Article
Published: 15 April 2021

Microglia use TAM receptors to detect and engulf amyloid β plaques

Youtong Huang^1,2,
Kaisa E. Happonen¹,
Patrick G. Burrola¹,
Carolyn O’Connor³,
Nasun Hah⁴,
Ling Huang⁵,
Axel Nimmerjahn ORCID: orcid.org/0000-0002-0875-7855⁶ &
…
Greg Lemke ORCID: orcid.org/0000-0001-5805-9855¹

Nature Immunology volume 22, pages 586–594 (2021)Cite this article

25k Accesses
194 Citations
250 Altmetric
Metrics details

Subjects

Abstract

Two microglial TAM receptor tyrosine kinases, Axl and Mer, have been linked to Alzheimer’s disease, but their roles in disease have not been tested experimentally. We find that in Alzheimer’s disease and its mouse models, induced expression of Axl and Mer in amyloid plaque–associated microglia was coupled to induced plaque decoration by the TAM ligand Gas6 and its co-ligand phosphatidylserine. In the APP/PS1 mouse model of Alzheimer’s disease, genetic ablation of Axl and Mer resulted in microglia that were unable to normally detect, respond to, organize or phagocytose amyloid-β plaques. These major deficits notwithstanding, TAM-deficient APP/PS1 mice developed fewer dense-core plaques than APP/PS1 mice with normal microglia. Our findings reveal that the TAM system is an essential mediator of microglial recognition and engulfment of amyloid plaques and that TAM-driven microglial phagocytosis does not inhibit, but rather promotes, dense-core plaque development.

You have full access to this article via your institution.

Identification of senescent, TREM2-expressing microglia in aging and Alzheimer’s disease model mouse brain

Article 18 April 2024

APOE4/4 is linked to damaging lipid droplets in Alzheimer’s disease microglia

Article Open access 13 March 2024

The Amyloid-β Pathway in Alzheimer’s Disease

Article Open access 30 August 2021

Main

The TAM receptor tyrosine kinases (RTKs) Axl and Mer (gene name Mertk) exert two key functions in macrophages and other immune sentinels^1,2. First, they are required for the phagocytosis of apoptotic cells (ACs)^3,4,5. Mer, which is expressed by all phagocytic macrophages, is especially important for this process³. Second, both receptors, but most prominently Axl, function as negative feedback inhibitors of Toll-like and cytokine receptor signaling in dendritic cells and macrophages^1,2,6. We have previously shown that these two functions are TAM-controlled in microglia, the tissue macrophages of the central nervous system (CNS)⁷. As genetic analyses have revealed that many of the genes associated with increased risk for Alzheimer’s disease (AD) and other neurodegenerative disorders are expressed in the CNS predominantly or exclusively by microglia, these macrophages have attracted increasing interest in the context of neurodegeneration⁸.

Human and mouse microglia normally express high levels of Mer and low levels of Axl^7,9 and we have shown that these RTKs play a critical role in microglial phagocytosis of the ACs that are generated during adult neurogenesis⁷. The third TAM receptor, Tyro3, is not expressed by microglia^7,9, but is prominently expressed by neurons¹⁰. We and others have also shown that Axl gene expression is markedly upregulated in microglia and other macrophages by inflammatory stimuli⁵ and that Axl upregulation also occurs when macrophages are activated during the course of infection, tissue trauma or disease^3,7. Correspondingly, RNA-sequencing (RNA-seq) analyses of laser-captured plaque-associated tissue from human AD and single-cell RNA-seq (scRNA-seq) analyses of disease-associated microglia (DAM) in the 5xFAD mouse model of AD, have both documented upregulation of microglial Axl messenger RNA in late-stage disease^11,12. The Axl tyrosine kinase is activated by Axl ectodomain binding of the ligand Gas6, which is in turn bound by the co-ligand phosphatidylserine¹³. Subsequent to Axl activation, metalloproteases cleave the Axl ectodomain from the cell surface⁵ and elevated cerebrospinal fluid levels of the soluble ectodomain (sAxl), in complex with Gas6, are predictive of disease development and progression in human AD^14,15. All of these observations have led to the classification of Axl as an ‘AD-associated gene’¹². This notwithstanding, the roles that Axl and Mer might play in AD have not been studied.

We therefore examined the function of microglial TAM receptors and their ligands in AD and its mouse models. We studied hemizygous APP/PS1 mice, which carry an amyloidogenic ‘Swedish’ mutation of the human amyloid precursor protein (APP) gene and a pathogenic mutation of the human presenilin 1 (PSEN1) gene^16,17, both as transgenes driven by the mouse prion protein promoter. In humans, these mutations result in familial AD, due to elevated proteolytic production of amyloid β (Aβ) peptides from the amino terminus of APP. We also examined hemizygous APP41 mice, which express a ‘Swedish’ + ‘London’ (V717I) mutant human APP under the control of the Thy-1 promoter¹⁸. Finally, we examined TAM regulation in human AD. We found that the full TAM system, including its universal ligand Gas6 and essential co-ligand phosphatidylserine, was engaged concomitantly with the deposition of Aβ plaques in the brain. Using scRNA-seq, we observed that Axl^−/−Mertk^−/− microglia display a blunted transcriptomic response to Aβ plaques in APP/PS1 mice. Using single-cell live two-photon imaging of compound mutant mice, we found that Axl^−/−Mertk^−/− microglia are also unable to normally detect, respond to, organize or phagocytose plaques. Remarkably, we observed that TAM-deficient APP/PS1 mice have fewer dense-core plaques than their TAM-expressing counterparts. We propose a model to account for these results.

Results

Engagement of TAM receptors during disease

We detected pronounced elevation of microglial Axl protein in APP/PS1 mice (Fig. 1). This induction was tightly plaque-associated and was not seen before the deposition of Aβ plaques. We identified microglia as ramified, multipolar cells expressing Iba1 (ionized calcium-binding adaptor molecule 1) or green fluorescent protein (GFP) in Cx3cr1-eGFP mice⁷. Perivascular macrophages are also GFP⁺ in these mice, but are readily distinguished from microglia. Iba1 and Cx3cr1 are also expressed by peripheral myeloid cells⁹ that may enter the brain and assume microglia-like properties in neurodegenerative disease¹⁹. Microglia near to but not touching plaques, identified with the 6E10 antibody to the amino terminus of human APP, were Axl⁻ in aged APP/PS1 mice, whereas those contacting plaques were strongly Axl⁺ (Fig. 1a). Notably, Axl protein upregulation is also seen in plaque-associated microglia in human AD²⁰. Quantification of Axl fluorescence in plaque-associated microglia (PAM) versus nonplaque-associated microglia (NPAM) over the course of disease in APP/PS1 cortex revealed a modest Axl increase in NPAM between 7 and 12 months (mo) (Fig. 1b). However, robust upregulation only occurred in plaque-associated cells (Fig. 1b). We did not detect appreciable plaque accumulation until ~5 mo in APP/PS1 mice and did not measure microglial Axl upregulation before this age (Fig. 1b and Extended Data Fig. 1a). Axl is not detectable by immunohistochemistry in healthy wild-type (WT) microglia⁷ (Extended Data Fig. 1a). At 9.5 mo, Axl levels were ~25-fold higher in PAM versus NPAM (Fig. 1b). We detected a similar Axl upregulation in PAM in aged APP41 mice, in which Axl was also frequently concentrated at the center of plaques (Extended Data Fig. 1b). As Axl activation is always accompanied by cleavage of the Axl ectodomain⁵, this plaque-deposited Axl is almost certainly sAxl.

**Fig. 1: TAM receptor upregulation is strictly plaque-associated.**

In contrast to Axl, Mer is abundant in all normal human and mouse microglia^7,9. Although modest reductions in Mertk mRNA have been reported for PAM in APP/PS1 mice²¹, Mer protein expression was upregulated in PAM in aged APP/PS1 (Fig. 1c) and APP41 (Extended Data Fig. 1c) mice. In the 9.5 mo APP/PS1 brain, we quantified a 3.5-fold elevation in Mer in PAM versus NPAM (Extended Data Fig. 1d). Axl⁺Mer^hi microglia were the same plaque-associated cells that also upregulated Trem2 (Fig. 1d and Extended Data Fig. 1e), mutations in which increase the risk of developing human AD²². Although upregulated Axl, Mer and Trem2 were always present in the same cells, Trem2 and TAMs seemed to occupy distinct membrane domains (Fig. 1d). Most Iba1⁺Axl⁺Mer^hiTrem2⁺ microglia in both APP/PS1 and APP41 mice were negative for TMEM119, a marker of homeostatic microglia²³, although 1–2 cells at the very center of plaques were occasionally TMEM119⁺ (Extended Data Fig. 1f,g). Plaque-associated astrocytes display reactive astrocytosis in AD and its mouse models, but GFAP⁺ and S100b⁺ reactive astrocytes expressed no detectable Mer (Extended Data Fig. 1h and ref. ⁷). Together, these results demonstrate that the concerted expression of Axl, Mer and Trem2 is specific to PAM.

Engagement of TAM ligand and co-ligand during disease

Expression of the obligate Axl and pan-TAM ligand Gas6 (ref. ¹³) was also strongly upregulated at plaques in aged AD mice (Fig. 2a). This upregulation was missed in earlier scRNA-seq analyses of mouse AD models because microglial expression of Gas6 mRNA, which is already high in normal microglia⁹, does not change over the course of disease in these models¹¹. We found that the presence of plaque-associated Gas6 in APP/PS1 mice was tied to Axl upregulation and to an unusual feature of Gas6–Axl interaction; it has been shown that most of the Axl on cells in WT mouse tissues is constitutively bound by Gas6 and that expression of Gas6 protein in these tissues is Axl-dependent⁵. Gas6 expression is lost in Axl^−/− but not Mertk^−/− tissues, even though Gas6 mRNA levels are unchanged between Axl^−/−, Mertk^−/− and WT tissues⁵. We therefore conclude that Gas6 protein is undetectable in nondiseased brains because these brains express little or no Axl (Extended Data Fig. 1a); and further, that acquired Axl in PAM leads to acquired plaque-localized Gas6. Consistent with this conclusion, plaque-associated Gas6 was sharply reduced in APP/PS1Axl^−/− and eliminated entirely in APP/PS1Axl^−/−Mertk^−/− mice (Fig. 2a,b).

**Fig. 2: Gas6 and phosphatidylserine decorate all Aβ plaques.**

Amyloid plaques did not develop and Gas6 did not appear in aged Axl^−/−Mertk^−/− mice without the APP/PS1 allele (Fig. 2a) and Gas6 mRNA levels were constant across WT, APP/PS1, Axl^−/−Mertk^−/− and APP/PS1Axl^−/−Mertk^−/− cortices at 12 mo (Fig. 2b). We also detected expression of Gas6 surrounding plaques in sections of human AD brain at Braak stage 6, but not in age-matched controls (Extended Data Fig. 2a), indicating that acquired Gas6 expression also marks plaques in human AD. Gas6 activates both Axl and Mer, but Mer can also be activated by the related ligand Protein S (Pros1)¹³. Although antibodies for high-resolution immunohistochemical detection of Pros1 in mouse are not currently available, microglia express abundant Pros1 mRNA⁹ and deep proteomic profiling of the 5xFAD hippocampus has revealed that elevated Pros1 is a new biomarker for disease severity in this model²⁴. It is therefore very possible that plaques are also decorated with Pros1.

In addition to ligands and receptors, TAM signaling requires binding of the phospholipid phosphatidylserine (PtdSer) to the amino-terminal ‘Gla’ domains of Gas6 and Pros1 (refs. ^3,13). For example, a ‘Gla-less’ truncation mutant of Gas6 binds Axl with the same affinity as the full-length ligand, but is incapable of activating the receptor¹³. With respect to TAM receptor interaction, PtdSer is generally externalized on the surface of a cell (often an AC) that is apposed to a TAM-expressing cell^1,3. Although PtdSer is a component of every plasma membrane, in all normal cells it is enzymatically confined to the inner leaflet of the membrane bilayer³. Translocation of PtdSer to the plasma membrane surface is carried out by a set of PtdSer scramblases³ during apoptosis and in response to Ca²⁺ entry, viral infection and disease stressors³. During AC phagocytosis, Gas6 and/or Pros1 ‘bridge’ PtdSer that is externalized on the AC surface to TAM receptors expressed on the surface of microglia and other phagocytes³. To detect externalization of PtdSer in APP/PS1 mice, we stereotaxically injected their cortices with pSIVA, a ‘polarity-sensitive indicator for viability and apoptosis’ that fluoresces strongly only when bound to PtdSer²⁵. The needle track of this injection (where cells were damaged) was strongly pSIVA⁺, but radiating out from the injection site we detected coincident pSIVA labeling of PtdSer on all 6E10⁺ Aβ plaques (Fig. 2c and Extended Data Fig. 2b). No pSIVA puncta were labeled outside the needle track after injection of WT brains (Extended Data Fig. 2c). Externalized plaque-associated PtdSer may mark the dystrophic neurites that are present in plaques and their immediate surroundings. Super-resolution microscopy suggested that Gas6 does indeed bridge PtdSer⁺ Aβ plaques to TAM⁺ microglia in the aged APP/PS1 brain (Extended Data Fig. 2d). Together, these results demonstrate that all of the essential components of the TAM system (the receptors, their ligand Gas6 and the essential co-factor PtdSer) are specifically upregulated on and around Aβ plaques in AD.

