. 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.


Please login to recommend the paper.


  1. There are points of real interest here, and it's potentially informative to take a step back and consider what's been detected using the elegant scRNA-Seq and computational modeling approach. At a first approximation, the investigators appear to have rediscovered and expression-profiled the plaque-associated macrophages (in their view, derived from microglia), which have been known in AD since initial pathological descriptions. What's intriguing and provocative is that they did so by identifying a subgroup defined through expression profiling, and subsequently localized this subset to the macrophages around plaques.

    This work confirms that these cells are biologically distinct from those remaining tiled through the parenchyma or in unaffected brain regions such as the cerebellum. Temporal resolution of the plaque-associated macrophage expression profile suggested that it emerges in stages, first accompanied by suppression of regulators of the microglial phenotype, such as Cx3cr1, and later mediated by signaling through TREM2 and TyroBP/DAP12.

    The crucial unresolved question (finessed rather aggressively by the article’s title but nonetheless crucial and unresolved) concerns the proposed protective functions of these microglia. Plaque-associated macrophages are induced by effective anti-amyloid passive immunization (Sevigny et al., 2016) and have been beautifully shown to limit nearby neuritic pathology (Condello et al., 2015), a function enhanced by genetic disruption of Cx3cr1. Amit and colleagues document the pathophysiological downregulation of Cx3cr1 en route to the DAM phenotype, a finding of genuine interest. However, the efficiency of Cx3cr1-deficient cells in plaque clearance (Liu et al., 2010; Lee et al., 2010) should by no means be interpreted as indicating their uniformly beneficial nature: in a model of tau pathology uncomplicated by expression of mutant tau, Cx3cr1 deficiency markedly worsens pathology and cognition (Bhaskar et al., 2010). Therefore, these DAM cells may be regarded as exerting a highly temporally restricted beneficial function in the initial phases of AD, but may become deleterious subsequently. One salient corollary: application of what the authors term “checkpoint” therapeutics should be precisely timed as defined by objective biomarkers of a target pathological process.

    Other questions and research avenues opened by the present research report:

    1. These DAM cells (it’s simply fun to write that phrase) emerge in other contexts, including aging and SOD1-G93A mutant mice. This observation suggests that the DAM expression phenotype is rather non-specifically triggered by altered brain homeostasis. It remains completely speculative whether the DAM-like cells in ALS models (or disease) and in aging play deleterious, helpful or neutral roles. The (probable) epigenetic basis for this expression phenotype will be of considerable interest to decipher.
    2. Genetic background of the control animals was not specified. The 5xFAD models were on a mixed background, so it will be important to determine to what extent the observations reported here were related to strain background. Similarly, the sex of the mice was not reported (in the Methods section, at least) and gender dimorphism of microglia may play a significant role in defining transcriptional profiles. Both sexes need to be characterized in detail.
    3. CD11c+ microglia were previously characterized in detail as to morphology and localization (Prodinger et al., 2011). It would be useful to know the relationship between DAM cells and CD11c+ microglia in the healthy brain.
    4. The authors propose that downregulation of P2y12 may be a physiological adaptation enabling acquisition of the DAM phenotype. In contrast to the situation with Cx3cr1, evidence for transcriptional regulation downstream of P2y12 has not been presented. Instead, loss of this purinergic receptor will render microglia “blind” to ATP signaling, and is likely to be deleterious with regard to their physiological and wound-limiting functions (Abiega et al., 2016; Davalos et al., 2005). Such lesions can include microvascular hemorrhage characteristic of AD (Lou et al., 2016). 
    5. It should also be noted that increased CSF-soluble TREM2 was incisively studied in dominantly inherited AD (Suárez-Calvet et al., 2016), and shown conclusively to increase along with CSF biomarkers of amyloid and tau pathology. As sTREM2 is shed from the membrane-associated form, these data suggest that TREM2 upregulation in human AD occurs relatively late in the pathogenic cascade, raising uncertainty whether the DAM phenotype is implicated in the response to preclinical AD pathology, the main focus of present therapeutic efforts (Sperling et al., 2013). 

    In summary, DAM cells appear to constitute a distinct response of microglia to a number of states involving altered CNS homeostasis. This report provides a preliminary characterization and points to a sequential acquisition of the phenotype. Their regulation and activities in disease states, for good or ill, remain undefined.


    . The antibody aducanumab reduces Aβ plaques in Alzheimer's disease. Nature. 2016 Aug 31;537(7618):50-6. PubMed.

    . Microglia constitute a barrier that prevents neurotoxic protofibrillar Aβ42 hotspots around plaques. Nat Commun. 2015 Jan 29;6:6176. PubMed.

    . CX3CR1 in microglia regulates brain amyloid deposition through selective protofibrillar amyloid-β phagocytosis. J Neurosci. 2010 Dec 15;30(50):17091-101. PubMed.

