Unlike the simple response of mouse microglia to amyloid plaques, human microglia react in diverse ways that researchers are just beginning to understand. Data from two different model systems, one in vivo and one in vitro, now provide a remarkably congruent catalogue of these reactive states. The in vivo data was described in a July 7 bioRxiv preprint. Scientists led by Renzo Mancuso at the University of Antwerp, and Bart De Strooper at KU Leuven, both in Belgium, transplanted human microglia into the brains of newborn mouse models of amyloidosis and analyzed glial gene expression six months later. They identified transcriptional profiles that resembled those of disease-associated mouse microglia (DAM), as well as distinct antigen-presenting, cytokine-based, and interferon-based profiles. Meanwhile, researchers led by Beth Stevens and Evan Macosko at The Broad Institute of MIT and Harvard, developed an in vitro system for generating microglia from human iPS cells. In a paper uploaded to bioRxiv on May 2, they reported that microglia assumed DAM, antigen-presenting, and interferon-based transcriptional states in the presence of fibrillar amyloid.

  • In chimeric mice, human microglia assume distinct states in response to amyloid.
  • So do human microglia exposed to amyloid in vitro.
  • These include DAM, antigen-presenting, cytokine-, and interferon-response states.

“We are close to untangling the diversity of microglial responses,” Mancuso told Alzforum. Other researchers agreed the field may be approaching a consensus on the major activation states of human microglia. Marco Colonna, Yonghua Zhou, and Khai Nguyen at Washington University School of Medicine in St. Louis noted that transcriptional datasets from human brain have turned up similar findings (comment below).

The versatility of the in vivo and in vitro platforms also generated enthusiasm. “We are excited about the prospect of future studies using these models to test the impact of candidate therapeutics on microglial function and other disease endpoints,” Joseph Lewcock and Kathryn Monroe at Denali Therapeutics wrote to Alzforum (comment below).

Microglial Lineage. Data from chimeric mice suggests two main pathways (brown and turquoise arrows) whereby human homeostatic microglial (HM) respond to amyloid: Cytokine response microglia (CRM) diverge from human leukocyte antigen-presenting microglia (HLA), which transition through a DAM-like state. The former seems to be triggered by Aβ oligomers, the latter by plaques. [Courtesy of Mancuso et al., 2022.]

Moving Beyond DAM
The DAM signature, driven by APOE and TREM2 genes, was first identified in microglia around plaques in mouse models of amyloidosis (Jun 2017 newsSep 2017 news). However, it was unclear if this signature occurred in the human brain, because RNA-Seq studies of microglia from postmortem Alzheimer’s brain tissue found only partial overlap with DAM genes (May 2019 news; Jan 2020 news).

To better dissect microglial responses under different conditions, researchers would like to manipulate these cells in model systems. However, human microglia rapidly change their expression profiles when placed in culture (Jun 2017 news). Turning to a more physiological environment, De Strooper developed a chimeric mouse model, in which microglial precursors generated from human iPSCs are transplanted into the brains of newborn mice (Apr 2019 conference news; Aug 2019 news).

For the current study, joint first authors Mancuso, Nicola Fattorelli, and Anna Martinez-Muriana, placed the precursors in APPNL-G-F knock-in pups, then isolated the human microglia from 6-month-old mice for single-cell RNA-Seq. They identified four main reactive states: DAM, the antigen-presenting form (HLA), the cytokine/chemokine-response microglia (CRM), and interferon-response microglia (IRM). In addition, they found a transitory form with elevated expression of ribosomal genes, as well as homeostatic microglia.

A comparative analysis of the gene expression changes shed light on how microglia might morph from homeostatic to reactive. According to this computational model, homeostatic microglia can take one of two paths, either the cytokine or the DAM response. The CRM pathway includes three successive substates: a transitioning CRM form, CRM1, and CRM2. DAM microglia evolve into the antigen-presenting HLA form. The origin of IRM microglia is less clear, perhaps branching off the DAM path. The transient ribosomal response seems to bridge homeostatic and DAM states (see image above).