TAM regulation of the microglial transcriptome in AD

To obtain a global perspective on how TAM signaling influences gene expression in PAM and NPAM, we performed comparative scRNA-seq of CD45⁺ cells sorted (Extended Data Fig. 3a) from plaque-burdened cortices at ~18 mo, when plaques were abundant, using a 10X Genomics platform. Several previous RNA-seq analyses have been performed with human and mouse microglia in neurodegenerative settings^11,21,26 and these provide data for comparison with our results. We performed high-throughput single-cell transcriptomics of a total of 14,628 CD45⁺ immune cells from APP/PS1 and APP/PS1Axl^−/−Mertk^−/− cortices and detected nine immune clusters by applying Seurat analyses²⁷ (Extended Data Fig. 3b). Annotation using 18 marker genes (Extended Data Fig. 3c) allowed us to associate these clusters with specific cell types, of which microglia comprised the largest cluster. Also present in the CD45⁺ sort were natural killer and T cells, neutrophils, monocytes and nonmicroglial macrophages and smaller populations of B cells, oligodendrocytes, endothelial cells and proliferating cells of various lineages (Extended Data Fig. 3b).

In a second clustering round, we identified seven microglial clusters and numbered and staged these in the APP/PS1 and APP/PS1Axl^−/−Mertk^−/− replicates (Fig. 3a). Clusters were numbered based on transcriptomic state, with the largest cluster 0 (c0; Fig. 3a) representing the homeostatic microglial ground state. Mertk mRNA was distributed across all clusters, whereas Axl mRNA was largely restricted to clusters 4–6 (Extended Data Fig. 3d). Prominently expressed mRNAs in cluster 0 included Tmem119, Cx3cr1, P2ry12 and Maf (Fig. 3b and Extended Data Fig. 3e,f).

**Fig. 3: TAM regulation of the microglial transcriptome in disease.**

Cluster 1 microglia seem to represent a small one-way radiation from cluster 0, in which mRNAs indicative of immune stimulation, including Cst7, Axl and type I interferon response genes (for example, Ifitm3), were upregulated (Fig. 3a,b). All remaining transcriptomic development from cluster 0 proceeded in order from clusters 2 through 6, with most microglia stopping at a fully activated cluster 5 (Fig. 3a,b and Extended Data Fig. 3e). By tracking microglial marker expression (Fig. 3b and Extended Data Fig. 3e,f), we determined that clusters 2–4 represented cell sets in sequential transition states between the predominant clusters 0 and 5. These predominant clusters were largely equivalent to the homeostatic and DAM cells, respectively, defined for 5xFAD microglia¹¹ (Extended Data Fig. 3f). Although cluster 2 microglia exhibit a transcriptomic phenotype distinct from cluster 0, with diminished expression of Tmem119 and P2ry12 and increased expression of Apoe and Clec7a, among others (Fig. 3b), these cells mostly fell into the homeostatic cluster of Keren-Shaul et al.¹¹ (Extended Data Fig. 3f). Similarly, cells in clusters 3 and 4 substantially, but not entirely, resembled those of 5xFAD DAM stage 1, whereas cells in cluster 5 resembled those of DAM stage 2 (ref. ¹¹) (Extended Data Fig. 3f). RNAs abundant in cluster 5 cells included Cst7, Lpl, Trem2 and Csf1 (Extended Data Fig. 3e). The small cluster 6 seems to represent a further activation of cluster 5, in which microglia are primed for antigen presentation, in that cells in this cluster expressed elevated levels of mRNAs for the major histocompatibility complex (MHC) class II trafficking adaptor CD74 and for multiple MHC class II proteins (Fig. 3b). This is of interest given evidence for T cell regulation of neuroinflammation in AD²⁸.

Three features of microglial transcriptomic evolution in APP/PS1 versus APP/PS1Axl^−/−Mertk^−/− mice were apparent. First, we found that the same seven clusters were equivalently represented in the APP/PS1 and APP/PS1Axl^−/−Mertk^−/− replicates (Fig. 3c), indicating that Axl and Mer expression does not influence the evolution of microglial transcriptomic states in APP/PS1 mice. This is in contrast to the mutation of Trem2 in 5xFAD mice, which blocks the transition of cells to a maximally activated DAM stage 2 state¹¹. Notably, one of the mRNAs prominent in DAM stage 2 is Axl, whose full microglial expression has been found to be Trem2-dependent¹¹. While Trem2 protein is higher in PAM versus NPAM (Fig. 1d), Trem2 mRNA is only modestly increased in state 5 versus state 0 microglia (Fig. 3b) and expression of this mRNA is unaffected by mutation of Axl and Mertk.

Second, we observed that both the statistical significance and magnitude of induction of many mRNAs upregulated in cluster 5 versus cluster 0 microglia was blunted upon mutation of Axl and Mertk (Fig. 3d,e and Extended Data Fig. 3g). This effect were seen for many DAM signature genes, including Spp1, Cd74 and Gpnmb. It was also apparent for genes involved in lipid metabolism (Lpl, Apoc4, Ch25h, Apoe) and MHC class II antigen presentation (H2-Aa, H2-Ab1, H2-Eb1 and Cd74) and for the ‘plaque-forming’ gene Ccl6 (ref. ²⁹). Although RNA for the phagocytosis regulator SIRPα was reduced in both transcriptomic states 5 and 0 in APP/PS1Axl^−/−Mertk^−/− microglia (Extended Data Fig. 3g), the principal phagocytic mediators effected by Axl and Mertk mutation were Axl and Mer themselves. (The Mertk mutation deletes an essential exon in the kinase domain and produces a protein null, but results in the synthesis of a nonfunctional mRNA that is detectable by PCR, RNase protection and RNA-seq^5,30.) Conversely, we found that both the statistical significance (Fig. 3d) and magnitude (Fig. 3e and Extended Data Fig. 3g) of downregulation of many genes reduced in cluster 5 relative to cluster 0 was dampened in APP/PS1Axl^−/−Mertk^−/− relative to APP/PS1 microglia. These included the homeostatic genes Malat1 and Nfkbia (Fig. 3d,e and Extended Data Fig. 3g).

Finally, we did not observe large-scale upregulation of cytokine or chemokine mRNAs in state 5 versus state 0 microglia (Extended Data Fig. 3h) or in total mRNA purified from whole 12-mo cortex (Extended Data Fig. 3i), as a consequence of the combined loss of both microglial TAMs. Thus, it is unlikely that the phenotypes documented below result from major changes in APP/PS1 neuroinflammation consequent to Axl and Mertk mutation.

Microglial TAM activity at plaques

To determine whether the TAM system perturbations documented above are of significance to microglial interaction with plaques, we used live two-photon imaging in APP/PS1 mice at 16 mo. We analyzed APP/PS1 hemizygotes that carried a single Cx3cr1^GFP allele³¹ to label microglia and then crossed these mice with both WT and Axl^−/−Mertk^−/− mutants. We labeled Aβ plaques using intraperitoneal injection of methoxy-X04 (ref. ³²) (MX04), a fluorescent amyloid dye and then live-imaged 4–7 plaque-containing volumes per mouse in layers 1/2 of somatosensory cortex for ~90 min, using modifications of two-photon methods described previously^7,33. Representative time-lapse recordings of these analyses are shown in Supplementary Videos 1 and 2.

Quantification of microglial behavior allowed us to establish multiple differences in plaque recognition and interaction between WT and Axl^−/−Mertk^−/− microglia. First, we observed many fewer microglia encapsulating Aβ plaques in Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− versus Cx3cr1^GFPAPP/PS1 mice (Fig. 4a). Most plaques were enveloped by tightly bound and morphologically activated microglia in the latter (Fig. 4a), whereas plaques were often unattended by microglia in Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− mice (Fig. 4a). Surface building reconstruction and quantification of imaging volumes (Fig. 4b) yielded a flat distribution of microglia cell body distances from plaques in Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− mice, but showed a strong bias for microglia to be in contact with plaques in Cx3cr1^GFPAPP/PS1 mice (Fig. 4c). This difference was independent of plaque size, although larger plaques were bound by more microglia in Cx3cr1^GFPAPP/PS1 mice (Fig. 4c and Extended Data Fig. 4a).

**Fig. 4: Microglia use Axl and Mer to detect, engage and react to Aβ plaques.**

Similarly, there were obvious differences in microglial morphology and motility (Fig. 4a,d–g). Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− microglia often displayed a ramified ‘resting’ configuration that was similar to that seen in WT mice (Fig. 4a and Supplementary Video 1). In contrast, Cx3cr1^GFPAPP/PS1 microglia displayed an activated ‘ameboid’ morphology, with larger cell bodies and fewer, shorter processes (Fig. 4a and Supplementary Video 1). Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− PAM elaborated substantially more (Fig. 4d and Extended Data Fig. 4b) and longer (Fig. 4e and Extended Data Fig. 4c) primary processes than Cx3cr1^GFPAPP/PS1 PAM.

Processes from microglia positioned >20 μm from plaques (NPAM) were strongly oriented toward plaques in the Cx3cr1^GFPAPP/PS1 cortex, but showed no orientation bias in the Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− cortex (Fig. 4f and Extended Data Fig. 4d). In addition, microglial process motility, a reduction in which is indicative of microglial activation³⁴, was markedly reduced in APP/PS1 mice relative to WT mice (Fig. 4g). In contrast, process motility was only modestly lower than WT for APP/PS1Axl^−/−Mertk^−/− PAM (microglia < 5 μm from the nearest plaque) (Fig. 4g) and was not statistically different from WT for APP/PS1Axl^−/−Mertk^−/− NPAM (Fig. 4g). Finally, AD microglia normally proliferate near plaques. Mutation of Trem2 in the 5xFAD mouse model blunts this microgliosis ~twofold^35,36 and we measured a similarly blunted proliferation of microglia, primarily those in close proximity to plaques, in the cortex of Cx3cr1^GFPAPP/PS1Axl^−/−Mertk^−/− relative to Cx3cr1^GFPAPP/PS1 mice (Extended Data Fig. 4e). Thus, by all of the above measures, TAM-deficient microglia were strikingly deficient in their ability to detect, bind and respond to amyloid plaques.

TAM-dependent phagocytosis and shaping of plaques

Microglia are vigorous phagocytes that exhibit an absolute TAM receptor requirement for AC engulfment⁷. As TAM-mediated phagocytosis of ACs is entirely dependent on PtdSer externalization on the AC surface^3,13 and because all Aβ plaques are decorated with both PtdSer (Fig. 2c) and Gas6 (Fig. 2a,b), we used two-photon imaging to measure engulfment of MX04-labeled amyloid plaque material into GFP⁺ microglia that were either WT or Axl^−/−Mertk^−/−. We quantified internalized MX04 signals both per normalized volume and per GFP volume. These measurements revealed a clear deficit; we quantified ~tenfold lower levels of engulfed MX04-labeled plaque material inside of Axl^−/−Mertk^−/− as compared to WT microglia by both measures (Fig. 5a,b). Note that internalized amyloid material, which may be routed to acidic lysosomes, occupies a remarkable 9% of WT microglial volume in APP/PS1 mice (Fig. 5b). These data indicate that microglia cannot effectively phagocytose amyloid material without TAM receptors.

**Fig. 5: TAM-deficient microglia can neither phagocytose nor organize plaques.**

Amyloid plaques are typically surrounded by a halo of dystrophic neuronal membranes³⁷. This halo is shaped by microglial phagocytic activity³⁷ and is marked by the expression of several proteins, including the autophagic and endolysosomal vesicular marker LAMP1 (ref. ³⁸) (examples in Fig. 5c). When we measured the area ratio of LAMP1⁺ membranes to 6E10⁺ plaques (examples in Extended Data Fig. 5), we quantified a tenfold increase in this ratio in APP/PS1Axl^−/−Mertk^−/− versus APP/PS1 mice at 12 mo (Fig. 5d). This difference was primarily driven by an expansion of LAMP1 around weakly staining, diffuse plaques (Fig. 5c), but was also seen for dense-core plaques (Extended Data Fig. 6a,b). A second protein that displays a similar but more circumscribed distribution in dystrophic membranes is the endoplasmic reticulum protein reticulon-3 (RTN3)³⁹. When we performed an area ratio analysis for RTN3 surrounding 6E10-labeled plaques, we again measured a large increase in this ratio in APP/PS1Axl^−/−Mertk^−/− mice (Fig. 5e). This difference was also driven by a large expansion of RTN3 around weakly staining, diffuse plaques, but was also seen for dense-core plaques (Extended Data Fig. 6c,d). Inspection of 6E10⁺ profiles suggested that diffuse, poorly organized 6E10⁺ plaques were more common in APP/PS1Axl^−/−Mertk^−/− than in APP/PS1 mice (Extended Data Fig. 5) and quantification confirmed this impression (Extended Data Fig. 6e).

We found that these multiple plaque perturbations were accompanied a ~twofold increase, relative to APP/PS1, in the deposition of Aβ within and surrounding blood vessels in the APP/PS1Axl^−/−Mertk^−/− brain, indicative of cerebral amyloid angiopathy (CAA)⁴⁰ (Extended Data Fig. 6f,g). This increased CAA, which is a common feature of AD⁴⁰, is of interest given that a similar increase has been seen when microglia are depleted in 5xFAD mice²⁹. Finally and in keeping with the phagocytic deficits documented above, we always detected uncleared cCasp3⁺ apoptotic debris surrounding 6E10⁺ plaques in aged APP/PS1Axl^−/−Mertk^−/− mice, but never in APP/PS1 mice (Extended Data Fig. 6h).