    . Regulation of tau pathology by the microglial fractalkine receptor. Neuron. 2010 Oct 6;68(1):19-31. PubMed.

    . CD11c-expressing cells reside in the juxtavascular parenchyma and extend processes into the glia limitans of the mouse nervous system. Acta Neuropathol. 2011 Apr;121(4):445-58. Epub 2010 Nov 13 PubMed.

    . Neuronal Hyperactivity Disturbs ATP Microgradients, Impairs Microglial Motility, and Reduces Phagocytic Receptor Expression Triggering Apoptosis/Microglial Phagocytosis Uncoupling. PLoS Biol. 2016 May;14(5):e1002466. Epub 2016 May 26 PubMed.

    . ATP mediates rapid microglial response to local brain injury in vivo. Nat Neurosci. 2005 Jun;8(6):752-8. PubMed.

    . Purinergic receptor P2RY12-dependent microglial closure of the injured blood-brain barrier. Proc Natl Acad Sci U S A. 2016 Jan 26;113(4):1074-9. Epub 2016 Jan 11 PubMed.

    . Early changes in CSF sTREM2 in dominantly inherited Alzheimer's disease occur after amyloid deposition and neuronal injury. Sci Transl Med. 2016 Dec 14;8(369):369ra178. PubMed.

    . Preclinical Alzheimer disease-the challenges ahead. Nat Rev Neurol. 2013 Jan;9(1):54-8. Epub 2012 Nov 27 PubMed.

    View all comments by Richard Ransohoff
  2. Work by Keren-Shaul et al. provides a very compelling demonstration that microglia activation in the context of chronic diseases, and possibly aging, does not seem to operate in turns of simple “digital–on/off” states. Transition states are involved, and these may be also critically involved in brain pathologies. To my knowledge, this is the first time such a state of “transition” has been described for microglia, at least in such a comprehensive manner. I also think that the data here demonstrate the limitation of the paradigm of M1–M2 state of microglia (or macrophage) activation. As illustrated by the current work, microglia are incredibly plastic cells, and that quality of plasticity is poorly captured by this simple paradigm. Thus, the current work helps significantly refine our conceptualization of microglia activation phenotypes in the brain.

    In addition, there seems to be an important role for TREM2 in the transition of microglia from the cluster II to the cluster III phenotype, or the DAM state. In human Alzheimer’s disease, TREM2 functions appear to be protective as the R47H mutation, which results in loss-of-function of TREM2, predisposes for the development of AD (Guerreiro et al., 2013). TREM2 loss-of-function studies in mice corroborated the human data (Wang et al., 2016). 

    Combining these studies/observations, Keren-Shaul et al. propose the hypothesis that the DAM state is neuroprotective. It would have been very informative for the authors to show how absence of TREM2 in their AD model affects cognitive functions in mice. Nonetheless, I think the hypothesis is reasonable. Thus, it could be that one key role of TREM2 is to promote an overall state of microglia activation that enables these cells to perform better functions that help neutralize disease mechanisms relevant to AD.

    What triggers this microglia-activating function of TREM2 is still not well understood, but evidence suggest that lipids and/or proteins involved in lipid biology like ApoE might have a role (Wang et al., 2015; Yeh et al., 2016). Also, what is it about the DAM state that helps microglia prevent AD? Keren-Shaul et al. note that many genes linked with the DAM state are related to lysosomal/phagocytosis functions and lipid metabolism. Thus, another very important contribution of that study is that it “nominates” genes whose activity might be critical to that protective state of DAM. This should prove very beneficial to the AD research community. Finally, whether the DAM state can be promoted or induced pharmacologically needs to be studied.

    A few outstanding points remain. First, the molecular mechanisms that trigger cluster I microglia to transition to the cluster II phenotype remain unknown. This is a point raised by the authors, and one for which we still lack robust answers in 2017. Interestingly, some evidence from multiple studies suggests that a type I interferon signaling may be at play. One of the more peripheral results from Keren-Shaul et al. is that DAMs are also present in the brains of older mice (Figure S4E). In addition, a previous study by the groups of Michal Schwartz and Ido Amit showed that aging is associated in the brain with an increase in type I interferon signaling (Baruch et al., 2014). This latter study, however, did not investigate how microglia are integrated with this response. Nonetheless, putting the current study in the context of the previous one, it is tempting to hypothesize that the transition from cluster I to the cluster II and III state of microglia may be driven, at least in part, by type I interferon signaling. This is a relatively straightforward hypothesis, and we expect that it will be the focus of a study in the near future.