Cell-State Triggers
What causes these different states? One clue came from crossing APPNL-G-F mice with APOE knockout mice. In these animals, the lack of APOE causes fewer plaques to form, and the researchers found correspondingly fewer DAM and HLA cells. The HLA form is likely a response to plaques, the authors suggested. Why would microglia assume an antigen-presenting form when faced with plaques? This is unclear, but the authors speculated that these microglia may be trying to stimulate an adaptive immune response by presenting phagocytosed amyloid to T cells, which are known to infiltrate AD brain (Jan 2020 news).

On the other hand, the cytokine response seemed to be triggered by soluble forms of Aβ. The authors injected synthetic Aβ42 oligomers into the cerebral ventricles of 3-month-old chimeric mice with wild-type human APP knocked in (APPhu/hu). Six hours later, most of the human microglia in their brains were in the CRM state.

By contrast, the IRM response remains a complete mystery. These cells were equally present in APPNL-G-F and APPhu/hu mice, indicating it is not an amyloid-specific response, Mancuso told Alzforum.

Other Alzheimer’s risk genes affect these microglial states. For example, knocking out TREM2 in the human microglial precursors before they were grafted into the mouse brain suppressed DAM and HLA, in keeping with other data that TREM2 drives the DAM transitions. The pathogenic R47H variant of TREM2 and the epsilon 4 isoform of APOE also stifled HLA microglia, the authors found.

In fact, out of 85 GWAS candidates, more than half were differentially expressed in microglia in APPNL-G-F versus APPhu/hu mice. Specific gene expression changes were associated with each microglial state. For example, TREM2, APOE, COX7C, and HLA-DQA1 expression ticked up with DAM/HLA microglia, while SPI1, SPPL2A, TNIP1, LILRB2, SEC61G, NCK2, and CTSH expression rose with CRM. The authors are transplanting iPS cells generated from people with specific AD risk variants into their chimeric mice to find out how they affect microglial transitions.

Though these human microglia came from mouse brains, the data dovetail with RNA-Seq studies of human brain tissue. Two previous analyses of living microglia from fresh autopsy tissue or surgical resections described antigen-presenting, cytokine-based, or interferon-based microglial states (Nov 2019 news; Dec 2020 news). In addition, De Strooper and colleagues reanalyzed RNA-Seq data from studies of human postmortem brain and pulled out DAM, HLA, CRM, and IRM transcriptional profiles (Zhou et al., 2020; Gerrits et al., 2021; Dec 2021 news).

Marta Olah at Columbia University Medical Center, New York, who led one of the human brain studies, agreed that De Strooper’s data closely match hers. She was intrigued by the authors’ finding that APOE4 and R47H TREM2 risk variants suppressed HLA microglia. “[This] suggests that this particular state of microglia is protective in the course of AD pathogenesis, further providing support to the conclusion of our own study of human ex vivo microglia,” she wrote (full comment below).

In Vitro Profiles. Heat map of transcriptional profiles from in vitro human microglia, identifying proliferative (yellow), interferon-response (red), homeostatic (blue), antigen-presenting (green), and DAM (orange) states. Columns represent each microglial subtype. Some of the key genes altered in each are noted on the right. [Courtesy of Dolan et al., 2022.]

High-Throughput Microglial Screens
If researchers could study human microglia in a dish, they could better manipulate the cells and determine what affects them. However, cultured microglia generated from iPS cells bear little resemblance to their in-vivo cousins. Stevens wondered if co-culturing these cells with human brain tissue, instead of in monocultures, would evoke more physiological gene expression. To test this, joint first authors Michael-John Dolan and Martine Therrien challenged cultured microglia with various stimuli and stressors from human brain, including synaptosomes, myelin debris, apoptotic neurons, and synthetic Aβ fibrils. Each material induced a different pattern of responses. The transcriptional profiles were similar to those reported in microglia from human postmortem brain, suggesting this in vitro platform is able to trigger states like those found in the complex brain environment.