Dense-core plaque burden in the absence of TAM signaling

Microglial phagocytosis has heretofore been thought to inhibit the growth of plaques⁸. Thus, a prediction from the relative inability of APP/PS1Axl^−/−Mertk^−/− microglia to engulf Aβ material is that aged TAM-deficient AD mice should display a much higher plaque burden than AD mice with WT microglia. Our initial indication that this was not the case came from the APP/PS1Axl^−/−Mertk^−/− two-photon volumes that we recorded, most of which contained fewer MX04-labeled plaques than APP/PS1 volumes with WT microglia (examples in Fig. 6a). To measure plaque burden in these populations, we used semi-automated quantification of thioflavin S (Thio S)-stained dense-core plaques in whole-brain sections to assess the effect of Axl and Mer deletion on dense-core plaque accumulation, size and number across disease development. We also quantified dense-core plaques in APP/PS1Axl^−/− and APP/PS1Mertk^−/− single-compound mutants at 12 mo. Representative Thio S–stained sagittal sections of APP/PS1 and APP/PS1Axl^−/−Mertk^−/− brains at 12 mo are shown in Fig. 6b.

**Fig. 6: TAM-driven microglial phagocytosis favors dense-core plaque formation.**

Quantification revealed that, from initial plaque appearance to profuse deposition, APP/PS1Axl^−/−Mertk^−/− cortices displayed many fewer, not many more, dense-core plaques than APP/PS1 cortices (Fig. 6c). At 12 mo, APP/PS1Axl^−/−Mertk^−/− mice showed a 35% reduction in cortical Thio S⁺ plaque density relative to APP/PS1 (Fig. 6d). This reduction was seen across all plaque sizes, but was most pronounced for smaller plaques (Extended Data Fig. 7a). A similar reduction was seen in the hippocampus (Extended Data Fig. 7b). Notably, these reductions in dense-core plaque density were not due to any change in the production of Aβ peptide, as soluble Aβ levels in the cortex and hippocampus at 4 and 12 mo were not significantly different between APP/PS1 and APP/PS1Axl^−/−Mertk^−/− mice (Extended Data Fig. 7c). Similarly, no change was apparent between these genotypes in cortical expression of APP (Extended Data Fig. 7d).

When we quantified dense-core plaque burden in the cortex and hippocampus of single-compound mutants at 12 mo, we found that nearly all of the effect observed in the double mutants was due to mutation of Mer alone. APP/PS1Mertk^−/− mice displayed reductions in dense-core plaque burden that largely phenocopied those seen in APP/PS1Axl^−/−Mertk^−/− mice (Extended Data Fig. 7e), whereas APP/PS1Axl^−/− mice, while exhibiting a trend toward reduced plaque density in the hippocampus, displayed cortical plaque burdens that were equivalent to APP/PS1 (Extended Data Fig. 7f). Thus, while Axl is upregulated in PAM, most of the phagocytosis-dependent construction of dense-core plaques is performed by Mer, the receptor that is constitutively present in microglia. Together, these data demonstrate that TAM-mediated phagocytosis promotes the deposition of dense-core Aβ plaques (Extended Data Fig. 7g).

Consistent with the findings of equivalent APP and Aβ levels in APP/PS1 and APP/PS1Axl^−/−Mertk^−/− mice, we detected equivalent reductions in an excitatory synaptogenesis index (the physically paired expression of the pre- and post-synaptic markers vGlut1 and PSD95 (ref. ⁴¹)) between these genotypes and WT mice within CA1 of the 15-mo hippocampus (Extended Data Fig. 8a). When we performed a fear-conditioning assay assessing cognitive deficits in WT, Axl^−/−Mertk^−/−, APP/PS1 and APP/PS1Axl^−/−Mertk^−/− mice, we detected a statistically significant deficit in fear acquisition in APP/PS1Axl^−/−Mertk^−/− relative to APP/PS1 mice (Extended Data Fig. 8b,c), but no significant difference in contextual fear memory between these genotypes (Extended Data Fig. 8d).

Discussion

Our results lead to two principal conclusions. First, the TAM system is required for microglial recognition of, response to and phagocytosis of Aβ plaques. And second, TAM-mediated microglial phagocytosis of Aβ material does not inhibit, but rather promotes, the formation of dense-core plaques.

The first of these conclusions is consistent with previous knowledge as to TAM action in microglia and other macrophages^3,5,7. All Aβ plaques are decorated with externalized PtdSer, an essential TAM co-ligand¹³ and this PtdSer is bound to the amino-terminal ‘Gla’ domain of Gas6 (ref. ¹³). The carboxy-terminal ‘SHBG’ domain of this ligand^1,13 is then bound to Axl and Mer on microglia. In this arrangement, the TAM ligand serves as a bridge between microglia and an Aβ plaque in the same way that TAM ligands bridge macrophages to PtdSer-expressing ACs during engulfment^1,3,4,5,7,13. In the absence of Axl and Mer, microglia display a diminished transcriptomic response to plaques, are compromised in their attachment to plaques, do not re-orient their processes toward plaques and are blunted in their proliferative, process extension and process motility responses to plaques. Most notably, Axl^−/−Mertk^−/− microglia cannot effectively phagocytose Aβ plaque material.

There are some similarities between these phenotypes and those seen when Trem2^−/− mice are crossed into amyloidogenic AD mouse models^{22,35,36,42,43,44}, but in general the TAM phenotypes are far stronger. The closest similarities are seen with respect to microglial proliferation and plaque binding. Microglial proliferation around plaques is reduced ~twofold in Trem2^−/− compound mutants^36,42 and also ~twofold for Axl^−/−Mertk^−/− compound mutants. Similarly, the number of microglia bound per plaque is reduced two- to fourfold in Trem2^−/− compound mutants^35,36,42,43 and ~fivefold in the Axl^−/−Mertk^−/− compound mutants. Loss of Trem2 from the APP/PS1-21 line results in a ~3.5-fold increase in the volume of LAMP1⁺ membranes surrounding ‘filamentous’ plaques and a ~1.5-fold increase for compact plaques in late-stage disease⁴⁴, whereas these numbers for the APP/PS1Axl^−/−Mertk^−/− mice are ~eightfold and ~fourfold at a comparable stage of disease. As dystrophic neurites arise around plaques devoid of microglial coverage³⁷, the paucity of microglial plaque association seen in APP/PS1Axl^−/−Mertk^−/− mice may result in collateral axonal damage.

There is no consensus among research groups as to the effect of Trem2 deletion on Aβ plaque burden in amyloidogenic AD models^22,42, with different groups reporting either marginal increases or decreases or no changes, in plaque density (measured with Thio S, 6E10 or amyloid dyes) that vary with the model employed, age analyzed and brain region assessed. In contrast, APP/PS1Axl^−/−Mertk^−/− mice display a consistent reduction in dense-core plaque density, in both the cortex and the hippocampus, across the full course of disease in APP/PS1 mice. Most notably, Trem2^−/− microglia exhibit a modest ~twofold reduction in their ability to phagocytose MX04-labeled Aβ material³⁶ and display normal phagocytosis of ACs in in vitro assays³⁵, whereas Axl^−/−Mertk^−/− microglia exhibit a tenfold deficit in their ability to phagocytose MX04-labeled Aβ material in vivo and are incapable of any AC phagocytosis in vitro⁷. Together, these results argue that Mer and Axl are the main receptors that microglia use to detect, engage and engulf amyloid plaques. Indeed, it is possible that some of phenotypes observed in Trem2^−/− mice are in part TAM-mediated, as scRNA-seq analyses have indicated that upregulation of Axl mRNA in the 5xFAD model is Trem2-dependent¹¹.

The second of our principal conclusions is contrary to expectation. How does a tenfold reduction in phagocytic capacity, coupled with a twofold reduction in microglial numbers and a fivefold reduction in plaque binding, result in 35% fewer dense-core plaques? This finding should be considered in light of two sets of previous analyses in 5xFAD mice^29,45,46. First, using two-photon imaging methods similar to ours, Baik and colleagues found that microglial deposition of previously phagocytosed Aβ material, via exocytosis or microglial death, was essential for the growth of dense-core plaques⁴⁵. An acidic environment, as is found in the lysosomes into which engulfed Aβ material is deposited⁴⁷, is known to promote the formation of densely packed, protease-resistant Aβ fibrils^45,48. As noted above, we find that internalized Aβ material accounts for nearly 10% of WT microglial volume in APP/PS1 mice and studies in another amyloidogenic model have reported even larger Aβ-laden lysosomal volumes within these cells²⁹. This suggests that once routed to lysosomes, Aβ fibrils are compacted into dense-core material that is ‘indigestible’.

A second key observation is provided by studies in which microglia were killed by pharmacological inhibition of the CSF1 receptor^29,46. These studies demonstrated that plaques never appear in the 5xFAD brain when microglia are killed, except in the limited regions where microglia are spared from death²⁹. Together, these findings and our results suggest that dense-core Aβ plaques do not form spontaneously, but are instead constructed from loosely-organized Aβ material by phagocytic microglia and that TAM receptors are key components of the molecular machinery through which this is achieved. They are consistent with the hypothesis that dense-core plaques may represent a macrophage-mediated confinement mechanism, perhaps analogous to the granulomas of tuberculosis and other infections⁴⁹, which limits the dissemination of toxic pre-plaque Aβ oligomers throughout the brain. This may in part explain why agents that disaggregate dense-core plaques but do little to alter the production of Aβ peptides have largely failed as AD therapeutics^8,50.

Methods

Mice

C57BL/6J WT mice were obtained from The Jackson Laboratory. Mice were typically group-housed at approximately 22 °C and provided with bedding and nesting material. All animals were maintained on a 12 h light–dark cycle and given ad libitum access to standard rodent chow and water. The Axl^−/−, Mertk^−/−, Axl^−/−Mertk^−/−³⁰ and Cx3cr1^GFP/+³¹ strains have been described previously. B6.Cg-Tg(APPSwePSEN1dE9) hemizygous mice^16,17 (APP/PS1) (JAX no. 005864) were crossed with Axl^−/−, Mertk^−/− and/or Axl^−/−Mertk^−/− lines to generate APP/PS1Axl^−/−, APP/PS1Mertk^−/− and APP/PS1Axl^−/−Mertk^−/− mice. For two-photon microscopy, APP/PS1 mice were crossed with Cx3cr1^GFP/GFP or Cx3cr1^GFP/GFP Axl^−/−Mertk^−/− strains. Only female APP/PS1Cx3cr1^GFP/+ WT or Axl^−/−Mertk^−/− and healthy littermates were used in two-photon studies to avoid potential sex biases in pathology and microglial responses. APP41 mice, which express a ‘Swedish’ + ‘London’ (V717I) mutant human APP under the Thy-1 promoter^18,51,52, were a kind gift of K.-F. Lee and J. Xu. All lines have been backcrossed for >ten generations to and maintained on a C57BL/6 background. All animal procedures were conducted in compliance with the ethical guidelines and according to protocols approved by the Salk Institute Animal Care and Use Committee (protocol no. 17-00009). Mice of both sexes were randomly allocated to experimental groups unless otherwise noted.

Reagents and antibodies

Antibodies used were as follows: anti-Axl (R&D AF854, 1:50 dilution), anti-Mer (eBioscience DS5MMER and R&D AF591, 1:200 dilution), anti-mouse Gas6 (R&D AF986, 1:50 dilution), anti-human Gas6 (R&D AF885, 1:25 dilution), anti-β-amyloid, clone 6E10 (BioLegend 803001, 1:500 dilution), anti-Iba1 (Wako 019-19741 and Novus NB100-1028, 1:250 and 1:200 dilution, respectively), anti-GFAP (Dako z-334, 1:500 dilution), cleaved Casp3 (Cell Signaling 9661, 1:200 dilution), anti-CD68 (BioRad MCA1957, 1:200 dilution), anti-RTN3 (EMD Millipore ABN1723, 1:500 dilution), anti-Tmem119 (Abcam AB209064, 1:200 dilution), anti-LAMP1 (BD Biosciences, 553792, clone 1D4B, 1:200 dilution) and anti-Trem2 (R&D systems, AF1729 and BAF 1729, 1:200 and 1:50 dilution, respectively), anti-CD31 (R&D, AF3628, 1:50 dilution), anti-laminin (Sigma, L-9393, 1:500 dilution). For flow cytometry, anti-mouse CD16/32 (BioLegend 101301, 1:100 dilution), anti-CD45-PE, 2D1 (Invitrogen 12-9459-42, 1:25 dilution), anti-CD11b-FITC, M1/70 (Invitrogen 11-0112-85, 1:25 dilution). Anti-vGlut1 (Millipore, AB5905) and anti-PSD95 (Life Technologies, 51-6900) were kind gifts of the N. Allen laboratory. Secondary antibodies for immunohistochemistry were all used at 1:500 dilution and were fluorophore-conjugated anti-rat (712- 545- 153 or 712- 165- 153 from Jackson ImmunoResearch), anti-goat (A-11056 from Life Technologies or 705-166-147 from Jackson ImmunoResearch), anti-rabbit (A-11071 or A-21206 from Life Technologies), anti-sheep (A21098 from Thermo Fisher Scientific), anti-mouse (A-11029 from Life Technologies, 715-166-150 or 715-176-150 from Jackson ImmunoResearch) and streptavidin 555 (Invitrogen S21381). For immunoblotting, anti-6E10 (see above, 1:1,000 dilution) and anti-GAPDH (Millipore MAB374, 1:10,000 dilution) were used as loading control. For Li-Cor detection, IRDye 680RG IgG secondary antibodies of the corresponding species were used. Antibodies used in the quantification of Aβ₄₂ peptides by ELISA were obtained as a kind gift from M. Mercken from Janssen Pharmaceuticals: JRF AB042/26 for Aβ₄₂ and detection antibody JRF/AbN/25, unlabeled. Key reagents and their usage were as follows: Thio S (Acros Organics 213150250); Hoechst 33258 in a 1% solution of bisbenzimide in water (Sigma Aldrich B-2883); TrueBlack Lipofuscin Autofluorescence Quencher (Biotium 23007). For biotinylation of detection antibody JRF/AbN/25 (EZ link Sulfo-NHS-LC Biotin, Thermo Fisher Scientific, 21217) and subsequent dialysis (Slide-A-lyzer dialysis cassette, Pierce 0728, 66415). Synthetic human Aβ_1-42 peptides for ELISA standards (Anaspec 20276). pSIVA-IANBD was included in the pSIVA apoptotic detection kit (Novus Bio NBP2-29382) originally described in Kim et al.²⁵. For in vivo amyloid plaque labeling, MX04 (Tocris 4920) was prepared and used as previously described³², following the manufacturer’s instructions. Stock MX04 was first dissolved in 100% dimethylsulfoxide then injected intraperitoneally at a final concentration of 10 mg kg⁻¹ in 10% dimethylsulfoxide, 45% propylene glycol and 45% PBS (pH 7.5) 24 h before imaging.

