Microglial mTOR Activation Upregulates Trem2 and Enhances β-Amyloid Plaque Clearance in the 5XFAD Alzheimer's Disease Model

Abstract

The mechanistic target of rapamycin (mTOR) signaling pathway plays a major role in key cellular processes including metabolism and differentiation; however, the role of mTOR in microglia and its importance in Alzheimer's disease (AD) have remained largely uncharacterized. We report that selective loss of Tsc1, a negative regulator of mTOR, in microglia in mice of both sexes, caused mTOR activation and upregulation of Trem2 with enhanced β-Amyloid (Aβ) clearance, reduced spine loss, and improved cognitive function in the 5XFAD AD mouse model. Combined loss of Tsc1 and Trem2 in microglia led to reduced Aβ clearance and increased Aβ plaque burden revealing that Trem2 functions downstream of mTOR. Tsc1 mutant microglia showed increased phagocytosis with upregulation of CD68 and Lamp1 lysosomal proteins. In vitro studies using Tsc1-deficient microglia revealed enhanced endocytosis of the lysosomal tracker indicator Green DND-26 suggesting increased lysosomal activity. Incubation of Tsc1-deficient microglia with fluorescent-labeled Aβ revealed enhanced Aβ uptake and clearance, which was blunted by rapamycin, an mTOR inhibitor. In vivo treatment of mice of relevant genotypes in the 5XFAD background with rapamycin, affected microglial activity, decreased Trem2 expression and reduced Aβ clearance causing an increase in Aβ plaque burden. Prolonged treatment with rapamycin caused even further reduction of mTOR activity, reduction in Trem2 expression, and increase in Aβ levels. Together, our findings reveal that mTOR signaling in microglia is critically linked to Trem2 regulation and lysosomal biogenesis, and that the upregulation of Trem2 in microglia through mTOR activation could be exploited toward better therapeutic avenues to Aβ-related AD pathologies.

SIGNIFICANCE STATEMENT Mechanistic target of rapamycin (mTOR) signaling pathway is a key regulator for major cellular metabolic processes. However, the link between mTOR signaling and Alzheimer's disease (AD) is not well understood. In this study, we provide compelling in vivo evidence that mTOR activation in microglia would benefit β-Amyloid (Aβ)-related AD pathologies, as it upregulates Trem2, a key receptor for Aβ plaque uptake. Inhibition of mTOR pathway with rapamycin, a well-established immunosuppressant, downregulated Trem2 in microglia and reduced Aβ plaque clearance indicating that mTOR inactivation may be detrimental in Aβ-associated AD patients. This finding will have a significant public health impact and benefit, regarding the usage of rapamycin in AD patients, which we believe will aggravate the Aβ-related AD pathologies.

Introduction

Alzheimer's disease (AD) is characterized by numerous neuropathological hallmarks, including β-Amyloid (Aβ) plaques and Tau neurofibrillary tangles at the disease onset followed by loss of synapses and neurons toward the later stages of the disease (Ferrari and Sorbi, 2021; Knopman et al., 2021). Genome-wide association studies (GWAS) have identified over 20 AD-associated genetic variations in humans, the majority of which are genes related to microglial immune responses and phagocytosis (Lambert et al., 2013; Karch et al., 2014; Griciuc and Tanzi, 2021). This has drawn tremendous attention to the role of microglia in AD pathogenesis. During brain development or brain injury, microglia respond and express a wide range of molecules to initiate immune responses (Ajami et al., 2007; Askew et al., 2017; Pierre et al., 2017; Leng and Edison, 2021). In many disease states or aging mice or humans, subsets of microglia are primed generating a specific population of microglia referred as disease associated microglia (DAM). These microglia display upregulation of lysosomal, phagocytic and lipid genes, and downregulation of “homeostatic” genes, such as P2ry12/13, Cx3cr1, Tmem119 (Butovsky et al., 2014; Keren-Shaul et al., 2017; Deczkowska et al., 2018). In the case of AD, the aggregation of Aβ triggers both molecular and morphologic changes in microglia to surround Aβ deposits and thus prevent further aggregation of new Aβ plaques with existing plaques and subsequently reduce axonal dystrophy in the nearby neuronal area (Condello et al., 2015). Upregulation of AD-associated genes including APOE, CR1, CLU, TREM2, and CD33, point to excessive activation of the complement system possibly because of degeneration of neural tissue (McQuade and Blurton-Jones, 2019). An emerging hotspot in AD research is TREM2, a single transmembrane immune receptor expressed primarily in microglia (Painter et al., 2015; Ulland et al., 2017; Ulrich et al., 2017; Yeh et al., 2017; Gratuze et al., 2018; Ulland and Colonna, 2018; Sierksma et al., 2020; Ellwanger et al., 2021; SH Lee et al., 2021). While the role of Trem2 has been well established in AD pathology, loss of Trem2 has also been shown to cause impairment in microglial mechanistic target of rapamycin (mTOR) activation and metabolism (Ulland et al., 2017).mTOR signaling pathways are considered as essential regulators of basic cellular and metabolic processes (Saxton and Sabatini, 2017), playing a critical role in the development of diseases, including AD (Mueed et al., 2018). A major player in the mTOR signaling pathway is the mTOR protein, a serine/threonine kinase that is negatively regulated by Tsc1/Tsc2 protein complex which maintain a balance of Rheb protein in its GDP or GTP bound forms. In the absence of Tsc1 or Tsc2, higher levels of Rheb-GTP are achieved causing activation of mTOR pathway and its downstream effectors. However, treatment of mice and microglia with rapamycin, the mTOR inhibitor, showed inconsistent effects toward microglial activation (van Vliet et al., 2012, 2016; Nguyen et al., 2015). Recent RNA-seq analyses of mTOR deficient microglia from aged mouse brains showed reduced cytokine protein levels, decreased microglial activation and behavioral changes (Keane et al., 2021). However, detailed mechanisms by which mTOR signaling in microglia contributes to the pathogenesis of AD remain to be fully elucidated (Querfurth and Lee, 2021).

Here, we report that loss of Tsc1 specifically in microglia in the 5XFAD AD mouse model led to unique molecular changes in microglia. The Tsc1-deficient microglia showed morphologic changes, increased Trem2 expression, coupled with increased phagocytosis, enhanced lysosomal function and Aβ clearance. The combined loss of Tsc1;Trem2 in microglia abrogated mTOR activated enhanced Aβ uptake and clearance indicating that Trem2 functions downstream of mTOR. In vivo mTOR inhibition by rapamycin treatment compromised microglial functions, decreased Trem2 expression, augmented Aβ burden and exacerbated AD pathogenesis in the 5XFAD mice. Together, our studies provide in vivo evidence that mTOR activation-dependent upregulation of Trem2 in microglia is beneficial in the Aβ-associated AD mouse model and that inhibition of mTOR signaling in microglia by rapamycin downregulates Trem2 causing aggravation of Aβ-related AD pathologies.

