory in the brain shapes neurological disease hallmarks. Nature. 2018 Apr;556(7701):332-338. Epub 2018 Apr 11 PubMed.

    . The Major Risk Factors for Alzheimer's Disease: Age, Sex, and Genes Modulate the Microglia Response to Aβ Plaques. Cell Rep. 2019 Apr 23;27(4):1293-1306.e6. PubMed.

    . Single-Cell RNA Sequencing of Microglia throughout the Mouse Lifespan and in the Injured Brain Reveals Complex Cell-State Changes. Immunity. 2019 Jan 15;50(1):253-271.e6. Epub 2018 Nov 21 PubMed.

    . Hypoxia compromises the mitochondrial metabolism of Alzheimer’s disease microglia via HIF1. Nature Aging. 1, 2021, pp.385–99. Nat. Aging.

  4. The work by Grubman et al. is a nice step toward understanding the complex transcriptional responses observed in Aβ-plaque-associated microglia (AβAM). Using an elegant approach, they confirm a trend observed in other works: The closer to Aβ deposits, the stronger/more complex the induced microglial transcriptional responses. Previous studies have shown that the HIF-1 pathway is highly active in microglia from AD mouse models (Baik et al., 2019March-Diaz et al., 2021; Ulland et al., 2017; Wendeln et al., 2018) and HIF-1 itself is included in the DAM signature (Keren-Shaul et al., 2017). This new work is a direct confirmation of the relevance of the hypoxic/HIF-1 pathway in AD microglia. However, what role HIF-1 plays in AD microglia, and what the therapeutic consequences of modulating the pathway could be, are open questions.

    HIF-1 is a well-known cell response to metabolic stress, which adapts cellular metabolism to low oxygen levels by increasing anaerobic glycolysis (glucose to lactate) and decreasing aerobic respiration (OXPHOS). Paradoxically, AβAM from AD mouse models and human samples upregulate OXPHOS (March-Diaz et al., 2021). Of note, the MX-04-positive microglia described by Grubman et al. and the human amyloid-responsive microglia also upregulated both OXPHOS and HIF-1(Nguyen et al., 2020). We also showed OXPHOS activation in all DAM we analyzed, suggesting that mitochondrial activity is a requirement for microglial activity, whereas HIF-1 activation was only observed in Aβ models, and could be an unwanted side response (March-Diaz et al., 2021). Our in vivo conditional genetic experiments to stabilize HIF-1 against degradation, or systemic hypoxia exposure, demonstrated that overactivation of HIF-1 reduces the microglial defensive responses against Aβ and aggravates neuropathology, phenocopying Trem2-deficient mutants (March-Diaz et al., 2021; see also Jonas Neher comment above). Further in vivo work with conditional microglial mutations of HIF-1 or HIF-1 transcriptional targets will be key to define the role of this pathway in AD progression.

    As stated in other comments, the in vitro analysis of the role of HIF-1 in microglial responses to Aβ should be taken with caution, as the optimal nutrient and oxygen conditions in which we culture microglia are far from those bathing microglia in vivo. We have recently shown that Aβ plaques constitute ischemic foci lacking blood vessels, shedding some light on the activation of HIF-1 observed around Aβ plaques (Alvarez-Vergara et al., 2021). Indeed, microglia are almost the only cells able to survive under those stressful conditions. Perhaps they pay a toll for it?


    . Non-productive angiogenesis disassembles Aß plaque-associated blood vessels. Nat Commun. 2021 May 25;12(1):3098. PubMed.

    . A Breakdown in Metabolic Reprogramming Causes Microglia Dysfunction in Alzheimer's Disease. Cell Metab. 2019 Sep 3;30(3):493-507.e6. Epub 2019 Jun 27 PubMed.

    . 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.

    . Hypoxia compromises the mitochondrial metabolism of Alzheimer’s disease microglia via HIF1. Nature Aging. 1, 2021, pp.385–99. Nat. Aging.

    . APOE and TREM2 regulate amyloid-responsive microglia in Alzheimer's disease. Acta Neuropathol. 2020 Oct;140(4):477-493. Epub 2020 Aug 25 PubMed. Correction.

    . TREM2 Maintains Microglial Metabolic Fitness in Alzheimer's Disease. Cell. 2017 Aug 10;170(4):649-663.e13. PubMed.

    . Innate immune memory in the brain shapes neurological disease hallmarks. Nature. 2018 Apr;556(7701):332-338. Epub 2018 Apr 11 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Microglia Build Plaques to Protect the Brain
  2. Hot DAM: Specific Microglia Engulf Plaques
  3. ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
  4. Striking Microgliosis in New APP Knock-in Mice
  5. Paper Alert: Those PIGs! Spatial Transcriptomics Add Human Data
  6. Paper Alert: Aβ’s Day Job—Slayer of Microbes?
  7. Like a Tiny Spider-Man, Aβ May Fight Infection by Cocooning Microbes
  8. Paper Alert: Microglia Mediate Synaptic Loss in Early Alzheimer’s Disease
  9. Stuck in the Past? Microglial Memories Dictate Response to Aβ

Research Models Citations

  1. 5xFAD (B6SJL)
  2. APPPS1

Paper Citations

  1. . Mammalian defensins in the antimicrobial immune response. Nat Immunol. 2005 Jun;6(6):551-7. PubMed.
  2. . Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses in Alzheimer's disease. Nat Med. 2020 Jan;26(1):131-142. Epub 2020 Jan 13 PubMed. Correction.
  3. . Single-cell transcriptomic analysis of Alzheimer's disease. Nature. 2019 Jun;570(7761):332-337. Epub 2019 May 1 PubMed.
  4. . A single-cell atlas of entorhinal cortex from individuals with Alzheimer's disease reveals cell-type-specific gene expression regulation. Nat Neurosci. 2019 Dec;22(12):2087-2097. PubMed.
  5. . Molecular characterization of selectively vulnerable neurons in Alzheimer's disease. Nat Neurosci. 2021 Feb;24(2):276-287. Epub 2021 Jan 11 PubMed.

Further Reading

Primary Papers

  1. . Transcriptional signature in microglia associated with Aβ plaque phagocytosis. Nat Commun. 2021 May 21;12(1):3015. PubMed.