Immunohistochemistry and histochemistry

Routine perfusion procedures were performed⁷. Briefly, P30 and adult mice (2.5–16 mo) were anesthetized with a final concentration of 100 mg kg⁻¹ ketamine and 10 mg kg⁻¹ xylazine in accordance with institutional animal care and use committee guidelines. Mice were then transcardially perfused with 20 U ml⁻¹ heparin in PBS followed by freshly prepared 4% PFA in PBS. Both hemispheres were post-fixed in 4% PFA in PBS overnight at 4 °C and subsequently infiltrated in 30% sucrose in PBS for 1 d at 4 °C and flash-frozen in TBS tissue freezing medium. The left hemisphere was sagittally cryosectioned at 15 μm, air-dried overnight and subsequently processed for staining. Antigen retrieval with citrate buffer was performed if needed by heating brain sections to 80 °C for 3 min in pre-warmed citrate buffer and cooling to room temperature. Nonspecific binding was blocked by 1 h incubation in blocking buffer (PBS containing 0.3% Triton-X 100, 0.1% Tween-20, 5% donkey serum and 2% IgG-free BSA). Sections were incubated overnight at 4 °C with primary antibody diluted in blocking buffer, then washed in PBS 0.1% Tween-20 and incubated for 2 h at 22–24 °C in the dark with fluorophore-coupled secondary antibodies and subsequently in Hoechst for 5 min diluted in blocking buffer. Sections were washed, sealed with Fluoromount-G (SouthernBiotech) and stored at 4 °C.

Cortical and hippocampal plaque density was quantified by staining with the fluorescent amyloid dye Thio S. Thio S binds to the β pleated sheets of dense-core senile amyloid plaques and provides for more reliable semi-automated quantification of plaques than immunostaining with 6E10. A serial set of 15-μm sagittal sections (five sections per mouse) were obtained 0.25–0.85 mm from the midline, each spaced 0.15 mm apart. Plaque abundance was quantified by histochemistry with Thio S staining using published protocols. Image analysis of the number of cortical and hippocampal Thio S⁺ plaques was performed blind with respect to genotypes and in a semi-automated fashion with set intensity thresholds for all conditions, using the ‘analyze particles’ function on ImageJ or FIJI software. For evaluating pSIVA-injected APP/PS1 and WT mice, a serial set of 15-μm sagittal sections were obtained between 1–2 mm from the midline (0.5 mm before and after the corresponding pSIVA injection sites, 15 sections per mouse) from both injected and un-injected control hemispheres. Serially cut sections from both hemispheres and genotypes (APP/PS1 versus WT) were further immunostained with 6E10 antibody and evaluated for colocalization of fluorescence emitted from pSIVA and amyloid.

Human postmortem AD brain sections

Paraffin-embedded thin sections of postmortem human brain tissues were obtained from the University of California, San Diego (UCSD) Alzheimer’s Disease Research Center (ADRC), courtesy of R. Rissman. Postmortem tissues used in this project were collected from the UCSD ADRC neuropathology core. All participants consented to brain donation at the time of enrollment in the ADRC. Three age-matched pairs of AD BRAAK stage 6 and clinically healthy brain sections were evaluated. Sections underwent standard deparaffinization and permeabilization with 0.3% Triton-X 100 before treated for 5 min with 1× TrueBlack in 70% ethanol. Subsequent immunostaining for human Gas6 and Aβ plaques using the antibodies detailed above was conducted with the standard immunofluorescence protocol detailed above without detergents in the buffers.

Isolation and purification of brain immune single cells for sequencing

Mature APP/PS1 or APP/PS1Axl^−/−Mertk^−/− mice (18 mo) were transcardially perfused with ice-cold D-PBS containing Ca²⁺ and Mg²⁺ and brains were promptly dissected out and placed on pre-chilled Petri dishes on ice to extract cortices. Cortices from two mice were pooled for each sample and two biological samples per genotype were included in our study. All steps were carried out on ice, unless otherwise specified. Pooled cortices were minced with a razor blade, suspended in 5 ml D-PBS in 15-ml tubes for tissue chunks to settle before removal of the supernatant. Single-cell suspensions were prepared following a modified version of the Neural Tissue Dissociation kit from Miltenyi Biotec (130-094-802). Briefly, for each sample, an enzyme mixture 1 containing 50 μl enzyme P in 1,910 μl buffer Z supplemented with 0.2 μg DNase (Sigma) and an enzyme mixture 2 containing 30 μl Buffer Y and 15 μl of enzyme A were prepared and these enzyme mixtures were kept on ice before usage. The settled tissue pellet was resuspended in mixture 1 and horizontally agitated at room temperature for 15 min. Thirty microliters of enzyme mixture 2 was added before mechanical dissociation on ice using a fire-polished Pasteur pipette. The cell suspension was further incubated at room temperature with constant rotation for 10 min, after which the remaining mixture 2 was added. Samples were subjected to a second round of mechanical dissociation on ice with fire-polished Pasteur pipettes with decreasing diameter until no observable tissue pieces remained. The resulting cell suspensions were filtered using 70-μm cell strainers, which were subsequently washed with 10 ml D-PBS supplemented with 0.5% BSA. The cells were centrifuged at 300g for 5 min at +4 °C and pellets were suspended in 10 ml 30% isotonic Percoll (GE Healthcare 17-0891-01) diluted in 1× HBSS and then centrifuged again at 700g for 15 min at +4 °C with minimum acceleration and braking. The bottom 5 ml of the gradient containing microglia was collected, resuspended and passed through a 70-μm cell strainer. The cell suspension was washed in ice-cold 1× HBSS for a total volume of 40 ml and pelleted at 300g for 10 min at +4 °C.

Flow cytometry and fluorescence activated cell sorting

Dissociated cells were resuspended in 300 μl of FACS buffer (2% FBS and 1 mM EDTA in D-PBS, sterile). Fluorescence-labeling procedures were then carried out on ice. Fc-receptors were blocked by addition of anti-CD16/32 antibody (1:100 dilution, BioLegend) for 15 min followed by the addition of labeled antibodies: anti-CD45-PE (1:25 dilution, BioLegend), anti-CD11b-FITC (1:25 dilution, BioLegend) and Hoechst 33258 (1:1,000 dilution) for 1 h. Finally, samples were washed twice with FACS buffer and immediately taken to the Salk Institute Flow Cytometry core facility (samples were kept on ice from this point onwards, or chilled at 4 °C while undergoing FACS purification).

FACS purification was carried out on a BD FACS Aria Fusion sorter with 1× PBS for sheath fluid. For high viability, concentrated cells suitable for downstream 10X Genomics analysis, an 85-μm nozzle was used with sheath pressure set to 45 PSI. Live cells were gated first (Hoechst dye negative), followed by exclusion of debris using forward and side-scatter pulse area parameters (FSC-A and SSC-A), exclusion of aggregates using forward and side-scatter pulse width parameters (FSC-W and SSC-W), before finally gating on CD45⁺ cells to be isolated. Cells were purified using a one-drop single-cell sort mode (for counting accuracy). These were directly deposited into a 1.5-ml Eppendorf tube without additional buffer to yield a sufficient concentration that permitted direct loading onto the 10X chip.

Single-cell sequencing

scRNA-seq was conducted on FACS-sorted brain immune cells following the user guide on the Next GEM single cell 3′ v3.1 protocol. Briefly, single-cell suspensions of approximately 14,000–16,000 cells from each sample were directly loaded onto microfluidic chip with barcoded beads to generate Gel Bead-in-Emulsions (GEMs) using 10X Genomics Chromium Single-Cell Controller. Reverse transcription of GEMs for first-strand cDNA synthesis and complementary DNA amplification were carried out according to 10X Next GEM single cell 3′ v3.1 protocol. Following indexed scRNA-seq library construction, the final library size distribution was determined using TapeStation (Agilent) and the concentration was measured by a Qubit fluorometer (Thermo Fisher Scientific). The libraries were pooled in equal molar ratio, quantified by qPCR and sequenced on Illumina NextSeq500 at 28 cycles for read 1, 8 cycles for i7 index and 91 cycles for read 2 at an average sequencing read depth of 33,000–41,000 reads per cell.

Analysis of scRNA-seq

Raw sequencing data were aligned to the reference built using Ensembl primary assembly and annotation (release-93) that included both protein coding genes and polymorphic pseudogenes by cell ranger pipeline (v.3.1.0). The resulting filtered gene expression matrix was further analyzed by Seurat (3.2.1.9002)⁵³. Cells that had <10% of reads mapped to mitochondria genes and 200–10,000 expressed genes were used in the analysis. Samples were integrated by Seurat standard workflow ‘FindIntegrationAnchors’ and ‘IntegrateData’ functions with default settings. After manually examining the principal component (PC) elbow plot, the top 25 PCs of the integrated data were used for clustering analysis and UMAP analysis with default settings.

Doublet scores were calculated by R package ‘scds’ (v.1.2.3)⁵⁴ using the raw count matrix as input. The top 5% cells ranked by hybrid scores and clusters enriched (>50%) with doublets were removed from the downstream analysis. Data were re-scaled and re-clustered after doublet removal. The expression patterns of a large group of well-known immune markers (Extended Data Fig. 3b) were carefully examined to annotate the clusters. A second round of analysis was performed on microglia to further reveal the relationship of cells within the population. The top 20 PCs were used for clustering and UMAP analysis. One cluster that coexpressed both microglia markers (P2ry12 and Fcrls) and T cell/natural killer cell markers (Cd3g and Nkg7) was regarded as a doublet and was filtered out. Clustering resolution was set at 0.2 because clusters highly correlated with UMAP topologies. DE analysis was performed by Seurat function ‘FindAllMarkers’ and ‘FindMarkers’ with default Wilcoxon rank-sum test and logFC > 0.25 on pooled biological replicates. Genes with Bonferroni-adjusted P value <0.05 were considered to be significant. When plotted, the adjusted P value was log₁₀ transformed. Each adjusted P value was added to an extremely small number (1 × 10⁻³¹⁰) to avoid infinite values before transformation. Then the value was signed by the up/downregulation of the gene.

Supplementary Table 7 containing Trem2⁺ expression from different stages of disease progression from Keren-Shaul et al.¹¹ was used as validation dataset. Common symbols between this study and our top ten DE genes ranked by logFC from each microglia cluster were used for heat map plotting. Average expression values from both datasets were offset by 1 and log₂ transformed and z-scaled.

Quantitative PCR with reverse transcription

RNA from snap-frozen cortex or hippocampal tissue was isolated with TRIzol (Thermo Fisher Scientific, 15596026) according to the manufacturer’s instructions. The RT Transcriptor First Strand cDNA Synthesis kit (Roche) with anchored oligo(dT) primers (Roche) was used for reverse transcription. Quantitative PCR was run in a 384-well plate format on a QuantStudio Q5. Oligonucleotide primer sequences were ordered from Integrated DNA Technologies and reconstituted in house. Primer sequences used were: Gas6 forward primer 5′–3′ AACTGGCTGAACGGGGAAG; and reverse primer 5′–3′ CTTCCCAGGTGGTTTCCGT. TNFα forward primer 5′–3′ GCCACCACGCTCTTCTGTCT and reverse primer 5′–3′ CAGCTGCTCCTCCACTTG GT. IFNα4 forward primer 5′–3′ CCCACAGCCCAGAGAGTGAC and reverse primer 5′–3′ GCCCTCTTGTTCCCGAGGT. Il-6 forward primer AGACAAAGCCAGAGTCCTTCAGA and reverse primer 5′–3′ GCCACTCCTTCTGTGACTCCA. Il-1β forward primer 5′–3′ CCTCTCCAGCCAAGCTTCC reverse primer 5′–3′ CTCATCAGGACAGCCCAGGT. IFNγ forward primer CAATCAGGCCATCAGCAACA and reverse primer 5′–3′ AACAGCTGGTGGACCAC CG. Relative expression of genes of interest was normalized to Gapdh (primer pairs: forward sequence 5′–3′ AGGTCGGTGTGAACGGATTTG; reverse sequence 5′–3′ GGGGTCGTTGATGGCAAC).

Extraction of soluble Aβ and quantification of Aβ₄₂ by ELISA

Briefly, whole cortices and hippocampi were swiftly dissected from freshly extracted brains on an ice-cold platform and snap-frozen in liquid nitrogen. Tissues were lysed on ice in RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, 1% Triton-X 100 and 0.5% deoxycholate) with protease and phosphatase inhibitor (Roche, Sigma) at 4:1 RIPA volume:brain wet weight. Homogenates were spun at 5,000g first to deplete debris and then ultracentrifuged at 100,000 r.p.m. at 4 °C for 1 h. Supernatants (containing soluble Aβ) from the cortical and hippocampal preparation were aliquoted and stored at −80 °C until further use. Extracts from APP/PS1 and APP/PS1Axl^−/−Mertk^−/− and their nondiseased littermates were prepared simultaneously and identically.