Materials and Methods

Animals

Mouse strains Tsc1^Flox (#005680), Trem2^Flox (#029853), Cx3cr1-Cre (#025524), Cx3cr1-Cre-ERT2 (#021160) and 5XFAD (#006554) were all acquired from The Jackson Laboratory. All strains are cross bred with C57/BL6 wild-type (WT; Jax#000664) to maintain the line. The heterozygous littermate progeny from breeding were used as controls in all experiments, unless stated otherwise. All transgenic mice exhibited no gross anatomic defects, except for Tsc1^Flox/Flox;Cx3cr1-Cre (Tsc1^cKO) mice, which died around postnatal day (P)25 as has been reported previously (Zhao et al., 2018a). Tsc1^cKO mice were used at P0–P3 for establishing microglia cultures reported in this study. Matched numbers of mice of both sexes (males and females) were used in the study. The age, number and sex of mice used for individual experiments are indicated in the figure legends or main text. To allow inducible Cre-dependent recombination (for Cx3cr1-Cre-ERT line), all mice (including WT and 5XFAD) were administrated with tamoxifen (TAM; 0.8 mg/μl sunflower oil per kilogram of body weight) for 5 d at P30 (Saifetiarova et al., 2017), and then tested at two months of age for microglia phenotypic and/or morphometric analysis, or six to seven months of age (for transgenic mice in the 5XFAD background) for all AD-related Aβ pathologic analyses. All procedures involving mice were approved by the IACUC Committee for ethical treatment of experimental animals of University of Texas (UT) Health Science Center at San Antonio. All mouse lines and other reagents are listed in the Table 1.

Table 1.

List of reagents

Antibodies	Resource	Identifier
Rabbit anti-Tsc1	Cell Signaling	#6935
Rabbit anti-p-S6k (Thr389)	Cell Signaling	#97596
Rabbit anti-S6k	Cell Signaling	#2708
Rabbit anti-p-4EBP1(Thr37/46)	Cell Signaling	#2855
Rabbit anti-4EBP1	Cell Signaling	#9644
Rabbit anti-Iba1	Wako Chemicals USA	#019-19741
Rat anti-F4/80	Invitrogen	MA5-16624
Rat anti-CD68	AbD Serotec	MCA1957
Mouse anti-LAMP1	Abcam	#ab25630
Rabbit anti-NeuN	Cell Signaling	#24307
Mouse anti-NeuN	Millipore Sigma	MAB377
Sheep anti-mouse Trem2	R&D Systems	#AF1729
Rat anti-mouse Trem2	Abcam	#ab252876
Rabbit anti-APP (6e10)	BioLegend	#SIG-39320
Mouse anti-β-actin	Sigma	#A2228
Key chemicals
Methoxy-X04 (MX04)	TOCRIS	Catalog #4920
LysoTracker Green DND-26	Cell Signaling	Catalog #8783
Rapamycin	TOCRIS	Catalog #1292
FAM-Aβ	ANASPEC	AS-23525
Rapamycin and control diet	Rapamycin Holding Inc. & Lab Supply Diet	Custom made
Commercial kits
SuperGolgi kit	Bioenno Lifesciences	Catalog #003010
Anti-mouse β-Amyloid ELISA	ANASPEC	AS-55554
Softwares
NeuronStudio Software	Icahn School of Medicine at Mount Sinai
AnyMaze software	Stoelting Co
Mouse strains
C57/BL6 WT	The Jackson Laboratory	#000664
Tsc1Flox	The Jackson Laboratory	#005680
Trem2Flox	The Jackson Laboratory	#029853
Cx3cr1-Cre	The Jackson Laboratory	#025524
Cx3cr1-Cre-ERT2	The Jackson Laboratory	#021160
5xFAD	The Jackson Laboratory	#006554
Trem2KO	The Jackson Laboratory	#027197

Tissue preparation and immunofluorescence staining

Mice were anesthetized and pericardially perfused with 4% paraformaldehyde (PFA; in PBS) as described previously (Shi et al., 2018). The whole brain was removed and subjected to postfixation in 4% PFA overnight at 4°C. Following postfixation, the brains were washed in PBS three times and sliced into sections at 30-μm thickness, using a Vibratome (Leica). The sections were then incubated in blocking buffer (3% BSA, 1% goat serum, and 0.2% Triton X-100 in PBS) for 1 h at room temperature (RT), followed by primary antibodies in blocking buffer for overnight incubation and subjected to fluorescence secondary antibodies next day. All images were acquired under a Zeiss LSM 710 confocal microscope, and intensities of staining were quantified using ImageJ software. To assess Aβ plaque burden, three coronal sections spanning the cortex and hippocampus regions were imaged for each animal. The amyloid plaque burden (the size of the amyloid plaque or the area percentage of all plaques in total imaging area) was estimated in both the cortex and hippocampus for all sections with the Analyze Particles plugin of Fiji ImageJ software (NIH). Values from all sections for one animal were averaged to generate a mean plaque burden for each animal. The sex and number of animals in each experiment are indicated in the figure legends. The number of microglia (Iba1⁺ cells) was quantified in all sections using Cell Counter plugin of ImageJ. The immunofluorescence intensities of Iba1, F4/80, and Trem2 were quantified using the Measure function of ImageJ. The investigators were blind to the genotype for all the quantifications.

Measurement of aβ₄₂ by ELISA

Levels of Aβ₄₂ in cortical or hippocampal tissues of 5XFAD and all experimental mice were detected by using a colorimetric Aβ₄₂ ELISA kit (ANASPEC). In brief, tissue samples were extracted with Triton X-100 first and centrifuged for 30 min at 12,000 × g. The supernatant was used for estimating levels of the soluble Aβ₄₂, and the pellet was again homogenized in guanidine-HCl buffer: 5 m guanidine HCl, 50 mm Tris HCl (pH 8.0) and protease inhibitor cocktail for estimating the levels of Aβ₄₂ in the insoluble fraction. Diluted Aβ₄₂ peptides were used as standards and the estimation procedure was followed according to the manufacturer's instructions.

Immunoblotting

Cells or tissues were collected/homogenized and lysed on ice in RIPA buffer (25 mm Tris-HCl, pH 7.5, 150 mm NaCl, 1 mm EDTA, 1% NP-40, and 5% glycerol) with protease inhibitors (#88660SPCL, Thermo Fisher Scientific) and phosphatase inhibitor mix (sc-45044, Santa Cruz). All antibodies used for immunostaining and immunoblotting are listed in Table 1. For all protein level measurements, β-Actin was used as the loading control.