    Finally, there is some reason to think that DAM might also be seen in the human brain in AD. For one, the preliminary evidence from Keren-Shaul et al. on Lpl mRNA expression, which is induced in DAM in mice and is also seen with microglia associated with senile plaque in the human brain, is encouraging. It is also important to highlight recent work by Erik Boddeke’s group at University of Groningen, which reported that senile plaque-associated microglia in the brains of individuals diagnosed with AD expressed high levels of APOE, AXL, TREM2, and TYROBP, which were all components of clusters II and/or DAM microglia (Yin et al., 2017). Lastly, the recent characterization of human microglia by Christopher Glass’ laboratory at UC San Diego showed that key signaling pathways that are important to specify gene expression in mouse microglia are relatively well conserved in human microglia (Gosselin et al., 2017). 

    Thus, there is certainly some early evidence suggesting that the mouse DAM phenotype may be relevant to human microglia in AD, but the full extent of these similarities needs to be thoroughly investigated. Given the huge progress we have made in our understanding of microglia biology over the past 10 years, we expect that we will have an answer to that critical question as well in the near future.


    . TREM2 variants in Alzheimer's disease. N Engl J Med. 2013 Jan 10;368(2):117-27. Epub 2012 Nov 14 PubMed.

    . 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.

    . 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.

    . TREM2 Binds to Apolipoproteins, Including APOE and CLU/APOJ, and Thereby Facilitates Uptake of Amyloid-Beta by Microglia. Neuron. 2016 Jul 20;91(2):328-40. PubMed.

    . Aging-induced type I interferon response at the choroid plexus negatively affects brain function. Science. 2014 Aug 21; PubMed.

    . Immune hyperreactivity of Aβ plaque-associated microglia in Alzheimer's disease. Neurobiol Aging. 2017 Jul;55:115-122. Epub 2017 Mar 27 PubMed.

    . An environment-dependent transcriptional network specifies human microglia identity. Science. 2017 Jun 23;356(6344) Epub 2017 May 25 PubMed.

    View all comments by David Gosselin
  3. The DAM microglia reported by Keren-Shaul et al. share several markers with "dark microglia,” a phenotype we recently characterized at the ultrastructural level (Bisht et al., 2016). Dark microglia are rarely present under steady-state conditions, in the hippocampus, cerebral cortex, amygdala, and hypothalamus, but become prevalent upon chronic stress, aging, fractalkine signaling deficiency (CX3CR1 KOs), and in the APP-PS1 model of Alzheimer’s disease.

    Dark microglia show reduced expression of CX3CR1 (but also of IBA1, contrary to the DAM microglia), and were not found to co-localize with P2RY12. They are strongly positive for CD11b (but not for CD11c using immunoEM), 4D4, and TREM2, when they associa Aβ te with Aβ plaques. Besides their phagocytosis of Aβ (we very frequently observed engulfments), dark microglia appear extremely active at synapses, even more than normal microglia, suggesting their implication in the pathological/traumatic remodeling of neuronal circuits. 

    Regarding the DAM microglia, I agree with David Gosselin that it would be extremely important to determine their consequences on cognitive function by conducting behavioral experiments.


    . Dark microglia: A new phenotype predominantly associated with pathological states. Glia. 2016 May;64(5):826-39. Epub 2016 Feb 5 PubMed.

    View all comments by Marie-Eve Tremblay
  4. The existence of plaque-associated microglia as a distinct molecular entity was previously documented by the labs of Elly Hol and Javier Vitorica. The current paper is a very elegant dissection and analysis of microglia subtypes, but there is no demonstration that the DAM microglia engulf plaques or restrict neurodegeneration, as respectively stated in the titles of the Alzforum piece and the article itself. Clearly, the mice have plaques galore.

    My view is that DAM microglia might represent a maladaptive response of surveillant microglia, which, in the presence of excessive amounts of Aβ, suffer a phenotypical involution such that vestigial pathways from the macrophage precursors that give rise to microglia become activated in an aberrant manner. The resulting DAM microglia might be a defective microglia that neglects regulation of neuronal circuits, and a defective macrophage that has no capacity to efficiently phagocytose plaques. 

    With regard to the role of TREM2 in the phenotypical transformation of microglia, I find quite on target recent studies from the Haass lab, which have shown alterations in microglia motility in a model of FTD caused by a TREM2 mutation. I posit that impaired microglia surveillance might alter synaptic scaling, thereby causing neuronal hyperexcitation and ensuing “burn out,” which may explain the striking decrease in brain metabolism shown in the mice.

    We recently published an opinion piece proposing a revision of the notion of "neuroinflammation": Masgrau et al., 2017


    . Should We Stop Saying 'Glia' and 'Neuroinflammation'?. Trends Mol Med. 2017 Jun;23(6):486-500. Epub 2017 May 9 PubMed.