Notably, amyloid fibrils induced DAM, antigen-presenting, and interferon-response states, similar to those seen in De Strooper’s chimeric mice. The authors also identified homeostatic and proliferative microglia, but not the cytokine-response state. This also fits with De Strooper’s data, since the cultured microglia were not stimulated with Aβ oligomers.

“This work demonstrates for the first time that microglia heterogeneity can be recapitulated in a cell culture dish,” noted Lewcock and Monroe. They were struck by how much human microglial expression varied from that of mouse microglia. “Taken together, these [papers] suggest that mouse models markedly underrepresent the biological states in which microglia exist in AD, and they shine a spotlight on critical species differences.”

Researchers agree more work is needed before these transcriptional findings could be used to identify ways to manipulate microglia therapeutically. Studies will have to determine what genes regulate microglial states and how to push cells in one direction or another, Stevens said. Microglia also present a technical challenge for such genetic screens, because they are notoriously difficult to transduce. The cells express an anti-viral protein, SAMHD1, that chews up viral vectors (Maes et al., 2019). However, Dolan and colleagues found a way around this limitation in their in vitro system by co-delivering the viral protein Vpx, which degrades SAMHD1, along with lentiviral vectors. Vpx brought cellular transduction efficiency from 4 to 89 percent.

They used the lentiviral/Vpx system to overexpress melanocyte-inducing transcription factor (MITF), which is highly expressed in DAM and has been proposed to regulate this state (Friedman et al., 2018). Indeed, upping MITF boosted DAM genes and stimulated phagocytosis.

This system could be used to rapidly test the effects of numerous candidate genes, Stevens said. She believes the in vitro and in vivo models are complementary, noting that the most promising gene candidates could be selected for follow-up in the chimeric mice to dig deeper into the underlying biology. Stevens plans to collaborate with De Strooper’s group to screen GWAS hits in this way. In addition, her group is adding proteomic data, which will offer more clues about the function of different genes.

Scientists stressed it is crucial to understand the biological consequences of each state. “Extensive functional experiments are warranted to determine whether [these states] are beneficial or detrimental, before therapeutic interventions can be confidently applied,” Colonna, Zhou and Nguyen wrote to Alzforum.—Madolyn Bowman Rogers


  1. Mancuso et al. presented an exciting piece of work, taking advantage of the xenotransplantation model for AD, where human microglia precursors were transplanted into immunodeficient amyloid pathology mouse models. Through single-cell RNA-Seq, they identified cytokine response (CRM), interferon response (IFN), disease-associated (DAM), antigen-presenting response (HLA), and transitory microglia clusters in response to amyloid pathology. All of these clusters have been identified previously in mouse models of amyloid pathology (Ellwanger et al., 2021), except for CRM, which was shown to be APOE- or TREM2-independent. CRM, DAM, and HLA microglia subsets were corroborated in various postmortem human AD snRNA-Seq datasets, confirming the validity of this xenotransplantation model.

    Interestingly, pseudotime analysis suggested the development of CRM fell in a distinct trajectory from the rest of the reactive states. Along the line, it is very intriguing to see that oligomeric Aβ induced the majority of homeostatic microglia to the CRM state, not to others. This is consistent with results from Dolan et al., in that CRM was not observed in response to Aβ fibrils or other CNS substrates. It is nice to see the CRM response was specifically induced by Aβ oligomers, supporting the heterogeneity of microglial responses. The fact that this cluster was not identified previously in mouse models was likely due to lack of resolution, since Ccl3 and Ccl4, which were upregulated in CRM microglia, were also upregulated in the DAM cluster in the 5XFAD model (Zhou et al., 2020). 

    Besides CRM and DAM, it still remains an open question as to how the other reactive states are induced in response to Aβ pathology. Pseudotime analysis suggested that HLA and IFN clusters may derive from DAM. However, lineage-tracing analyses are needed to validate this finding. In addition, the IFN cluster was not observed in any of the postmortem human AD snRNA-Seq datasets reanalyzed, questioning its biological relevance in human pathology.