Microglia morphology analysis

A total of 20 of 0.25-μm interval stacked images were taken to perform the morphologic analysis for each microglia. Microglia were analyzed for changes in ramified morphology (processes and branches) using computer-aided skeleton analysis methods (Analyze Skeleton Plugin in ImageJ), as described previously (Young and Morrison, 2018). And 40–50 microglia from cortex and hippocampus areas were analyzed and averaged for each animal. At least three animals were taken for each genotype and/or conditions as described in the figure legends.

Primary microglia cell culture

Primary microglia cultures were prepared from P0 to P3 mouse pups. Briefly, brain tissues from pups were dissected and digested with Trypsin at 37°C for 30 min, and then passed through 70- and 45-μm cell strainer sequentially. Single cells were seeded in poly-D lysine-coated dishes, and the medium (DMEM/F12 + 10% FBS) was replaced every 3 days. After two weeks in culture, mild trypsinization steps were followed to obtain single layers of microglia (Saura et al., 2003). The microglia culture was again trypsinized and seeded in the proper 12-well dish for Immunoblotting or four-chamber slides for immunostaining or later dye-uptake experiments. Experiments were repeated at least three times using three different animals from the same genotypes and/or conditions.

In vitro phagocytosis assay

To assess the Aβ uptake, microglia cells derived from control, Tsc1^cKO, Trem2^cKO, or Dual^cKO, noted as +/+, Tsc1^−/−, Trem2^−/−, or Dual^−/−, were incubated with FAM-Aβ (200 nm) for 1 h at 37°C, and then fixed in 4% PFA and subjected to fluorescence microscopy immediately. To assess the dye clearance, FAM-Aβ containing medium was replaced after 1 h of incubation with fresh medium without FAM-Aβ, and then cells were imaged after 30 min, 1 h, and 2 h, respectively. Images were obtained using an LSM 710 confocal microscope (Zeiss). The percentage of FAM-Aβ positive cells or FAM-Aβ intensity in each cell were quantified using ImageJ software, and at least 150–200 cells from each group were analyzed from three independent experiments (cells for each genotype were derived and cultured from three different animals).

Labeling lysosomes in live microglia with LysoTracker

Primary microglia were cultured as described above. After obtaining pure microglia, cells were distributed evenly in four-chamber slides; and the next day, cells were replaced with HBSS containing 50 nm LysoTracker Green DND-26 (catalog #8783, Cell Signaling) for 30 min in a 37°C CO₂ incubator. Cells were washed with PBS for three times and directly subjected to confocal imaging. Eight to 10 random images were taken for each well, and 40–50 cells were quantified and averaged per cell for each genotype. The experiments were repeated three times for each genotype.

Microglia isolation from adult mouse brains

Fresh microglia were isolated from adult mouse brains using CD11b⁺-conjugated magnetic beads per Manufacturer's instructions (Miltenyi Biotec). Briefly, after perfusion with PBS, brains were dissected and enzymatically digested using Neural Tissue Dissociation kit (Miltenyi Biotec) for 45 min at 37°C. Then, tissue debris was removed by passing the whole suspension through a 40-µm cell strainer twice. Cells were re-suspended in 30% of Percoll (GE Healthcare) and centrifuged for 15 min at 700 × g without stopping brakes. The cell pellets were then washed again with HBSS and subjected to anti-CD11b+ beads. CD11b-positive fractions were collected and used for biochemical experiments.

RNA extraction and RT-PCR analyses

Total RNA from microglia cells or tissues was isolated by using PureLink RNA mini kit (Ambion) with DNase treatment. cDNA reverse transcription was performed with iScript Reverse Transcription Supermix from Bio-Rad. Quantitative RT-PCR was performed with SYBR green PCR Master Mix (Applied Bio System) on Applied Biosystems 7500 Real-Time PCR System with primers listed in Table 2. All experiments were repeated at least three times from three different animals for each genotype/condition.

Table 2.

List of primers for RT-PCR

Name of gene	Forward primer	Reverse primer
Gapdh	AGGTCGGTGTGAACGGATTTG	TGTAGACCATGTAGTTGAGGTCA
Tsc1	ATGGCCCAGTTAGCCAACATT	CAGAATTGAGGGACTCCTTGAAG
Trem2	CTGGAACCGTCACCATCACTC	CGAAACTCGATGACTCCTCGG
Ctsc	CAACTGCACCTACCCTGATCT	TAAAATGCCCGGAATTGCCCA
Ctsb	CAGGCTGGACGCAACTTCTAC	TCACCGAACGCAACCCTTC
Ctsd	GCTTCCGGTCTTTGACAACCT	CACCAAGCATTAGTTCTCCTCC
Ctse	TGTATTCGAGTGTCAATGAACCC	GGCTGGTGCAGTACACAGAAG
C1qa	AAAGGCAATCCAGGCAATATCA	TGGTTCTGGTATGGACTCTCC
C1qb	CGTCGGCCCTAAGGGTACT	GGGGCTGTTGATGGTCCTC
C1qc	CCCAGTTGCCAGCCTCAAT	GGAGTCCATCATGCCCGTC

mRNA sequencing and analysis

mRNA samples from whole brains were prepared with PureLink RNA mini kit (Ambion), and then submitted to UT Health Genome Sequencing Facility. cDNA libraries were prepared using the TruSeq Stranded mRNA LT Prep kit (Illumina) and following Manufacturer's instruction. Libraries were sequenced on a HiSeq 3000 instrument (Illumina) using single-end 50-bp sequencing. Data were analyzed with Ingenuity Pathway Analysis software (IPA; QIAGEN) at the UT Health Bioinformatics core.

Golgi–Cox staining and dendrite analysis

Golgi–Cox staining was used to analyze the dendritic spines of hippocampal neurons. Mice were anesthetized and pericardially perfused with 20 ml of PBS. The brains were removed and processed with Golgi–Cox impregnation and staining with SuperGolgi kit (#003010, Bioenno Tech, LLC). After impregnation, 150-μm sections were obtained using a Vibratome (Lecia), mounted on gelatin-coated glass slides, and stained. Stacked images were taken with a Zeiss Imager II deconvolution microscope with SlideBook software using a 100× oil objective. Dendritic spines were then reconstructed and analyzed with NeuronStudio (v0.9.92, Mount Sinai School of Medicine). A total of 100–150 dendrites were taken for images and analyzed for each mouse. Three animals were taken for each genotype, as indicated in figure legends.

Rapamycin diet and treatment

Rapamycin compound was obtained from Rapamycin Holdings Inc as microencapsulated forms, which has been used widely in other publications (Lin et al., 2013, 2017; Kaeberlein and Galvan, 2019; Van Skike et al., 2021). The diet was then custom-made at LabSupplies at 143 ppm encapsulated rapamycin (14 ppm active rapamycin), which is equivalent to 2.24 mg/kg body weight/d, given the average food consumption per day is 5 g for the mice weighed around 30 g. The control diet was made in the same manner with the same microencapsulated materials without rapamycin.