    View all comments by Elena Galea
  5. We thank Richard Ransohoff for his insightful comments. We would like to clarify several points:

    1. Indeed, plaque-associated macrophages have been known in AD since initial pathological descriptions. However, this is the first study that describes the disease-associated microglia population in precise molecular terms. Previous attempts over the last decade to identify these cells classified an entire zoo of different myeloid populations, and as such overlooked the important pathways and genes at play in DAM, while attributing inaccurate (and sometimes deleterious) functions to these cells. We believe the cells and pathways we describe are an important stepping-stone to move the field forward.

    2. The genetic background of the control mice is identical to the 5xFAD. They are raised in the same facility and cages to avoid any unrelated genetic or environmental effects. We fully agree that both sexes need to be characterized in detail. We reported the sex of the mice in both the figure legends and methods section. Our conclusions are based on single-cell profiling of more than 30 independent mice replicates, both male and female. As can be seen in Figure 1b, aged-matched 5xFAD females tend to have slightly more DAM cells than males.

    3. We do not identify any DAM cells in healthy mice at six months of age, at least to the resolution that we profile. It is important to note that while all DAM cells are CD11c+, many CD11c+ cells in Alzheimer’s disease brains are not DAM. This may cause artifactual results, as seen in the many papers that reported on such mixtures of cell populations. See, for example, the following figure plotting CD11c intensity (measured with index sorting on the single cell level) and the DAM program.

     [[{"fid":"58261","view_mode":"large","fields":{"format":"large","field_brf_image_display_title[und][0][value]":"","field_brf_image_display_caption[und][0][value]":"","field_brf_image_display_credit[und][0][value]":"","field_internal_usage_notes[und][0][value]":""},"type":"media","field_deltas":{"1":{"format":"large","field_brf_image_display_title[und][0][value]":"","field_brf_image_display_caption[und][0][value]":"","field_brf_image_display_credit[und][0][value]":"","field_internal_usage_notes[und][0][value]":""}},"link_text":null,"attributes":{"height":298,"width":627,"class":"media-element file-large","data-delta":"1"}}]]


    (A) Scatter plot showing the correlation to the average DAM transcription program and index-sorting intensities of CD11c. (B) tSNE plot of CD11b+ cells from both wild-type and AD mouse. Cell color based on cluster association as shown in A. All DAM cells (red) show high levels (3.5E3) of CD11c but not all CD11c cells are DAM.

    4. A note regarding the statement in the Alzforum news article “The epigenetic basis for this expression phenotype will be of considerable interest to decipher.” We show in the paper that the epigenetic profile of DAM and homeostatic microglia are almost identical (S2F and S2G). Focusing on DAM-specific genes, we observed active H3K4me2 regions in both the microglia and DAM, demonstrating that the DAM program is already primed in homeostatic microglia (Figures S2F and S2G). This finding is in line with our single-cell profiling of WT and TREM2 KO. The DAM program is highly anticipated by the microglia and regulated; these are not cells losing control in the face of neuronal damage.

    5. The finding that myeloid cells around plaques are brain-resident microglia, not bone marrow-derived macrophages, has been definitively demonstrated through recent parabiosis experiments in two distinct models of AD, including that presented in our study. The detailed molecular characterization of DAM, and the two stages, strongly support these studies, setting these cells aside from conventional macrophages. In fact, previous attempts to characterize the myeloid cells around plaques based on cellular markers have generated opposite results regarding TREM2-expressing myeloid cells. This can be explained by the impurity of the myeloid cells due to the markers used.

    6. Genetic studies have shown that TREM2 polymorphisms impairing TREM2 function increase the risk of AD three- to fivefold. Our identification of a TREM2-dependent stage in the activation of DAM in models of Aβ accumulation support a protective function of DAM in this type of lesion. Whether DAM have protective functions in other lesions associated with AD, such as taupathy, is an important question for future studies. It can be addressed with the cutting-edge approach we have developed and reported.

    Ido Amit of the Weizmann Institute of Science also contributed to this comment.

    View all comments by Ido Amit
  6. In addition to the certainty in life of death and taxes, we also have the guarantee of scientific advancements—although sometimes met with intermittent setbacks. This rings true when considering the back-and-forth play with respect to the role of TREM2 in AD. Amit’s team elegantly uses single-cell RNAseq supplemented by unbiased algorithms to illustrate the transition of microglial activation states from homoeostatic to neurodegenerative phenotypes.

    Although at face value this resembles the M1-M2 nomenclature, it’s a step forward because the authors acknowledge a dynamic shift in microglial activation states. Interestingly, they report on a heterogeneous CD11c myeloid population within the CNS. A logical next step would be to tease apart central versus peripherally infiltrating myeloid cells in the brain. 

    View all comments by Terrence Town

Make a Comment

To make a comment you must login or register.