Standard sandwich ELISA measurements of soluble Aβ₄₂ were performed based on published protocols⁵⁵. Briefly, 96-well plates were pre-coated with anti-Aβ₄₂ JRF AB042/26 capture antibody at a concentration of 1.5 μg ml⁻¹ at 50 μl per well overnight at 4 °C. The following day, plates were blocked at 150 μl per well with 1% casein blocking buffer for 2 h at room temperature. Plates were washed five times with PBS + 0.05% Tween-20. JRF/AbN/25 detection antibody to the N terminus of Aβ had previously been biotinylated and dialyzed following manufacturer’s instructions. The 25 μl biotinylated detection antibody was diluted in blocking buffer and then mixed with either 25 μl of standards (synthetic human Aβ_1-42 peptide) or 25 μl extracellular medium before loading 50 μl per well and in duplicate. Following overnight incubation at 4 °C, plates were rinsed and incubated in streptavidin–HRP at 50 μl per well for 30 min and developed in 1:1 mixture of Color Reagent A (H₂O₂) and Color Reagent B (tetramethylbenzidine) (cat. no. DY999). Plates were immediately measured on a TECAN Infinite 200 PRO reader at 450 nm.

pSIVA stereotaxic injections for in vivo PtdSer labeling

Stereotactic injections were performed following previously published protocols⁵⁶. Briefly, thin-wall glass pipettes were pulled on a Sutter Flaming/Brown puller and cut using sterile techniques resulting in tip diameters of 10–15 μm. Adult APP/PS1 and WT control mice (15–16 mo) were anesthetized with isoflurane (4% for induction; 1–2% during surgery). Mice were head-fixed in a computer-assisted stereotactic system with digital coordinate readout. Three sites were serially injected with pSIVA on the right hemisphere of each animal. Neocortical coordinates were (anterior–posterior (AP) +1 mm, medial–lateral (ML) +1.5 mm, dorsal–ventral (DV) +1 mm), (AP +0 mm, ML +1.5 mm, DV +1 mm) and (AP −1 mm, ML +1.5 mm, DV +1 mm) along the motor and somatosensory cortex around layer III/IV. First, craniotomy sites were marked and an electrical micro-drill with a fluted bit (0.5-mm tip diameter) was used to thin a 0.5–1-mm diameter part of the bone over the target injection site. Care was taken to ensure uninjured uplifting of the bone segment. Next, undiluted pSIVA dye was loaded into the glass pipette with sufficient volume and was gently lowered to the desired depth (using the DV coordinates). A total of 500 μl pSIVA per injection site was injected slowly at 1 nl s⁻¹ over a period of 15–20 min and with a 5-min break halfway and at the end of each injection to avoid backflow before carefully retracting the injection pipette to the next target injection site (using AP and ML coordinates). At the end of the last injection, mice were sutured along the incision of the scalp and given subcutaneous Buprenex SR (0.5 mg kg⁻¹). Each mouse was allowed to recover before placement in their home cage for 1.5–2 h before undergoing routine perfusion and processing of both injected right hemisphere and control left hemisphere (see previous section).

Surgery and animal preparation for in vivo two-photon imaging

Surgical procedures closely followed established protocols^7,33,56. Briefly, mice were anesthetized with isoflurane (4–5% for induction; 1–1.5% for maintenance), head-fixed with blunt ear bars and kept at 36–37 °C on a custom surgical bed (Thorlabs). Eyes were protected with vet ophthalmic ointment (Puralube). Depilator cream (Nair) was used to remove hair on top of the mouse’s head. The scalp was thoroughly cleansed and disinfected with a two-stage scrub of betadine and 70% ethanol. A scalp portion was surgically removed to expose frontal, parietal and interparietal skull segments. Scalp edges were attached to the lateral sides of the skull using tissue compatible adhesive (3M Vetbond). A custom-machined metal plate was affixed to the skull with dental cement (Coltene Whaledent, cat. no. H00335). Ear bars were removed and the head was stabilized by clamping the skull-attached plate with a custom holder. An approximately 3-mm diameter craniotomy was made over somatosensory cortex (center coordinates: AP −1.5 mm, ML 1.5 mm). A 1.5% agarose solution and coverslip were applied to the exposed tissue and the coverslip was fixed to the skull with dental cement to control tissue motion. Imaging commenced immediately after optical window preparation (corresponding to 13–14 h after intraperitoneal MX04 injection). Depth of anesthesia was monitored throughout the experiment and adjusted as needed to maintain a breath rate of approximately 55–65 breaths min⁻¹. Animal temperature was maintained at 36–37 °C and saline supplemented as needed to compensate for fluid loss.

Confocal microscopy

One-photon laser scanning confocal images were acquired with a Zeiss LSM 710 confocal microscope using Plan-Apochromat ×20 0.8-NA air-matched or ×63 1.4-NA oil objectives (laser lines, 405 nm, 488 nm, 594 nm and 633 nm). Image size was 1,024 × 1,024 pixels. Stack thickness was typically 15 μm for mouse brain sections and 5 μm for postmortem paraffin-embedded sections. For synaptic quantification specifically, CA1 apical dendrites were imaged at 3 μm thickness and z-stack image were obtained (optical slice 0.29 μm, 11 slices per 3-μm stack). Exposure acquisition was set according to the WT or nonmutant samples in all experiments. Airyscan super-resolution images were acquired with a Zeiss LSM 880 Rear Port Laser Scanning Confocal and Airyscan FAST Microscope using ×63 oil objective. Images were obtained and processed via the Zen Black and Zen Blue editions. Images for cortical and hippocampal plaque quantification were acquired with Olympus VS-120 Virtual Slide Scanning Microscope using a ×10 objective.

Two-photon microscopy

Live animal imaging was performed as previously described^7,33,56. Briefly, a Sutter Movable Objective Microscope equipped with a pulsed femtosecond Ti:Sapphire laser (Chameleon Ultra II, Coherent) with two fluorescence detection channels was used for imaging (dichroic beamsplitter, FF520-Di02 (Semrock); blue emission filter, FF01-452/45 (Semrock); green emission filter, ET525/70M (Chroma); photomultiplier tubes, H7422-40 GaAsP (Hamamatsu)). Laser excitation wavelength was set to 830 nm. Average laser power was <10–15 mW at the tissue surface and adjusted with depth as needed to compensate for signal loss due to scattering and absorption. An Olympus ×20 1.0-NA water immersion objective was used for light delivery and collection. Z-stacks included up to 350 images, acquired at 1-μm axial step size, and used a two-frame average, 512 × 512 pixel resolution and ×2.0–10 zoom (corresponding to 350 μm × 350 μm–72 μm × 72 μm fields of view). Time-lapse recordings typically included 60–70 images per stack, acquired at 1.0–1.2 μm axial step size, used a two-frame average, 60-stack repeat (corresponding to approximately 94 min total recording duration), 512 × 512 pixel resolution and ×3.3–5 zoom (corresponding to 212 μm–142 μm fields of view). Up to 13 z-stacks and four to seven time-lapse recordings were acquired per animal in layers 1 and 2 of the somatosensory cortex.

Imaging data analysis

For fixed-brain thin sections, maximum-intensity projection images of 212 μm × 212 μm (1,024 × 1,024 pixel resolution) were analyzed in FIJI. Approximately 5–7 plaques per brain sections (3–5 sections per animal) were randomly chosen in the prefrontal cortex using the 6E10 channel. MFI for Mer, Axl and Gas6 expression analysis, was calculated as the quotient of integrated density divided by region of interest, for example, Iba1 area or 6E10 area, respectively, which were gated based on their intensity and applied for all images in each experiment. Plaque-associated and nonplaque-associated Iba1 areas were classified manually by investigator blind with respect to genotypes. For quantification of dystrophic neurite area using LAMP1 or RTN3, area above set fluorescent thresholds for either LAMP1 or RTN3-marked neuritic dystrophy and 6E10 immuno-positive area were calculated and summed using the FIJI ‘analyze particle’ tool and the ratio of summed area (LAMP1 or RTN3 area/6E10 area) was calculated on per plaque basis. For LAMP1 analysis for convenience of binary categorization, dense-core plaques were defined as plaques that contain a single or solid (usually bright) 6E10⁺ core area ≥100 μm² and diffuse plaques were defined as the complement of these dense-core plaques that are characterized as those devoid of compact 6E10⁺ area ≥100 μm². To assess excitatory synapse changes in the mouse hippocampus, colocalization of vGlut1 and PSD95 in the apical dendrite area of CA1 was analyzed using Imaris software (Bitplane). Three-dimensional (3D) z-stack images were background subtracted and positive puncta of vGlut1 and PSD95 were selected and built as ‘spots’ by uniformly thresholding size and intensity across all sections and genotypes analyzed. PSD95 spots were then transformed into a distance vector using the ‘distance transform’ function, which was followed by a calculation of distance between vGlut1 spots and PSD95 spots. Puncta were considered colocalized if the distance between vGlut1 and PSD95 was ≤0.7 μm. Number of colocalized puncta from each image was averaged for one section or for one animal and compared between experimental groups that contain three animals. Three or more images per section and ≥3–5 sections per animal were analyzed.

For analyses of two-photon image stacks, Imaris software (v.9.1.2; Bitplane) was used for 3D reconstruction of GFP microglia and MX04-labeled dense-core plaques and also for the analysis of (1) the distance of the centroid of GFP⁺ microglial cell bodies to the edge of a MX04⁺ mass; (2) microglial cell body volume and (3) intracellular Aβ, from 1-μm-step z-series stack images of both genotypes. Two types of ‘surfaces’, as digital representations that capture the volume of either a microglia cell body or a dense-core (not diffuse) plaque for the GFP channel and MX04 channel, respectively, were created under the same threshold for all stack images. For MX04 objects, a surface cutoff was set to be at the clear-cut border of the outlining of dense-core plaques, exclusive of surrounding diffuse material. Only dense-core plaques whose diameters fell between 10–40 μm were included in the following analysis. For GFP surfaces, filters for ‘surface grain size’, ‘diameter of the largest sphere which fits into object’, ‘seed points diameter’, ‘quality for seed points’ (for example to determine whether a structure belongs to the same or a neighboring cell) and volume cutoff were set to 0.3 μm, 7.5 μm, 3.75 μm, 65 a.u. and 65 μm³, respectively, for capturing the structure of GFP microglial cell bodies in imaging volume of 142 μm × 142 μm × 70 μm for Fig. 4c and Extended Data Fig. 4a and a normalized imaging volume of 350 μm × 350 μm × 300 μm in somatosensory cortex (512 × 512 pixel resolution). The same thresholds were applied to all images of both genotypes. Structures on edge and/or captured incompletely were manually deleted. Created surfaces were then used to determine distance of the centroid of individual GFP surfaces to the closest edge of MX04 objects with the built-in MATLAB function in Imaris.

For quantification of intracellular Aβ, a GFP ‘cell body only’ channel was built with GFP surfaces as mask borders using the ‘set outside voxel to 0’ mask. MX04 and the ‘cell body only’ GFP channel were then thresholded to capture the overlapping volume between MX04 and GFP channel using the ‘colocalization’ function in Imaris. The overlapping volume was then normalized to the total volume of GFP of the image as a representation of the percentage of GFP microglia cell body volume occupied by MX04 material. Primary process number, total length for PAM and process polarization ratio for NPAM around dense-core plaques whose diameters fell between 10–30 μm were analyzed with FIJI based on maximum-intensity projection stack images of 142 µm × 142 μm × 60 μm in volume. The following parameters were measured using volumetric analysis^57,58. For primary process number and total length, PAM were defined as those whose cell body centroids were 0–5 μm away from the edge of the nearest MX04 edge³⁷. On average 3–5 microglia per plaque that were unambiguously distinct, showed clear process visibility and were not between two plaques were selected. Primary process number and length from PAM were measured using the ‘free-hand’ line tool on FIJI. Z-stacks of these projected images were gone through to ensure processes continuity from the same cell across layers. Mainly distal primary processes were analyzed in this group of cells due to the fact that majority of their somata were next to or in direct contact with a plaque surface. For polarization index, NPAM were defined as those whose cell body centroids were 20–50 μm from a dense-core MX04 edge and whose processes were clearly visible. Cells equidistant between two plaques were not included in the analysis. Calculation of this index was performed⁵⁸ using the following criteria: processes oriented toward plaques were classified as those on the plaque side of a line drawn through the microglial cell body perpendicular to a line drawn to the nearest plaque. The polarization index for one microglia cell is the ratio of the summed length of all ‘toward’ primary processes divided by the summed length of all processes.

For structural dynamics of individual microglia processes, process velocity was assessed as previously described^33,59, by manually tracing process tips over time using the Manual Tracking plug-in on FIJI on maximum-intensity projected time-lapse videos, typically over 90-min time spans that cover a field of view of 212 μm × 212 μm × 70 μm in the somatosensory cortices of age-matched AD and healthy littermates of WT and Axl^−/−Mertk^−/− mice. As mentioned in the previous section, measurements were made for AD microglia whose cell body centroids were within 5 μm from the edge of dense-core MeX04 edge as measured on projected images (PAM) and for those whose centroids were over 20 μm from any plaque in the field of view (NPAM). Resting microglia were microglia from non-AD age-matched littermates. At least 2–3 and 4–7 processes were quantified from PAM and NPAM, respectively, to obtain an average process velocity per microglia.