Behavioral tests

The Morris water maze test was performed with minor adjustments as previously described (Vorhees and Williams, 2006). The spatial memory testing was performed in a circular tank (2 m in diameter) filled with opacified water at RT. All movement and tracking of mice were captured and analyzed by ANY-maze software. The water area was virtually divided into four quadrants, with one containing a hidden platform (10 cm in diameter) that was submerged 1 cm below the water level. Four visual cues were placed on the wall of the tank above the water as spatial references. For each trial, mice were placed in the water facing the tank wall at different starting points to prevent potential strategy learning. Mice were allowed to swim to search for the platform for 1 min; and they were guided toward it if they could not find the platform. We performed five trials for each mouse per day for five consecutive days. On day 6, the platform was removed, and mice were allowed to swim precisely for 1 min, and the number of entries into the target quadrant was recorded and analyzed.

Quantification and statistical analyses

For most quantitative results, statistical analysis between multiple groups was performed by one-way ANOVA (nonparametric), and post hoc test analysis was performed using Tukey's test. An unpaired t test with Welch's correction was performed for normally distributed datasets between two groups. Data from behavioral duration studies were analyzed by two-way ANOVA followed by Tukey's post hoc test. All data were analyzed and presented with GraphPad Prism software. All n and p values and statistical tests are indicated in figure legends.

Results

mTOR activation causes metabolic changes and upregulation of Trem2 in microglia

Given the importance of microglia in proper neural development, and the many functions of mTOR signaling pathways, we used an inducible microglia-specific Cre strain (Cx3cr1-Cre/ERT2) to specifically ablate Tsc1 in Tsc1^Fx adult microglia (Tsc1^iKO). After TAM injection at P30, Tsc1 was selectively deleted in microglia obtained from two-month-old mouse brains (one month after TAM injection; Fig. 1A,B). Loss of Tsc1 led to activation of mTOR signaling (as assessed by increased levels of pS6K and p4EBP1; Fig. 1A,C,D; Kwiatkowski et al., 2002; McMahon et al., 2012). The microglia in Tsc1^iKO mice adopted a unique morphology with more round, less ramified and enlarged cell bodies (Fig. 1F,H vs E,G). Phase-contrast images of primary microglia culture from Cx3cr1-Cre;Tsc1^Fx (Tsc1^cKO) show short processes and fewer branches (Fig. 1J vs I) which was also observed in immunofluorescence analysis (Fig. 1E,G and F,H; quantified in Fig. 1L–N). More importantly, Tsc1^iKO microglia were programmed toward an inflammatory-like state, as revealed by significant upregulation of a known microglial marker F4/80 (Fig. 1Fb vs Eb), which has been previously proposed to be highly expressed by inflammatory macrophages (YS Lee et al., 2018b). In addition, we observed that mTOR-activated microglia proliferated extensively throughout the entire brain in two-month-old Tsc1^iKO mice, with a nearly 2-fold increase as compared with the microglia in age-matched control mice (Fig. 1K). These data suggest that elevated mTOR signaling not only altered the morphology of microglia, but also promoted microglial proliferation.

Figure 1.

Loss of Tsc1 causes mTOR activation and morphologic and physiological changes in microglia. A, Representative immunoblots showing protein levels of Tsc1, pS6K (Thr389), total S6K, p4EBP1 (Thr37/46), and total 4EBP1 in Tsc1^Fx and Tsc1^iKO microglia. Microglia were freshly isolated by CD11b antibody-conjugated magnet beads from two-month-old mice (one month after TAM injection for both genotypes). Each lane represents microglia lysates from individual animals. B–D, Relative protein levels of Tsc1, pS6K, and p4EBP1 were normalized and quantified (n = 3M/3F for each genotype). E, F, Representative immunostaining images of brain cortex regions using Iba1 (a, green), F4/80 (b, blue), and NeuN (c, red) from two-month-old of Tsc1^Fx and Tsc1^iKO mice. Arrows in Fb point to F4/80 positive microglia (scale bar: 100 μm). G, H, High-magnification images of Iba1+ single microglia from Tsc1^Fx and Tsc1^iKO. Arrows point to microglia soma and arrowheads point to processes. Scale bar: 50 μm. I, J, Phase-contrast images of microglia from in vitro culture. Arrows point to soma and arrowheads point to processes. Scale bar: 5 μm. K–N, Morphologic parameters of 40–50 microglia per animal were quantified from Tsc1^Fx and Tsc1^iKO (microglia numbers or images were obtained in cortex and hippocampus regions, n = 3M/3F). O, P, Two heatmaps display hierarchical clustering of top 50 upregulated (O) and 13 downregulated DGEs (P) in all samples analyzed (n = 2 M/1F). Q, The pathway analysis showing the top eight pathways that were significantly altered in Tsc1^iKO comparing to Tsc1^Fx mouse brains. R–T, The fold changes for selected gene families showing C1q family (R), cathepsin proteins (S), and lysosomal proteins (T). From B–D and K–N, unpaired Student's t test has been performed between two groups, **p < 0.01, ***p < 0.001 as compared with WT. All data are presented as mean ± SEM.

As activated microglia are known to produce unique inflammatory cytokines or mediators, we evaluated Tsc1^iKO mouse brains for their transcriptional profiles by RNA-seq (dataset deposited to GEO), which identified 277 upregulated (50 shown in Fig. 1O) and 13 downregulated (Fig. 1P) differentially expressed genes (DEGs). Cluster analysis showed that “glycoprotein,” “immune system process,” “lysosome” are the top three pathways that are regulated in Tsc1^iKO mutant mouse brains (Fig. 1Q). Complement pathway genes, such as C1qa, C1qb, and C1qc (Fig. 1R) are markedly increased; Cathepsin family genes, such as Ctsc, Ctsz, Ctsd, and Ctse (Fig. 1S), and lysosomal genes, such as Laptm5, and CD68 (Fig. 1T) and others, are significantly upregulated, suggesting that Tsc1-deficient microglia transformed to an activated state with augmented lysosomal activities (Lowry and Klegeris, 2018). Surprisingly, there was no significant induction of microglia proinflammation-signature genes, including IL-1a, IL-1b, IL-6, Nos2, Hmgb1, Cox2, Mmp family genes, which is consistent with previous reports in uninducible Cx3-cr1-Cre;Tsc1^Fx mice (Zhao et al., 2018a). The activation of mTOR in microglia downregulated genes specific to homeostatic microglia including purinergic receptors (P2ry12) and Tmem119 (important for cell invasion and migration (Fig. 1P) suggesting that such changes greatly diminished the induced motility of microglia (Haynes et al., 2006; van Wageningen et al., 2019). Together, these data suggest that constitutive activation of mTOR signaling in microglia does not provoke a typical proinflammatory response, despite the elevation of F4/80 and lysosomal proteins.