    Despite the characterization of distinct microglia subsets in amyloid pathology, a more fascinating question, the answer to which remains elusive, is what functions these transcriptionally distinctive cells play. Extensive functional experiments are warranted to determine whether they are beneficial or detrimental before therapeutic interventions can be confidently applied.


    . Prior activation state shapes the microglia response to antihuman TREM2 in a mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A. 2021 Jan 19;118(3) PubMed.

    . 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.

  2. While the iPSC-derived microglia (iMGL) in vitro model system demonstrates unexpected levels of microglia heterogeneity and inducibility of different phenotypes (as demonstrated by Dolan and colleagues), the coming of age of the xenotransplantation model of iMGLs holds the greatest promise for dissecting the roles of the different microglia phenotypes in a mechanistic way.

    Recently we reported that human microglia exist in several distinct functional states in the aged human brain (Olah et al., 2020), including homeostatic, interferon response, cytokine response, HLA overexpressing with enrichment of the DAM signature, and proliferative. The study by Mancuso and colleagues was able to recapitulate these human microglia states with high fidelity in a xenotransplant mouse model of Alzheimer’s disease (AD). Moreover, speaking to the strength of this approach, Mancuso et al. have shown that the risk alleles APOE4/4 and TREM2R47H both result in the reduction of the HLA phenotype, suggesting that this particular state of microglia is protective in the course of AD pathogenesis—further providing support to the conclusion of our own study of human ex vivo microglia.

    That human iPSC-derived microglia can acquire the exact same phenotypes as human microglia when transplanted to a corresponding mouse brain environment suggests that these transcriptional states are hardwired in the human microglia epigenome and can be evoked by environmental cues. Nonetheless, our own study and now Mancuso and colleagues show that this plasticity/inducibility is probably lost due to aging, and/or genetic predisposition. Accordingly, it will be critical for the field to gain an in-depth understanding of the epigenetic and metabolomic regulation of these different microglia states to be able to restore microglia plasticity, and to fine-tune microglia phenotypes in the context of therapeutic intervention for neurodegenerative diseases.

    In vivo Perturb-Seq experiments will be fundamental in our quest to identify the master regulators of human microglia subset-specific transcriptional programs and to provide us with clues to unlock the full potential of subset-specific differentiation in aged human microglia.


    . Single cell RNA sequencing of human microglia uncovers a subset associated with Alzheimer's disease. Nat Commun. 2020 Nov 30;11(1):6129. PubMed.

  3. These two new preprints describe exciting new approaches to utilize iPSC-derived human microglia in vitro and in vivo, which reveal important insights into how cell state relates to functional outcomes, and how the response of human microglia to Alzheimer’s disease pathology differs from what has been observed with mouse microglia.

    Mancuso et al. investigated human microglia states in a xenotransplant amyloid model and observed a remarkable increase in diversity and heterogeneity that more accurately reflects the expanded dynamic range of microglial states in human AD. These findings contrast with mouse microglia whose transcriptional signatures are limited to homeostatic or disease-associated microglial (DAM) states in amyloid models. Importantly, the human microglial states described from the xenotransplant model are similar to those detected in previously published single-nucleus RNA-Seq datasets from AD postmortem tissues, indicating a consensus is forming in the field regarding microglial states present in AD. How microglia states are altered during disease progression remains an open question, and these chimeric models could be useful for parsing states that are specific to certain stages of disease and that may protect against or exacerbate neurodegeneration. Taken together, these results suggest that mouse models markedly underrepresent the biological states in which microglia exist in AD and shine a spotlight on critical species differences. Future translational studies will likely need to address this challenge to accurately assess the therapeutic potential of microglial targets.