For CAA quantification, CAA area percentage was defined as ‘6E10⁺ vessel area’ by calculating the ratio of 6E10⁺ area within CD31⁺ area over total CD31 area in either somatosensory or prefrontal cortex of mice of each genotype. Measurements were averaged from images of respective brain areas from 15 sections (spanning >1.8 mm³ of brain volume) per mouse.

Immunoblot analysis

Tissues were snap-frozen in liquid nitrogen. Frozen tissues were lysed using RIPA buffer and phosphatase protease inhibitor (Roche, Sigma) for 30 min on ice. Supernatants were stored at −80 °C after samples were spun down at 12,000 r.p.m. for 5 min. Immunoblotting was as previously published⁵. Briefly, equal amounts of protein (10 μg) in 3× Laemmli buffer and 0.1 M dithiothreitol were separated by electrophoresis through 4–12% Bis-Tris polyacrylamide gels (Novex, Life Technologies) and were transferred to polyvinylidene difluoride membranes. Membranes were blocked with 1% casein block in PBS (BioRad 1610782), subsequently incubated with primary and secondary antibodies and washed with TBST (50 mM Tris-HCl, pH 7.5, 0.15 M NaCl and 0.25% Tween-20) between incubations. Blots were developed using an Odyssey Gel Imaging System (Li-Cor) and quantified using ImageStudio with GAPDH as loading control.

Contextual fear-conditioning behavior test

Fear acquisition and contextual memory tests were adapted from an established fear-conditioning paradigm⁶⁰. Briefly, mice were habituated for 5 min per animal for 5 consecutive days before the day of fear acquisition. One day before fear acquisition, mice were allowed to explore the conditioning context for 10 min. All testing was performed between 09:30 and 14:30 h and all mice were acclimated to the testing room for at least 1 h before the beginning of each behavioral assay. For fear acquisition (day 1), mice were evaluated for their abilities to learn the association between a context (conditional stimulus) and an aversive mild electrical stimulus (unconditional stimulus). Freezing behavior during and after three pairs of co-terminated stimuli of an auditory cue (30 s, 90 db, 2.5 kHz) and an electric shock (0.5 mA, 2 s) was calculated as ‘% freezing’ (the time the mouse was immobile during the time period in the presence of the paired context even when the aversive stimulus was absent). Freezing behavior was monitored and analyzed by an automated video-freeze software (Med Associates). Animals were returned to their home cage after the fear-conditioning session. On day 2, 24 h after fear acquisition, freezing behavior was measured in the original conditioning context for 3 min to test the memory retention of the learned association between conditional stimulus and unconditional stimulus. To avoid estrogen-related and other cognition-irrelevant variations, fear conditioning was performed only on cohorts of group-housed male mice of matched age.

Statistical analyses and reproducibility

Numerical data analysis and statistics were performed with Microsoft Excel (v.15.36) and GraphPad Prism (v.8.0) software, except for Fig. 3 and Extended Data Fig. 3. All data in all figure panels of the paper are represented as mean ±1 × s.d. In most cases, n is used to denote the number of mice per experimental group and N denotes the number of brain sections/images/samples analyzed per mouse. Group sample sizes were chosen based on previous studies and/or power analysis. Experiments were performed with at least two independent replicates whenever applicable. No technical replicate samples across independent experiments were pooled in the datasets. Datasets displayed normal distribution and equal s.d. unless indicated by unequal variance test. Statistical tests were always performed two-sided. Bonferroni-adjusted P values were calculated for genes in Fig. 3 and Extended Data Fig. 3. P values were calculated using Mann–Whitney U-test in Figs. 4d–f, 5a,b,d,e and Extended Data Figs. 1d, 4b–d, 6a–e,g, 7a,c,d and 8c,d. A Student’s t-test was used in Fig. 6c,d and Extended Data Fig. 7b,e,f. For multiple group comparison, P values were calculated in Figs. 2b and 4g and Extended Data Fig. 8a using a Kruskal–Wallis test followed by Dunn’s multiple comparison test. P values were calculated in Fig. 4c and Extended Data Fig. 4a,e using two-way analysis of variance with Sidak’s multiple comparison test. NS (not significant) indicates P > 0.05; significant P values are noted as exact values on graphs.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

We declare that the main data supporting the findings of this study are available within the article and its Extended Data information files. Sequencing data were deposited in the Gene Expression Omnibus with the accession number GSE160523. Additional supporting raw data are available from the authors upon request. Source data are provided with this paper.

References

Lemke, G. Biology of the TAM receptors. Cold Spring Harb. Perspect. Biol. https://doi.org/10.1101/cshperspect.a009076 (2013).
Lu, Q. & Lemke, G. Homeostatic regulation of the immune system by receptor tyrosine kinases of the Tyro 3 family. Science 293, 306–311 (2001).
Article CAS PubMed Google Scholar
Lemke, G. How macrophages deal with death. Nat. Rev. Immunol. 19, 539–549 (2019).
Article CAS PubMed PubMed Central Google Scholar
Scott, R. S. et al. Phagocytosis and clearance of apoptotic cells is mediated by Mer. Nature 411, 207–211 (2001).
Article CAS PubMed Google Scholar
Zagórska, A., Través, P. G., Lew, E. D., Dransfield, I. & Lemke, G. Diversification of TAM receptor tyrosine kinase function. Nat. Immunol. 15, 920–928 (2014).
Article PubMed PubMed Central CAS Google Scholar
Rothlin, C. V., Ghosh, S., Zuniga, E. I., Oldstone, M. B. & Lemke, G. TAM receptors are pleiotropic inhibitors of the innate immune response. Cell 131, 1124–1136 (2007).
Article CAS PubMed Google Scholar
Fourgeaud, L. et al. TAM receptors regulate multiple features of microglial physiology. Nature 532, 240–244 (2016).
Article CAS PubMed PubMed Central Google Scholar
Long, J. M. & Holtzman, D. M. Alzheimer disease: an update on pathobiology and treatment strategies. Cell 179, 312–339 (2019).
Article CAS PubMed PubMed Central Google Scholar
The ImmGen Consortium. Open-source ImmGen: mononuclear phagocytes. Nat. Immunol. 17, 741 (2016).
Lai, C. & Lemke, G. An extended family of protein-tyrosine kinase genes differentially expressed in the vertebrate nervous system. Neuron 6, 691–704 (1991).
Article CAS PubMed Google Scholar
Keren-Shaul, H. et al. A unique microglia type associated with restricting development of Alzheimer’s disease. Cell 169, 1276–1290 (2017).
Article CAS PubMed Google Scholar
Yin, Z. et al. Immune hyperreactivity of Aβ plaque-associated microglia in Alzheimer’s disease. Neurobiol. Aging 55, 115–122 (2017).
Article CAS PubMed Google Scholar
Lew, E. D. et al. Differential TAM receptor–ligand–phospholipid interactions delimit differential TAM bioactivities. Elife 3, e03385 (2014).
Article PubMed Central Google Scholar
Mattsson, N. et al. CSF protein biomarkers predicting longitudinal reduction of CSF β-amyloid42 in cognitively healthy elders. Transl. Psychiatry 3, e293 (2013).
Article CAS PubMed PubMed Central Google Scholar
Sainaghi, P. P. et al. Growth arrest specific 6 concentration is increased in the cerebrospinal fluid of patients with Alzheimer’s disease. J. Alzheimers Dis. 55, 59–65 (2017).
Article PubMed Google Scholar
Jankowsky, J. L. et al. Mutant presenilins specifically elevate the levels of the 42 residue β-amyloid peptide in vivo: evidence for augmentation of a 42-specific γ secretase. Hum. Mol. Genet. 13, 159–170 (2004).
Article CAS PubMed Google Scholar
Jankowsky, J. L. et al. Co-expression of multiple transgenes in mouse CNS: a comparison of strategies. Biomol. Eng. 17, 157–165 (2001).
Article CAS PubMed Google Scholar
Rockenstein, E., Mallory, M., Mante, M., Sisk, A. & Masliaha, E. Early formation of mature amyloid-β protein deposits in a mutant APP transgenic model depends on levels of Aβ_1–42. J. Neurosci. Res. 66, 573–582 (2001).
Article CAS PubMed Google Scholar
Labzin, L. I., Heneka, M. T. & Latz, E. Innate immunity and neurodegeneration. Annu Rev. Med. 69, 437–449 (2018).
Article CAS PubMed Google Scholar
Roy, E. R. et al. Type I interferon response drives neuroinflammation and synapse loss in Alzheimer disease. J. Clin. Invest. 130, 1912–1930 (2020).
Article CAS PubMed PubMed Central Google Scholar
Krasemann, S. et al. The TREM2-APOE pathway drives the transcriptional phenotype of dysfunctional microglia in neurodegenerative diseases. Immunity 47, 566–581 (2017).
Article CAS PubMed PubMed Central Google Scholar
Ulland, T. K. & Colonna, M. TREM2 — a key player in microglial biology and Alzheimer disease. Nat. Rev. Neurol. 14, 667–675 (2018).
Article CAS PubMed Google Scholar
Bennett, M. L. et al. New tools for studying microglia in the mouse and human CNS. Proc. Natl Acad. Sci. USA 113, E1738–E1746 (2016).
Article CAS PubMed PubMed Central Google Scholar
Kim, D. K. et al. Deep proteome profiling of the hippocampus in the 5XFAD mouse model reveals biological process alterations and a novel biomarker of Alzheimer’s disease. Exp. Mol. Med. 51, 1–17 (2019).
PubMed PubMed Central Google Scholar
Kim, Y. E., Chen, J., Langen, R. & Chan, J. R. Monitoring apoptosis and neuronal degeneration by real-time detection of phosphatidylserine externalization using a polarity-sensitive indicator of viability and apoptosis. Nat. Protoc. 5, 1396–1405 (2010).
Article CAS PubMed PubMed Central Google Scholar
Friedman, B. A. et al. Diverse brain myeloid expression profiles reveal distinct microglial activation states and aspects of Alzheimer’s disease not evident in mouse models. Cell Rep. 22, 832–847 (2018).
Article CAS PubMed Google Scholar
Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial reconstruction of single-cell gene expression data. Nat. Biotechnol. 33, 495–502 (2015).
Article CAS PubMed PubMed Central Google Scholar
Gonzalez, H. & Pacheco, R. T-cell-mediated regulation of neuroinflammation involved in neurodegenerative diseases. J. Neuroinflammation 11, 201 (2014).
Article PubMed PubMed Central CAS Google Scholar
Spangenberg, E. et al. Sustained microglial depletion with CSF1R inhibitor impairs parenchymal plaque development in an Alzheimer’s disease model. Nat. Commun. 10, 3758 (2019).
Article PubMed PubMed Central CAS Google Scholar
Lu, Q. et al. Tyro-3 family receptors are essential regulators of mammalian spermatogenesis. Nature 398, 723–728 (1999).
Article CAS PubMed Google Scholar
Jung, S. et al. Analysis of fractaÿlkine receptor CX₃CR1 function by targeted deletion and green fluorescent protein reporter gene insertion. Mol. Cell. Biol. 20, 4106–4114 (2000).
Article CAS PubMed PubMed Central Google Scholar
Klunk, W. E. et al. Imaging Aβ plaques in living transgenic mice with multiphoton microscopy and methoxy-X04, a systemically administered Congo red derivative. J. Neuropathol. Exp. Neurol. 61, 797–805 (2002).
Article CAS PubMed Google Scholar
Nimmerjahn, A., Kirchhoff, F. & Helmchen, F. Resting microglial cells are highly dynamic surveillants of brain parenchyma in vivo. Science 308, 1314–1318 (2005).
Article CAS PubMed Google Scholar
Nimmerjahn, A. Two-photon imaging of microglia in the mouse cortex in vivo. Cold Spring Harb. Protoc. https://doi.org/10.1101/pdb.prot069294 (2012).
Wang, Y. et al. TREM2 lipid sensing sustains the microglial response in an Alzheimer’s disease model. Cell 160, 1061–1071 (2015).
Article CAS PubMed PubMed Central Google Scholar
Wang, Y. et al. TREM2-mediated early microglial response limits diffusion and toxicity of amyloid plaques. J. Exp. Med. 213, 667–675 (2016).
Article CAS PubMed PubMed Central Google Scholar
Condello, C., Yuan, P., Schain, A. & Grutzendler, J. Microglia constitute a barrier that prevents neurotoxic protofibrillar Aβ42 hotspots around plaques. Nat. Commun. 6, 6176 (2015).
Article CAS PubMed Google Scholar
Hassiotis, S. et al. Lysosomal LAMP1 immunoreactivity exists in both diffuse and neuritic amyloid plaques in the human hippocampus. Eur. J. Neurosci. 47, 1043–1053 (2018).
Article PubMed Google Scholar
Hu, X. et al. Transgenic mice overexpressing reticulon 3 develop neuritic abnormalities. EMBO J. 26, 2755–2767 (2007).
Article CAS PubMed PubMed Central Google Scholar
Greenberg, S. M. et al. Cerebral amyloid angiopathy and Alzheimer disease—one peptide, two pathways. Nat. Rev. Neurol. 16, 30–42 (2020).
Article CAS PubMed Google Scholar
Blanco-Suarez, E., Liu, T. F., Kopelevich, A. & Allen, N. J. Astrocyte-secreted chordin-like 1 drives synapse maturation and limits plasticity by increasing synaptic GluA2 AMPA receptors. Neuron 100, 1116–1132.e13 (2018).
Article CAS PubMed PubMed Central Google Scholar
Jay, T. R. et al. Disease progression-dependent effects of TREM2 deficiency in a mouse model of Alzheimer’s disease. J. Neurosci. 37, 637–647 (2017).
Article CAS PubMed PubMed Central Google Scholar
Jay, T. R. et al. TREM2 deficiency eliminates TREM2⁺ inflammatory macrophages and ameliorates pathology in Alzheimer’s disease mouse models. J. Exp. Med. 212, 287–295 (2015).
Article CAS PubMed PubMed Central Google Scholar
Yuan, P. et al. TREM2 haplodeficiency in mice and humans impairs the microglia barrier function leading to decreased amyloid compaction and severe axonal dystrophy. Neuron 90, 724–739 (2016).
Article CAS PubMed PubMed Central Google Scholar
Baik, S. H., Kang, S., Son, S. M. & Mook-Jung, I. Microglia contributes to plaque growth by cell death due to uptake of amyloid β in the brain of Alzheimer’s disease mouse model. Glia 64, 2274–2290 (2016).
Article PubMed Google Scholar
Sosna, J. et al. Early long-term administration of the CSF1R inhibitor PLX3397 ablates microglia and reduces accumulation of intraneuronal amyloid, neuritic plaque deposition and pre-fibrillar oligomers in 5XFAD mouse model of Alzheimer’s disease. Mol. Neurodegener. 13, 11 (2018).
Article PubMed PubMed Central CAS Google Scholar
Fu, H. et al. Complement component C3 and complement receptor type 3 contribute to the phagocytosis and clearance of fibrillar Aβ by microglia. Glia 60, 993–1003 (2012).
Article PubMed PubMed Central Google Scholar
Hu, X. et al. Amyloid seeds formed by cellular uptake, concentration, and aggregation of the amyloid-β peptide. Proc. Natl Acad. Sci. USA 106, 20324–20329 (2009).
Article CAS PubMed PubMed Central Google Scholar
Cambier, C. J., Falkow, S. & Ramakrishnan, L. Host evasion and exploitation schemes of Mycobacterium tuberculosis. Cell 159, 1497–1509 (2014).
Article CAS PubMed Google Scholar
Panza, F., Lozupone, M., Logroscino, G. & Imbimbo, B. P. A critical appraisal of amyloid-β-targeting therapies for Alzheimer disease. Nat. Rev. Neurol. 15, 73–88 (2019).
Article PubMed Google Scholar
Hutter-Paier, B. et al. The ACAT inhibitor CP-113,818 markedly reduces amyloid pathology in a mouse model of Alzheimer’s disease. Neuron 44, 227–238 (2004).
Article CAS PubMed Google Scholar
Imbimbo, B. P. et al. CHF5074, a novel γ-secretase modulator, attenuates brain β-amyloid pathology and learning deficit in a mouse model of Alzheimer’s disease. Br. J. Pharmacol. 156, 982–993 (2009).
Article CAS PubMed PubMed Central Google Scholar
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902 (2019).
Article CAS PubMed PubMed Central Google Scholar
Bais, A. S. & Kostka, D. scds: computational annotation of doublets in single-cell RNA sequencing data. Bioinformatics 36, 1150–1158 (2020).
Article CAS PubMed Google Scholar
Schmidt, S. D., Mazzella, M. J., Nixon, R. A. & Mathews, P. M. Aβ measurement by enzyme-linked immunosorbent assay. Methods Mol. Biol. 849, 507–527 (2012).
Article CAS PubMed Google Scholar
Tufail, Y. et al. Phosphatidylserine exposure controls viral innate immune responses by microglia. Neuron 93, 574–586 (2017).
Article CAS PubMed PubMed Central Google Scholar
Koenigsknecht-Talboo, J. et al. Rapid microglial response around amyloid pathology after systemic anti-Aβ antibody administration in PDAPP mice. J. Neurosci. 28, 14156–14164 (2008).
Article CAS PubMed PubMed Central Google Scholar
Bolmont, T. et al. Dynamics of the microglial/amyloid interaction indicate a role in plaque maintenance. J. Neurosci. 28, 4283–4292 (2008).
Article CAS PubMed PubMed Central Google Scholar
Bernier, L. P. et al. Nanoscale surveillance of the brain by microglia via cAMP-regulated filopodia. Cell Rep. 27, 2895–2908 (2019).
Article CAS PubMed Google Scholar
Kitamura, T. et al. Island cells control temporal association memory. Science 343, 896–901 (2014).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank J. Hash for technical assistance, M. Mercken (Janssen Pharmaceuticals) for Aβ₄₂ antibodies, R. Rissman and the ADRC at UCSD for AD brain sections, M. Shokhirev for advice on bioinformatic analyses, S. Parylak and F. Gage for advice on fear-conditioning assays and members of the Lemke laboratory and the Nomis Center for discussions. The research was supported by grants from the US National Institutes of Health (RF1 AG060748 and R01 AI101400 to G.L.; DP2 NS083038, R01 NS108034 and U01 NS103522 to A.N.; P30 AG062429 to the UCSD ADRC; and P30 CA014195 and S10-OD023689 to the Salk Institute), the Cure Alzheimer’s Fund and the Coins for Alzheimer’s Research Trust (to G.L.) and the Leona M. and Harry B. Helmsley Charitable Trust (to the Salk Institute); by Goeddel’s Chancellor’s, Marguerite Vogt and H. A. and Mary K. Chapman Charitable Trust graduate fellowships (to Y.H.) and by Anderson, Nomis and Sweden–America Foundation postdoctoral fellowships (to K.E.H.).