mTOR activation upregulates AD risk genes including Trem2

Surprisingly several AD-risk factor genes were also altered in Tsc1^iKO mouse brains, including Ms4a7 family, Trem2, Plcg2, and CD33 (Fig. 2A). Since the role of Trem2 is well established in AD pathogenesis, we wanted to further address how mTOR signaling is linked with Trem2. We first validated the Trem2^Fx strain used for the first time in the current studies using microglia from Actin-Cre;Trem2^Fx mice (labeled as Trem2^KO) with anti-Trem2 rat antibody (5F4, ab252876, Abcam), which confirmed loss of Trem2 (Fig. 2B). Most importantly, we observed a marked increase in Trem2 (>4-fold) at the protein level in Tsc1^iKO brains as determined by immunoblotting measurements (Fig. 2C; quantified in Fig. 2D). In Trem2^iKO and combined Tsc1^iKO;Trem2^iKO (Dual^iKO), the levels of Trem2 were barely detectable (Fig. 2C; quantified in Fig. 2D). The increase in Trem2 levels was further confirmed by immunofluorescence intensity measurements in Tsc1^iKO mice (Fig. 2F, compare to 2E; quantified in Fig. 2G). Tsc1^iKO mice also showed a 3- to 4-fold increase in Trem2 at the mRNA level (Fig. 2H). Interestingly, Tsc1^iKO and Dual^iKO mice at two months old showed an increase in the number of microglia compared with WT and Trem2^iKO mice (Fig. 2I). There was no difference between the total microglia numbers in the age-matched two-month-old brains in the WT and Trem2^iKO mice. Dual^iKO mice showed a slight reduction as compared with the microglia numbers in Tsc1^iKO mice (Fig. 2I). These data again confirm a nonessential role of Trem2 in regulating microglia proliferation.

Figure 2.

Loss of Tsc1 in microglia upregulates Trem2 and causes Trem2-dependent activation of microglia. A, The fold changes for selected AD-risk genes in Tsc1^iKO versus Tsc1^Fx mice. B, Immunoblot of total Trem2 protein levels from purified microglia from one-month-old WT and Trem2^KO (Trem2^Fx;Actin-Cre). Each lane presents mixed microglia isolation from three mice. C, Representative immunoblot of total Tsc1 and Trem2 protein levels from purified microglia from two-month-old Tsc1^Fx, Tsc1^iKO, Trem2^iKO, and Dual^iKO. Each lane presents a mixed microglia isolation from three mice. D, Relative levels of Trem2 normalized to β-Actin in Tsc1^Fx versus Tsc1^iKO (n = 3M/3F). E, F, Representative immunostaining of cortical brain sections for Trem2 (green) and Iba1 (red) from Tsc1^Fx (E) and Tsc1^iKO (F) mice. Arrowheads point to microglia soma. Scale bar: 5 μm. G, Quantification of Trem2 immunofluorescence staining in Iba1+ cells (50 microglia were analyzed for each animal; n = 3M/3F for each genotype). H, RT-PCR analysis for Trem2 mRNA levels from Tsc1^Fx and Tsc1^iKO microglia normalized to β-Actin (n = 3M/3F for each genotype). I, Quantification of the number of Iba1+ cells in cortex and hippocampus regions in two-month-old Tsc1^Fx, Tsc1^iKO, Trem2^iKO, and Dual^iKO (n = 3M/3F for each genotype). J–L, Representative immunostaining of brain section against p4EBP1 (green) and F4/80 (red) from cortex region of Tsc1^Fx (J), Tsc1^iKO (K), and Dual^iKO (L). M, N, Quantification of immunofluorescence staining of p4EBP1 (M) and F4/80 (N) in microglia from Tsc1^Fx, Tsc1^iKO, and Dual^iKO (50 microglia were analyzed for each animal, and n = 3M/3F for each genotype). O, Immunoblots showing protein levels of Trem2 and pS6K in primary microglia culture derived from Tsc1^cKO treated with DMSO or rapamycin. Each lane represents one batch of primary microglia culture derived from four to five P0–P3 pups, and rapamycin was applied to the culture at 10 nm in a serum-free medium for 24 h. P, Quantification of protein levels of Trem2 and pS6K in primary microglia culture derived from Tsc1^cKO microglia treated with DMSO (left) or rapamycin (right; n = 3). Q, In vitro uptake of FAM-Aβ (green) by primary microglia derived from Tsc1^Fx, Tsc1^cKO, or rapamycin-treated microglia cells from Tsc1^cKO and stained against Iba1 (red) and DAPI (blue). Scale bar: 25 μm. R, The average intensity of FAM-Aβ in each FAM-positive cell from Tsc1^Fx, Tsc1^cKO, or rapamycin-treated Tsc1^−/− microglia. Comparisons between genotypes were performed by one-way ANOVA with Tukey's post hoc test (D, I, M, N, R). And unpaired Student's t test has been performed between two groups (G, H, P). *p < 0.05, **p < 0.01, ***p < 0.001 as compared with WT, and #p < 0.05, ##p < 0.01, ###p < 0.001 as compared within experiment groups as indicated.

We next determined the correlation between microglia activation by mTOR and Trem2 in Dual^iKO in which both Tsc1 and Trem2 were deleted specifically in microglia. All mice were injected with TAM at one month of age and examined at two months (one month after injection). Furthermore, we observed that loss of Trem2 led to a significant reduction in the levels of the microglial surface protein, F4/80 in the Dual^iKO microglia (Fig. 2L, compare with 2K, red), although mTOR signaling was still activated as a result of loss of Tsc1. In addition, microglia in Dual^iKO mice underwent significant morphologic changes from Tsc1^iKO microglia (Fig. 2L vs 2K) and were in a less activated state, and closer to the Tsc1^Fx controls (Fig. 2L). Quantification of the fluorescent intensities of p4EBP1 (Fig. 2M) and F4/80 revealed a significant reduction in Dual^iKO compared with Tsc1^iKO microglia. These data suggest that although mTOR regulates the expression of Trem2, Trem2 might also cause dysfunction in mTOR signaling pathway in a feedback loop as has been suggested previously (Ulland et al., 2017).