    It is our hope that broader implementation of chimeric mouse models like this one will better elucidate the biological effects of AD genetic risk factors such as TREM2 and APOE, and will inform therapeutic strategies for modulating microglia. Our understanding is limited of how transcriptional states impact specific microglial functions and of which microglial functions are beneficial versus deleterious in AD. This study highlights the benefits of assessing AD risk variants in human microglia using xenotransplant models and highlights the risks surrounding mechanistic insights or therapeutic strategies based on deep phenotyping of mouse microglia alone. We are excited about the prospect of future studies using these models to test the impact of candidate therapeutics microglial function and other disease endpoints.

    While leveraging novel in vivo models is extremely valuable to study the complex behavior of microglia in disease, these models are resource- and time-intensive. The ability to conduct some of these experiments using an in vitro system would be highly advantageous. Dolan et al. lay a foundation for leveraging iPSC-derived human microglia for in vitro mechanistic studies to bridge a key gap in our understanding of how microglial state relates to function.

    This work demonstrates, for the first time, that microglia heterogeneity can be recapitulated in a cell culture dish, which was previously thought to be limited by monoculture conditions, yet hinted at in studies that showed only a subset of microglia are phagocytic, for example. Dolan’s work builds confidence in the iMGL system by demonstrating overlap of in vitro and in vivo states of human microglia, and it shows consistency across multiple iPSC lines. Furthermore, Dolan et al. could connect transcriptional cell state to microglial function by demonstrating that different phagocytic substrates contribute to microglial diversity.

    From a technical perspective, the study also shows the ability to genetically manipulate iMGLs using a lentiviral approach that ingeniously leverages an HIV restriction factor inhibitor, Vpx, to enable highly efficient microglial transduction, which is a method well worth pursuing to genetically alter these historically difficult-to-manipulate cells.

    Together, these two papers provide a range of insights into the function of human microglia, their role in AD, and novel insights into how they can be used as an effective research tool in the future.

  4. In their recent preprint in bioRXiv, Dolan and colleagues report an in vitro platform utilizing human stem-cell-derived microglia (iMGLs) that allows for viral manipulation and high-throughput screens. They show that exposing iMGLs to central nervous system (CNS) substrates can induce their differentiation into in vivo-like microglial subsets. Exposing the iMGLs to apoptotic neurons, or Aβ amyloid fibrils, triggers their differentiation into a state highly analogous to disease-associated microglia (DAMs) seen in vivo. By utilizing virus-like particles containing Vpx, which prevents degradation of deoxynucleoside triphosphates, the authors dramatically augment lentiviral transduction of iMGLs, allowing for manipulation of iMGLs through gene expression. They further validate this platform by utilizing it to identify melanocyte-inducing transcription factor as a critical regulator of DAM gene expression and phagocytosis. Lastly, they demonstrate, as proof of concept, that this platform can be utilized to study samples derived from multiple donors in parallel.

    To date, the study of microglia in vitro mainly has mainly relied on the use of microglial cell lines or on primary microglial cultures. However, these approaches are hindered by their inability to accurately model in vivo microglial states. Additionally, microglia are traditionally not amenable to manipulation with viral vectors, which limits in vitro microglial studies to mostly lower-throughput methods. The model in this study overcomes these technical challenges, as Dolan and colleagues demonstrate from multiple angles. While applications of this technology remain to be seen, it opens possibilities that are certainly very exciting.

    In recent years, microglia have increasingly been shown to play important roles in pathologies of the CNS, particularly in neurodegenerative conditions such as Alzheimer’s disease (AD) (Song and Colonna., 2018). Studies in both mice and humans have demonstrated a microglia subtype in AD, commonly referred to as DAM, marked by expression of a specific set of genes, including APOE, LPL, and TREM2, among others, and the concurrent downregulation of homeostatic microglial signatures, including P2RY12, CX3CR1, etc. (Zhou et al., 2020Keren-Shaul et al., 2017). Dolan et al., in addition to providing a new method for studying microglia, demonstrated induction of DAM signatures in stem cell-derived microglia exposed to Aβ fibrils or apoptotic neurons, which are elements of the microglial microenvironment in AD. Taken together, these findings suggest an intimate relationship between formation of the DAM subset and hallmarks of AD neuropathology such as Aβ plaques.