Author information

Authors and Affiliations

Molecular Neurobiology Laboratory, The Salk Institute for Biological Studies, La Jolla, CA, USA
Youtong Huang, Kaisa E. Happonen, Patrick G. Burrola & Greg Lemke
Division of Biological Sciences, University of California, San Diego, La Jolla, CA, USA
Youtong Huang
Flow Cytometry Core, The Salk Institute for Biological Studies, La Jolla, CA, USA
Carolyn O’Connor
Chapman Foundations Genomic Sequencing Core, The Salk Institute for Biological Studies, La Jolla, CA, USA
Nasun Hah
Razavi Newman Integrative Genomics and Bioinformatics Core, The Salk Institute for Biological Studies, La Jolla, CA, USA
Ling Huang
Waitt Advanced Biophotonics Center, The Salk Institute for Biological Studies, La Jolla, CA, USA
Axel Nimmerjahn

Authors

Youtong Huang
View author publications
You can also search for this author in PubMed Google Scholar
Kaisa E. Happonen
View author publications
You can also search for this author in PubMed Google Scholar
Patrick G. Burrola
View author publications
You can also search for this author in PubMed Google Scholar
Carolyn O’Connor
View author publications
You can also search for this author in PubMed Google Scholar
Nasun Hah
View author publications
You can also search for this author in PubMed Google Scholar
Ling Huang
View author publications
You can also search for this author in PubMed Google Scholar
Axel Nimmerjahn
View author publications
You can also search for this author in PubMed Google Scholar
Greg Lemke
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

Y.H. was an essential contributor to every aspect of the paper. K.E.H. made extensive contributions to experimental design and implementation. P.G.B. obtained and prepared tissue sections for immunostaining. C.O’C. contributed to the design and performance of flow cytometry. N.H. carried out scRNA-seq. L.H. performed clustering and bioinformatics for scRNA-seq analyses. A.N. designed and implemented methods for two-photon imaging and data analyses. G.L. conceived the project and designed experiments. All authors contributed to the writing of the manuscript.

Corresponding author

Correspondence to Greg Lemke.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Immunology thanks Cristoforo Comi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. L. A. Dempsey was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Expression of Axl, Mer, TMEM119, Trem2, and GFAP in plaque-burdened AD brains.

a, Axl protein (green, lower panels) is undetectable in WT (left) and APP/PS1 (right) microglia (Iba1, red, upper panels) in the 4mo cortex, prior to the appearance of plaques in APP/PS1 mice. n = 3 per genotype. b, Axl protein (green) in 15 mo APP41 mice (n = 3) is up-regulated in cortical microglia (Iba1, red) contacting Aβ plaques (6E10, white), and is also often concentrated in plaque centers, consistent with strong prior activation of Axl and subsequent cleavage of its ectodomain. (As is seen for other RTKs, robust activation of Axl results in nearly complete cleavage of the ectodomain from the cell surface.) Hoechst 33258 stains nuclei. c, Expression of Mer protein (green) in 15 mo APP41 mice is seen in all cortical microglia (Iba1, red), but is further up-regulated in microglia that invest Aβ plaques (6E10, white). d, Quantification of Mer up-regulation in Iba1⁺ plaque-associated microglia (PAM) versus non-plaque-associated microglia (NPAM) in APP/PS1 mice at 9.5 months. e, Axl (green) and Trem2 (red) are up-regulated in the same Iba1⁺ (white) microglia cell in the 9.5 mo APP/PS1 cortex. f, Expression of the homeostatic microglial marker TMEM119 (green) is lost in cortical microglia that surround plaques (6E10, white) in 15 mo APP41 mice, except for an occasional 1-2 cells at the center of plaques. g, This same TMEM119 down-regulation is seen in 15 mo APP/PS1 mice. TMEM119^- microglia surrounding 6E10 + plaques are strongly Iba1⁺. h, The up-regulated Mer expression (green) seen in 15 mo APP41 mice is not in GFAP⁺ reactive astrocytes (red). (Activated S100b⁺ astrocytes are also negative for Mer expression by IHC⁷). Circles mark the position of Aβ plaques. Representative images obtained from immunostaining of N ≥ 3 sections from n = 3 mice of each genotype. Scale bars: 10 μm (a-c, e, h), 100 μm (f, g). Mann-Whitney test (d). For all supplementary figure panels, data are represented as mean ±1 STD.

Source data

Extended Data Fig. 2 Gas6 and PtdSer decoration of Aβ plaques.

a, Gas6 protein (green) decorates 6E10⁺ plaques (white) on sections of postmortem cortex from human patients with advanced (BRAAK stage 6) AD (left panels), but is not evident in the plaque-free cortex of cognitively normal age-matched controls (right panels). Representative images obtained from immunostaining of >3 sections from 3 individuals of each condition. b, Visualization of externalized PtdSer in 15 mo APP/PS1 cortex following stereotaxic injection of pSIVA (green, left panel). pSIVA binds to externalized PtdSer in the needle track of the injection, where cells are damaged and undergoing apoptosis, and also to the PtdSer associated with all 6E10⁺ amyloid plaques adjacent to the injection site (white, right panel). (c) A similar pSIVA injection in 15 mo WT mice labels only the needle track of the injection, since there are no plaques in these mice. (d) Airyscan super-resolution image of the juxtaposition of plaque-associated microglia (Iba1, red), Gas6 (green), and Aβ plaque (6E10, white) in 12 mo APP/PS1 mice (n = 5). Scale bars: 100 μm (a), 50 μm (b), 200 μm (c), 10 μm (d). n = 3 and 2 for APP/PS1 and WT control, respectively (b, c) from two independent experiments.

Extended Data Fig. 3 Transcriptomics of APP/PS1 and APP/PS1Axl^−/−Mertk^−/− microglia as quantified by single cell RNA-seq.

a, Sorting scheme for isolation of CD45⁺ single cells. FSC and SSC, forward and side scatter, respectively. A, area; W, width. b, Uniform manifold approximation and projection (UMAP) clustering of CD45⁺ cells sorted from 18 mo APP/PS1 (A/PS) and APP/PS1Axl^−/−Mertk^−/− (A/PS A/M^−/−) cortices (combined) and annotated using the 18 marker genes in c. c, Violin plots of population log-transformed normalized expression of the indicated genes in the indicated cell types. Cell type key applies to b and c. d, Log-transformed normalized expression of Axl (left) and Mertk (right) mRNA in APP/PS1 cells within the microglial clusters defined in Fig. 3a. e, Heat map of the scaled expression levels of the indicated genes in individual cells within transcriptomic state (cluster) 0 and state 5 microglia in the APP/PS1 cortex at 18 mo. f, Comparative composite heat maps for the indicated genes across stages and transcriptomic states in 5xFAD and APP/PS1 mice, as quantified in this study (top five rows) and in Keren-Shaul et al.¹¹ (bottom three rows), respectively. Values shown as z-scaled log-transformed normalized average of each group. g, Violin plots of the log-transformed normalized expression distribution of the indicated genes at transcriptomic state 5 in A/PS (black) versus A/PS A/M^−/− (red) microglia, as determined by single cell RNA-seq (scRNA-seq). Dotted lines indicate mean. h, Mean expression level of the indicated cytokine and chemokine genes in state 5 versus state 0 microglia in A/PS (black) versus A/PS A/M^−/− (red) microglia at 18 mo, as determined by scRNA-seq. i, Relative expression level of the indicated inflammatory regulator mRNAs in RNA isolated from total cortex of mice of the indicated genotypes at 12 mo, as determined by qRT-PCR. n = 3-6. Kruskal-Wallis test with Dunn’s multiple comparison test. Data are represented as mean ±1 STD.

Source data

Extended Data Fig. 4 APP/PS1Axl^−/−Mertk^−/− microglia are unresponsive to Aβ plaques.

a, Distribution of distance of microglial cell body centroids, in 2 μm bins, from the edge of MX04-labeled Aβ plaques with diameters of 10-15 μm in APP/PS1 (gray) and APP/PS1Axl^−/−Mertk^−/− (red) cortex. Values obtained for 56 and 23 plaques from 3 and 4 mice for APP/PS1 and APP/PS1Axl^−/−Mertk^−/−, respectively. b, Number of imaged GFP⁺ primary processes per PAM (microglia < 5 μm from plaques) in APP/PS1 (gray, A/PS) and APP/PS1Axl^−/−Mertk^−/− (red, A/PS A/M^−/−) cortex. c, Summed length of primary microglial processes per PAM in APP/PS1 (gray) and APP/PS1Axl^−/−Mertk^−/− (red) cortex. d, Process polarization ratio to nearest plaque per NPAM (microglia > 20 μm from plaques; see Materials and Methods) in APP/PS1 (gray) and APP/PS1Axl^−/−Mertk^−/− (red) cortex. e, Quantification of microglial cell density in the cortex of 16 mo APP/PS1 (gray) and APP/PS1Axl^−/−Mertk^−/− (red) mice for microglia 0-10 μm, 10-20 μm, and >20 μm from the edge of the nearest plaque. Data points are from 45-129 cells (PAM) investing 10-29 plaques (b, c), and 21-49 cells peripheral to 7-24 plaques (d) from n = 3 mice per genotype (b-d), Points in e represent 3-5 imaging volumes from 3 APP/PS and 4 APP/PS1Axl^−/−Mertk^−/− mice. Two-way ANOVA with Sidak’s multiple comparison test (a, e) and Mann-Whitney’s test (b-d), Data are represented as mean ±1 STD.