Furthermore, we wanted to determine whether in vitro inhibition of mTOR signaling by specific inhibitor, rapamycin, will affect the expression of Trem2 in primary microglia cultures as mTOR activation caused an increase in Trem2 expression. We prepared primary microglia culture from Tsc1^cKO (Tsc1^Fx;Cx3cr1-Cre) P0-P3 pups, and applied rapamycin treatment (10 nm for 24 h). As shown in Figure 2O,P, rapamycin treatment significantly reduced the Trem2 protein levels close to 20% of the levels in Tsc1^−/− microglia and reduced the levels of the mTOR signaling downstream protein pS6K. Next, we evaluated whether mTOR activation affected the phagocytotic activity of cultured microglia from Tsc1^cKO. We conducted FAM-labeled Aβ fluorescent dye uptake assay as Trem2 has been shown to be the cell surface receptor for Aβ (Zhao et al., 2018b). FAM-Aβ uptake assay showed a marked increase in uptake of FAM-Aβ by Tsc1^cKO microglia compared with Tsc1^Fx microglia (Fig. 2Q; quantified in Fig. 2R) after only 1-h incubation with FAM-Aβ dye. However, on incubation with rapamycin, the uptake of FAM-Aβ was significantly reduced (Fig. 2Q; quantified in Fig. 2R) compared with Tsc1^−/− microglia. Together, these data show that suppression of mTOR signaling has significant inhibitory effects on Trem2 protein expression. In addition, mTOR activation promoted FAM-Aβ uptake while mTOR inhibition by rapamycin led to a reduction of FAM-Aβ uptake, most likely through downregulation of Trem2 expression, further highlighting the beneficial impact of mTOR activation in modulating phagocytosis by microglia.

mTOR activation in microglia in the 5XFAD AD mouse model ameliorates Aβ pathology

Since mTOR-activated microglia expressed elevated levels of Trem2 (Fig. 2), we investigated whether activation of mTOR in microglia will impact Aβ deposition in the 5XFAD AD mouse model (Oakley et al., 2006; Zhao et al., 2018b). We generated a series of mouse lines in the 5XFAD background lacking Tsc1 (5XFAD;Tsc1^iKO) or Trem2 (5XFAD;Trem2^iKO) or both Tsc1 and Trem2 (5XFAD;Dual^iKO; Fig. 3A, schematic). All mice were administrated with TAM for 5 d at P30, and then tested at six to seven months of age for Aβ deposition, morphometric analyses, and other pathologic parameters, such as spine densities, memory retention test and microglia responses. Coronal brain sections were analyzed for Aβ plaque burden in cortex and hippocampus using anti-Aβ antibody and Aβ-binding dye Methoxy-X04 (MX04; Ulland et al., 2017; Fig. 3B–I), which revealed a significant reduction of Aβ plaque load in 5XFAD;Tsc1^iKO mice (Fig. 3Da,b), in both the cortex and hippocampus (Fig. 3Ea,b); such reduction was abolished in 5XFAD;Trem2^iKO (Fig. 3F,G) and 5XFAD;Dual^iKO mice (Fig. 3H,I), indicating that mTOR activation in microglia led to a reduction in plaque numbers. Further quantification based on Aβ plaque numbers revealed that overall plaque numbers were significantly reduced in 5XFAD;Tsc1^iKO mice compared with 5XFAD and 5XFAD;Trem2^iKO or 5XFAD;Dual^iKO mice (Fig. 3J). More importantly, the 5XFAD;Trem2^iKO or 5XFAD;Dual^iKO mice showed increased plaque numbers compared with 5XFAD mice (Fig. 3J). We next determined whether mTOR activation in microglia affected the Aβ plaque size distribution in 5XFAD;Tsc1^iKO mice. We quantified plaques based on their diameter and created three groups of plaques <10, 10–40, and >40 μm. As shown in Figure 3K, there was a marked increase in <10 μm plaques and a significant decrease in >40 μm in 5XFAD;Tsc1^iKO mice compared with other genotypes indicating that plaque size distribution is affected in 5XFAD;Tsc1^iKO mice. On the other hand, there was a significant increase in >40 μm in 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO mice (Fig. 3K). These data demonstrate for the first time that prolonged activation of mTOR signaling in microglia has a significant beneficial impact on Aβ clearance in an AD mouse model. We further quantified the levels of Aβ₄₂ in the cortex and hippocampus of the 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO mice using Aβ₄₂ ELISA assay (AS-55554, ANASPEC). We detected a significant reduction in insoluble, guanidine-HCl-extracted Aβ₄₂ in both the cortex and hippocampus regions of 5XFAD;Tsc1^iKO mice (Fig. 3L). Interestingly, the levels of insoluble Aβ₄₂ in the hippocampal region of 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO mice significantly increased; while, in cortex region, only the Aβ₄₂ levels in 5XFAD;Dual^iKO showed a significant increase compared with 5XFAD mice. There was no significant difference between genotypes regarding the levels of Triton X-100 soluble Aβ₄₂ (data not shown)_. These data further underscore that mTOR activation reduced Aβ levels in Tsc1^iKO mice. Importantly, loss of Trem2 or combined loss of Tsc1/Trem2 indeed led to an increased plaque burden and an increase in the levels of Aβ compared with 5XFAD suggesting that Trem2 is indeed essential for Aβ plaque reduction and clearance by microglia.

Figure 3.

mTOR activation reduces Aβ plaque burden and ameliorates amyloid pathology. A, Schematic showing the timeline of experiments and mouse models (color-coded). B, D, F, H, Representative immunostaining of whole-brain coronal sections with anti-Aβ (Ba, Da, Fa, Ha, red) and staining with fluorescent dye MX04 (Bb, Db, Fb, Hb, green) from 5XFAD and 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. Scale bar: 1 mm. C, E, G, I, Representative immunostaining of magnified images of cortical (Ca, Ea, Ga, Ia) and hippocampal section (Cb, Eb, Gb, Ib) with anti-NeuN (red) and staining with fluorescent dye MX04 (green) from 5XFAD and 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. Scale bar: 50 μm. J, The number of plaques quantified in 5XFAD (n = 3M/4F) and 5XFAD;Tsc1^iKO (n = 5 M/4F), 5XFAD;Trem2^iKO (n = 3M/3F), and 5XFAD;Dual^iKO (n = 3M/3F). The number of plaques for each animal was averaged from three sections at different depths in the cortex and hippocampal areas. K, Number of plaques grouped according to the size of plaques (diameter) at <10 μm, 11–40 μm and >40 μm and quantified for the whole brains from 5XFAD and 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. Mice were the same cohort as in J. L, Protein levels of Aβ₄₂ in insoluble, guanidine-HCl-extracted fraction from cortex and hippocampus of 5XFAD (n = 4 M/4F), 5XFAD;Tsc1^iKO (n = 5 M/4F), 5XFAD;Trem2^iKO (n = 3M/5F) and 5XFAD;Dual^iKO (n = 3M/4F; normalized to total protein levels). M–P, Representative immunostaining of brain sections from 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO with anti-Aβ antibody (red) and MX04 dye (green) shown in M–P, respectively Scale bar: 25 μm. Q, Types of Aβ plaques defined by morphology and compactness. R, Quantification of plaque types in 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. Mice were the same cohort as in J. Comparisons between genotypes were performed by one-way ANOVA with Tukey's post hoc test; * or #p < 0.05, ** or ##p < 0.01, *** or ###p < 0.001. Data are presented as mean ± SEM.