    Whether DAMs are the result of pathology, or an active anti-AD cell type remains controversial. DAMs express multiple immune sensing signatures, which suggest an active role in restricting pathology (Deczkowska et al., 2018). On the other hand, a recent study has suggested that DAMs emerge as a result of excessive microglial proliferation, which results in replicative senescence (Hu et al., 2021). Furthermore, it has been suggested that DAMs are not the sole microglia subtype responsible for controlling AD pathology; this is evident by enrichment of AD polygenic risk gene expression across different microglial subsets

    The DAM subset was notable at the time of discovery, in part, because it was associated with disease restriction in AD mouse models. As such, microglia have been proposed to be a therapeutic target in AD. DAM-specific, high-throughput screens using the platform proposed by Dolan et al. will allow us to further understand the role of DAMs in AD, in addition to potentially uncovering druggable molecular targets. In vivo validation of hits found by these screening models will help clarify how, and whether, to target DAMs in AD, as well as determine which, if any, microglial state is desirable for disease restriction.


    . The identity and function of microglia in neurodegeneration. Nat Immunol. 2018 Oct;19(10):1048-1058. Epub 2018 Sep 24 PubMed.

    . 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.

    . 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.

    . Disease-Associated Microglia: A Universal Immune Sensor of Neurodegeneration. Cell. 2018 May 17;173(5):1073-1081. PubMed.

    . Replicative senescence dictates the emergence of disease-associated microglia and contributes to Aβ pathology. Cell Rep. 2021 Jun 8;35(10):109228. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Hot DAM: Specific Microglia Engulf Plaques
  2. ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
  3. When It Comes to Alzheimer’s Disease, Do Human Microglia Even Give a DAM?
  4. Human and Mouse Microglia React Differently to Amyloid
  5. What Makes a Microglia? Tales from the Transcriptome
  6. Chimeric Mice: Can They Model Human Microglial Responses?
  7. Human Microglia Make Themselves at Home in Mouse Brain
  8. Attack of the Clones? Memory CD8+ T Cells Stalk the AD, PD Brain
  9. The Human Brain Hosts a Menagerie of Microglia
  10. Most Detailed Look Yet at Activation States of Human Microglia
  11. TREM2 Risk Variant Eggs on Clique of Microglia in AD Brain

Research Models Citations

  1. APP NL-G-F Knock-in

Paper Citations

  1. . 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.
  2. . Distinct amyloid-β and tau-associated microglia profiles in Alzheimer's disease. Acta Neuropathol. 2021 May;141(5):681-696. Epub 2021 Feb 20 PubMed.
  3. . Targeting microglia with lentivirus and AAV: Recent advances and remaining challenges. Neurosci Lett. 2019 Aug 10;707:134310. Epub 2019 May 31 PubMed.
  4. . Diverse Brain Myeloid Expression Profiles Reveal Distinct Microglial Activation States and Aspects of Alzheimer's Disease Not Evident in Mouse Models. Cell Rep. 2018 Jan 16;22(3):832-847. PubMed.

Further Reading

Primary Papers

  1. . A multi-pronged human microglia response to Alzheimer’s disease Aβ pathology. bioRXiv. 7 Jul 2022 bioRxiv
  2. . A resource for generating and manipulating human microglial states in vitro. bioRxiv. May 2, 2022. bioRxiv

Follow-On Reading


  1. . Exposure of iPSC-derived human microglia to brain substrates enables the generation and manipulation of diverse transcriptional states in vitro. Nat Immunol. 2023 Aug;24(8):1382-1390. Epub 2023 Jul 27 PubMed.
  2. . Stem-cell-derived human microglia transplanted into mouse brain to study human disease. Nat Protoc. 2021 Feb;16(2):1013-1033. Epub 2021 Jan 11 PubMed.