Source data

Extended Data Fig. 5 Expansive areas of plaque-associated dystrophic LAMP1⁺ membrane and poorly compacted plaques in the APP/PS1Axl^−/−Mertk^−/− brain.

A montage of 24 paired sections in both APP/PS1 (top six rows) and APP/PS1Axl^−/−Mertk^−/− (bottom six rows) cortex, each stained with antibodies to both LAMP1 (green) and 6E10 (white). Each 6E10 image is paired with (from the same section as) the LAMP image immediately below. This montage, which is a subset of the images used to generate the data in Fig. 5d, is composed of images taken from three different mice of each genotype at 12 months. Note that: (a) 6E10⁺ Aβ plaques are in general more compact and brightly stained in APP/PS1 mice and more diffuse and weakly stained in APP/PS1Axl^−/−Mertk^−/− mice; and (b) the area occupied by LAMP1⁺ membrane is in general much larger in APP/PS1Axl^−/−Mertk^−/− mice. Scale bars: 10 μm.

Extended Data Fig. 6 Accumulation of LAMP1⁺ dystrophic membrane and apoptotic cell debris in the APP/PS1Axl^−/−Mertk^−/− brain.

a, Quantification of LAMP1/6E10 area ratio as in Fig. 5d, but only for dense-core plaques (plaques with solid 6E10⁺ cores with areas > 100 μm²). b, Quantification of LAMP1/6E10 area ratio as in Fig. 5d, but only for diffuse plaques (plaques without solid 6E10⁺ cores with areas > 100 μm²). c, Quantification of RTN3/6E10 area ratio as in Fig. 5e, but only for dense-core plaques. d, Quantification of RTN3/6E10 area ratio as in Fig. 5e, but only for diffuse plaques. e, Quantification of the density of diffuse plaques (defined as above) expressed as a fraction of total plaques in the cortex of mice of the indicated genotypes at 12 mo. Data represent diffuse plaques quantified from N = 4-5 sections from n = 6 mice per group. (f) Representative example of cerebral amyloid angiopathy (CAA) in the cortex of a 15 mo APP/PS1Axl^−/−Mertk^−/− mouse. 6E10⁺ Aβ material (white) is evident within laminin+ blood vessels (green). Asterisk marks an Aβ plaque in the parenchyma. (g) Quantification (see Methods) of CAA in the somatosensory cortex of 15 mo APP/PS1Axl^−/−Mertk^−/− mice (A/PS A/M^−/−) relative to APP/PS1 mice (A/PS). n = 4/group and measurements were averaged from N > 15 sections (spanning > 1.8mm³ of brain volume) per mouse. (h) cCasp3⁺ apoptotic debris (cyan, lower panels) accumulates around 6E10⁺ Aβ plaques (upper panels) in the APP/PS1Axl^−/−Mertk^−/− (right panels) but not the APP/PS1 (left panels) hippocampus at 12 mo. Images are representative of n = 3 mice per genotype from three independent experiments. Scale bars: 100 μm. Data are 18-47 (a), 67-78 (b), 30-43 (c) and 26-52 (d) plaques investigated from N ≥ 3 sections per mouse from n = 3 mice of each genotype from at least 3 independent replicates. Mann-Whitney test (a-e, g). Data are represented as mean ±1 STD.

Source data

Extended Data Fig. 7 TAM (Mer) signaling promotes dense-core Aβ plaque accumulation with functional consequences.

a, Thio S plaque density in APP/PS1 (gray) versus APP/PS1Axl^−/−Mertk^−/− (red) cortex for plaques of the indicated size at 12 months. b, Thio S⁺ plaque density (all plaque sizes) in APP/PS1 (gray) versus APP/PS1Axl^−/−Mertk^−/− (red) hippocampus at 12 months. c, Soluble Aβ42 levels quantified in APP/PS1 (gray) versus APP/PS1Axl^−/−Mertk^−/− (red) cortex and hippocampus at 4 and 12 mo, as indicated. n = 5-6 per genotype. d, Quantitative LI-COR western blot measurement of APP protein levels in the 12 mo cortex of 3 cohorts of mice (4 genotypes each cohort) of the indicated genotypes demonstrates no change in APP expression in APP/PS1 mice upon mutation of Axl and Mertk. Blots left and quantification right. e, ThioS⁺ plaque density (all plaque sizes) in APP/PS1 (gray) versus APP/PS1Mertk^−/− (pink) cortex and hippocampus at 12 months. f, Thio S⁺ plaque density (all plaque sizes) in APP/PS1 (gray) versus APP/PS1Axl^−/− (white) cortex and hippocampus at 12 months. Data points represent plaque density in n = 6-8 mice of the indicated genotypes averaged from N ≥ 5 cortical sections for each brain. Mann-Whitney test (a, c, d) and Student’s t-test (b, e, f). Data are represented as mean ±1 STD. g, TAM-mediated microglial recognition, phagocytosis, and consolidation of Aβ plaques. Microglial Axl and Mer are bridged to the PtdSer-rich dystrophic membranes of plaques via TAM ligands, whose amino-terminal and carboxy-terminal domains bind PtdSer and Axl/Mer, respectively^3,26. Gas6 is shown, but a role for the Mer ligand Pros1²⁶ is not excluded. Engagement of the PtdSer-TAM ligand-TAM receptor complex activates the TAM tyrosine kinases (TK), which drives phagocytosis of forming plaque material. Internalized phagocytic cargo is eventually transferred to lysosomes, whose acidic interiors promote the aggregation of large, insoluble Aβ fibrils. Exocytosis or microglial death then delivers this aggregated material to growing dense-core plaques.

Source data

Extended Data Fig. 8 Functional consequences of TAM deletion in APP/PS1 mice.

a, Quantification of the 3D colocalization of the excitatory pre- and post-synaptic markers vGlut1 and PSD95 in the 15-mo hippocampus (CA1) (see Methods), as an index of synaptic connectivity. The previously documented decrease in co-localization of these markers in APP/PS1 mice is not altered by the combined mutation of Axl and Mertk. Stack size is 85 x 85 x3 μm³ per image, averaged 3 images per CA1 section across 3-5 sections per mouse. Data points represent synaptic density index in n = 3 mice of the indicated genotypes. b, Acquisition of association between a 30 s auditory tone and a subsequent co-terminal 2 s 0.5 mA foot shock, expressed as percent time immobile (% freezing) during the indicated intervals, over three successive trial intervals in 15 mo mice of the indicated genotypes (see Methods). c, Data in b plotted for the indicated genotypes with the indicated statistical significance per interval. d, Contextual fear memory as assayed by percent of a 3 minute interval in which mice of the indicated genotypes were immobile (% freezing) when returned to the same testing cage 24 h after the fear acquisition trials of b (see Methods). A cohort of group-housed male mice (n = 12-20/group) were used in the behavioral assay. Data points in b represent the mean % freezing of each group in the interval duration immediately prior to the point. Each data point in d is the % freezing of one animal in the duration of the testing period. Kruskal Wallis test followed by Dunn’s multiple comparison test (a) and Mann-Whitney test (c, d). Data are represented as mean ±1 STD.

Source data

Supplementary information

Reporting Summary

Supplementary Video 1

Microglia show reduced interaction with amyloid plaques in somatosensory cortical layer 1 of 16-mo APP/PS1Axl^−/−Mertk^−/− as compared to APP/PS1 mice. Left, example two-photon time-lapse recording showing microglia (GFP signal, white) and amyloid plaques (MX04 signal, red) in somatosensory cortical layer 1 of an anesthetized APP/PS1Cx3cr1^GFP/+ mouse. Each image is a maximum-intensity projection of a 70-μm-thick z-stack (1.2 μm axial step size) recorded every ~93 s at depths between 100 μm and 30 μm below the pia. Right, example two-photon time-lapse recording showing microglia (GFP signal, white) and amyloid plaques (MX04 signal, red) in somatosensory cortical layer 1 of an age- and sex-matched anesthetized APP/PS1Axl^−/−Mertk^−/−Cx3cr1^GFP/+ mouse. Each image is a maximum-intensity projection of a 70-μm-thick z-stack (1.2 μm axial step size) recorded every ~93 s at depths between 94 μm and 24 μm below the pia. Elapsed time is indicated in the upper right corner. Amyloid plaques are closely encapsulated by highly polarized microglia in APP/PS1 but not APP/PS1Axl^−/−Mertk^−/− mice. Additionally, microglia near plaques typically show fewer and less-elaborate processes, higher process polarization and reduced process motility in APP/PS1 as compared to APP/PS1Axl^−/−Mertk^−/− mice. Supplementary Videos 1 and 2 were prepared using FIJI and custom-written scripts in Igor Pro (v.6.04; WaveMetrics). Scale bar, 30 μm.

Supplementary Video 2

Microglia show reduced interaction with amyloid plaques in somatosensory cortical layers 1/2 of 16-mo APP/PS1Axl^−/−Mertk^−/− as compared to APP/PS1 mice. Left, example two-photon time-lapse recording showing microglia (GFP signal, white) and amyloid plaques (MX04 signal, red) in somatosensory cortical layers 1/2 of an anesthetized APP/PS1Cx3cr1^GFP/+ mouse. Each image is a maximum-intensity projection of a 70-μm-thick z-stack (1.2 μm axial step size) recorded every ~93 s at depths between 148 μm and 78 μm below the pia. Right, example two-photon time-lapse recording showing microglia (GFP signal, white) and amyloid plaques (MX04 signal, red) in somatosensory cortical layers 1/2 of an age-and sex-matched anesthetized APP/PS1Axl^−/−Mertk^−/−Cx3cr1^GFP/+ mouse. Each image is a maximum-intensity projection of a 70-μm-thick z-stack (1.2 μm axial step size) recorded every ~93 s at depths between 135 μm and 65 μm below the pia. Elapsed time is indicated in the upper right corner. Amyloid plaques are closely encapsulated by highly polarized microglia in APP/PS1 but not APP/PS1Axl^−/−Mertk^−/− mice. Additionally, microglia near plaques typically show fewer and less-elaborate processes, higher process polarization and reduced process motility in APP/PS1 as compared to APP/PS1Axl^−/−Mertk^−/− mice. Scale bar, 30μm.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Huang, Y., Happonen, K.E., Burrola, P.G. et al. Microglia use TAM receptors to detect and engulf amyloid β plaques. Nat Immunol 22, 586–594 (2021). https://doi.org/10.1038/s41590-021-00913-5

Download citation

Received: 05 May 2020
Accepted: 05 March 2021
Published: 15 April 2021
Issue Date: May 2021
DOI: https://doi.org/10.1038/s41590-021-00913-5

This article is cited by

Modulation of hippocampal protein expression by a brain penetrant biologic TNF-α inhibitor in the 3xTg Alzheimer’s disease mice
- Nataraj Jagadeesan
- G. Chuli Roules
- Rachita K. Sumbria
Journal of Translational Medicine (2024)
Altered amyloid-β structure markedly reduces gliosis in the brain of mice harboring the Uppsala APP deletion
- María Pagnon de la Vega
- Stina Syvänen
- Dag Sehlin
Acta Neuropathologica Communications (2024)
Taming microglia: the promise of engineered microglia in treating neurological diseases
- Echo Yongqi Luo
- Rio Ryohichi Sugimura
Journal of Neuroinflammation (2024)
Exercise mimetics: a novel strategy to combat neuroinflammation and Alzheimer’s disease
- Renqing Zhao
Journal of Neuroinflammation (2024)
Microbiota–gut–brain axis and its therapeutic applications in neurodegenerative diseases
- Jian Sheng Loh
- Wen Qi Mak
- Kooi Yeong Khaw
Signal Transduction and Targeted Therapy (2024)

Subjects

Abstract

Similar content being viewed by others

Main

Results

Engagement of TAM receptors during disease

Engagement of TAM ligand and co-ligand during disease

TAM regulation of the microglial transcriptome in AD

Microglial TAM activity at plaques

TAM-dependent phagocytosis and shaping of plaques

Dense-core plaque burden in the absence of TAM signaling

Discussion

Methods

Mice

Reagents and antibodies

Immunohistochemistry and histochemistry

Human postmortem AD brain sections

Isolation and purification of brain immune single cells for sequencing

Flow cytometry and fluorescence activated cell sorting

Single-cell sequencing

Analysis of scRNA-seq

Quantitative PCR with reverse transcription

Extraction of soluble Aβ and quantification of Aβ42 by ELISA

pSIVA stereotaxic injections for in vivo PtdSer labeling

Surgery and animal preparation for in vivo two-photon imaging

Confocal microscopy

Two-photon microscopy

Imaging data analysis

Immunoblot analysis

Contextual fear-conditioning behavior test

Statistical analyses and reproducibility

Reporting Summary

Data availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

Extraction of soluble Aβ and quantification of Aβ₄₂ by ELISA