To further examine whether the activation of mTOR in microglia was able to alter the Aβ plaque properties through regulation of Trem2 expression based on previous studies (Wang et al., 2016; Yuan et al., 2016; CYD Lee et al., 2018a), we analyzed distinct forms of Aβ plaques in the cortex region of respective mutants using anti-Aβ antibody in combination with MX04 dye. In 5XFAD;Tsc1^iKO mice, there were more inert Aβ plaques (strong MX-04 with minor anti-Aβ staining), less compact plaques and significantly reduced filamentous Aβ plaques (less MX-04 with diffuse anti-Aβ staining; Fig. 3N), compared with 5XFAD mice (Fig. 3M). In 5XFAD;Trem2^iKO (Fig. 3O), or 5XFAD;Dual^iKO (Fig. 3P), the Aβ plaques exhibited the opposite phenotypes to 5XFAD;Tsc1^iKO mice and phenotypically were closer to the parental 5XFAD. However, the percentage of compact Aβ plaques was slightly increased in 5XFAD;Trem2^iKO, or 5XFAD;Dual^iKO (Fig. 3R) compared with parental 5XFAD mice, which is consistent with the overall increase in Aβ plaque protein levels in these mice. The types of plaques that are distinct and morphologically discernible (Yuan et al., 2016; Wang et al., 2020) are shown in Figure 3Q, and all plaque types across relevant genotypes were quantified according to these morphologic characteristics shown in Figure 3R. Taken together, our data reveal that mTOR activation in microglia strongly influences both the plaque numbers and morphologies, and ameliorates Aβ pathology in 5XFAD mice, and that genetic deletion of Trem2 worsens the Aβ load and pathology in 5XFAD;Tsc1^iKO mice.

mTOR activation in microglia in the 5XFAD AD mouse model improves cognitive function

Given the reduction in Aβ plaque burden in 5XFAD;Tsc1^iKO mice, we sought to determine whether mTOR activated microglia in 5XFAD mice had any effect on the dendritic spine density; as activated microglia have been shown to mediate synaptic pruning and remodeling of spine formation (Imai et al., 2007; Wake et al., 2009). We stained brains using the Golgi–Cox method (Koyama and Tohyama, 2012) and analyzed the number of dendritic spines on nonprimary apical dendrites in the CA1 region of hippocampi from relevant genotypes (Fig. 4A; quantified in Fig. 4B). The total spine density was significantly reduced in 5XFAD mice compared with WT controls (Fig. 4Aa,b). The spine density was significantly preserved in 5XFAD;Tsc1^iKO mice (Fig. 4Ac); however, the spine density was reduced in 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO mice, similar to 5XFAD (Fig. 4Ac,d; quantified in Fig. 4B).

Figure 4.

Activation of mTOR in microglia reduces dendritic spine loss and improves cognitive functions. A, Representative Golgi–Cox staining images of brain sections from WT (CONT), 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. Scale bar: 5 μm. B, Quantification of spine densities in axon segments from CONT, 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO (50 spines were taken for each animal, and the average of spine densities were calculated; n = 3M/3F for each genotype). C, D, Results of the Morris water maze tests CONT (open circle, n = 8 M/7F) 5XFAD (black circle, n = 7 M/8F), and 5XFAD;Tsc1^iKO (red circle, n = 8 M/7F). Escape latency showing the average time for each mouse to reach the hidden platform during the training course (C; two-way ANOVA with Tukey's post hoc test, *p < 0.05, **p < 0.01, ***p < 0.001). Path length showing the length of swimming path taken to find the platform during the training course (D; two-way ANOVA with Tukey's post hoc test). E, Typical escape route (the swimming path of mice) in CONT, 5XFAD, and 5XFAD;Tsc1^iKO. F, The time each mouse spent in the hidden platform quadrant on day 6, in 1-min duration. Each dot represents one mouse. G, The number of entries to the hidden platform area on day 6 by CONT, 5XFAD, 5XFAD;Tsc1^iKO mice recorded in 1 min. Each dot represents one mouse. Comparisons between genotypes were performed by one-way ANOVA with Tukey's post hoc test; *** or ###p < 0.001. All data are presented as mean ± SEM.

Furthermore, we evaluated whether mTOR activation in microglia has any impact on the spatial learning and memory function in 5XFAD;Tsc1^iKO by the Morris water maze at the age of seven months. Matched numbers of mice of both sexes (males and females) were used in the study. The age, number and sex of mice used for individual experiments are indicated in the figure legends. On the training day 1 for all groups, no significant differences were observed in swim speed or latencies to reach the visible platform (day 1 in Fig. 4C,D). For training duration (days 1–5), to measure the time the mice spent before finding the visible platform, a two-way repeated-measures ANOVA method were employed, and revealed significant effects for duration days (F_(4,210) = 18.48, p < 0.0001) and for genotype groups (F_(2,210) = 16.25, p < 0.0001), but not for interaction (F_(8,210) = 0.5368, p = 0.8280; Fig. 4C). Similarly, there is significant improvement for the distance the mice traveled before finding the platform, same ANOVA measures were used for days (F_(4,210) = 64.86, p < 0.0001) and for genotype groups (F_(2,210) = 12.41, p < 0.0001), but not for interaction (F_(8,210) = 1.497, p = 0.1597; Fig. 4C). 5XFAD;Tsc1^iKO mice displayed improved spatial learning versus 5XFAD mice on day 4 and day 5 (p = 0.0308 and p = 0.0452), respectively, two-way ANOVA, Tukey's test (Fig. 4D).

On day 6, memory retention was evaluated using the hidden platform test (represented swimming route for each genotype; Fig. 4E), and data were analyzed with one-way ANOVA (Fig. 4F,G) [(_2,42) = 19.20, p < 0.0001). The 5XFAD mice were much less efficient at finding the hidden platform than the CONT mice (less time spent in the platform area (Fig. 4F), and fewer entries to the quadrant (Fig. 4G; one way ANOVA, p < 0.0001); however, 5XFAD;Tsc1^iKO mice displayed improved memory retention learning versus 5XFAD mice (Fig. 4G; one-way ANOVA, p < 0.001), which showed no significant difference while comparing to CONT group (p = 0.4780 in Fig. 4F, p = 0.1261 in Fig. 4G). These data suggest that Tsc1 loss-dependent activation of microglia significantly improved memory retention in 5XFAD mice and that this improvement in 5XFAD;Tsc1^iKO mice could be attributed to either reduced Aβ amyloid plaque deposition or increased clearance allowing for better preservation and/or maintenance of dendritic spines.

mTOR activation alters microglial activation and response to Aβ deposition

To assess the microglial response in 5XFAD mice and how that compared with 5XFAD;Tsc1^iKO mice, we examined mTOR activation and microglial activation at six months after TAM injection. mTOR signaling was not significantly activated in 5XFAD microglia (Fig. 5Aa,c), but highly activated in 5XFAD;Tsc1^iKO microglia, which showed robust staining for p4EBP1 (Fig. 5Ba,c) and the microglial marker F4/80 was also upregulated in p4EBP1-positive cells in six-month-old 5XFAD;Tsc1^iKO (Fig. 5Bb,c), suggesting that sustained activation of mTOR still induced activation of microglia in aged mice. As expected, microglia in 5XFAD;Trem2^iKO did not increase the expression of either p4EBP1 or F4/80, similar to what we observed in the parental 5XFAD mice (Fig. 5C, compare to 5A). Interestingly, consistent with the data above, loss of both Tsc1 and Trem2 in 5XFAD;Dual^iKO microglia, significantly reduced the expression of F4/80 (Fig. 5Db vs 5Bb), but p4EBP1 expression was only modestly affected (Fig. 5Da vs Ba; quantified in Fig. 5E). These data show that microglial activation does not occur at significant levels in 5XFAD mice and that robust activation of microglia following mTOR activation requires Trem2.

Figure 5.

mTOR activation enhances microglial Aβ clearance in AD mouse models. A–D, Brain sections from 5XFAD (A), 5XFAD;Tsc1^iKO (B), 5XFAD;Trem2^iKO (C), and 5XFAD;Dual^iKO (D) were immunostained with p4EBP1 (a, green) and F4/80 (b, red), and merged pictures were shown in c (scale bar: 5 µm). The images were obtained from cortical regions. E, The relative p4EBP1 or F4/80 immunostaining per cell was quantified for all four genotypes. All mice used in this figure were the same mouse cohort used in Figure 3J. F–I, Representative immunostaining of brain sections from 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO using antibodies against Iba1 (green), Aβ (red), and F4/80 (blue). Arrows point to microglia showing F4/80 immunofluorescence (scale bar: 20 μm). J, Quantification for Iba1⁺ microglia numbers from cortex and hippocampus regions in 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. K, Quantification for Iba1⁺ microglia soma volume in 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. L, Quantification for the length of the microglial processes in 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. M, Quantification in percentage of F-480/Iba1+ cells in 5XFAD, 5XFAD;Tsc1^iKO, 5XFAD;Trem2^iKO, and 5XFAD;Dual^iKO. N, Quantification of the number of microglia associated with Aβ plaques, grouped by the sizes of plaques <10, 10–40, and >40 μm from 5XFAD (n = 3M/4F), 5XFAD;Tsc1^iKO (n = 5 M/4F), 5XFAD;Trem2^iKO (n = 3M/3F), and 5XFAD;Dual^iKO (n = 3M/3F). Each dot represents the number of microglia associated with one plaque. O, The relative mRNA expression levels of Cathepsin family genes, normalized to the house-keeping gene Gapdh, in all groups including age and sex marched WT controls (white bar, n = 3M/3F). P, The relative mRNA expression levels of Complement family genes, normalized to the house-keeping gene Gapdh, in all groups including age and sex marched WT controls (white bar, n = 3M/3F). Comparisons between genotypes were performed by one-way ANOVA with Tukey's post hoc test; * or #p < 0.05, ** or ##p < 0.01, *** or ###p < 0.001. All data are presented as mean ± SEM.

Consistent with previous observations (Stence et al., 2001), in the 5XFAD AD disease model microglia that respond to Aβ amyloid deposition exhibit hypertrophic amoeboid features with shortened and thickened processes (Fig. 5F). These typical reactive microglia occupy the center of the Aβ plaques and express noticeable levels of F4/80 (Fig. 5Fc,d, arrowheads). In 5XFAD;Tsc1^iKO mice, the total number of microglia was upregulated as similar as observed in normal Tsc1^iKO mice (Fig. 5J), and the F4/80 levels were upregulated in all microglia because of mTOR activation (Fig. 5Gc,d, arrows; quantified in Fig. 5M). mTOR-activated microglia in 5XFAD;Tsc1^iKO were morphologically different from Aβ-stimulated microglia and were generally more “activated” or “primed” with larger cell bodies, lesser branches and shorter processes (quantified in Fig. 5K,L). Furthermore, deletion of Trem2 in 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO further reduced the number of microglia (Fig. 5J) and abolished the expression of F4/80 (Fig. 5Hc,d,Ic,d, arrows; quantified in Fig. 5M), suggesting the critical role of mTOR-induced, but Trem2-dependent, activation in maintaining microglial homeostasis in AD disease states.

We also determined whether mTOR activation in microglia altered microglial association with Aβ plaques and whether the number of microglia per Aβ plaque was affected. As the volume of Aβ plaques varies across the brain regions in the 5XFAD AD mouse model, which generates significant differences in the number of microglia surrounding the plaques, we quantified and compared the number of microglia that surround smaller Aβ plaques (<10 μm), medium-size plaques (10–40 μm), and large plaques (>40 μm). As shown in Figure 5N, the 5XFAD;Tsc1^iKO mice showed a marked increase in the number of microglia (Iba1⁺) per Aβ plaques, especially in the categories of smaller Aβ plaques and medium-size Aβ plaques compared with other genotypes. As the total number of large plaques (>40 μm) was significantly less in 5XFAD;Tsc1^iKO mice, the number of microglia associated with these plaques was not significantly different from the ones in the 5XFAD mice (Fig. 5N). In addition, we observed a marked reduction in microglia numbers associated with all Aβ plaques in 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO mice, regardless of the plaque sizes, further underscoring that mTOR activation leads to enhanced Trem2-dependent Aβ plaque clearance. In addition, we validated the expression of key molecules that are related to mTOR-activated microglia (Fig. 1), such as Cathepsin family genes (Fig. 5O) and Complement pathway genes (Fig. 5P) that show mRNA expression levels in all genotypes in the 5XFAD background. Indeed, the priming status of microglia from 5XFAD;Tsc1^iKO mice was unique and fundamentally different from the Aβ-stimulated/activated microglia, with enhanced levels of Cathepsin and Complement pathway genes; however, such increase in expression was markedly abolished in 5XFAD;Trem2^iKO and 5XFAD;Dual^iKO mice, suggested an indispensable role of Trem2 in regulating the metabolic status of mTOR-activated microglia during the pathogenesis of AD. These specific changes in microglia may very likely contribute to the reduced Aβ plaque burden and improvement in cognitive functions in 5XFAD;Tsc1^iKO mice.

mTOR activation in microglia increases lysosomal functions and Aβ phagocytosis

The morphologic changes and upregulation of the clearance ability of Tsc1^iKO microglia suggested an altered phagocytic function (Brown and Neher, 2014). In the transcriptome profiling, we also noticed altered expression levels of lysosomal genes in Tsc1^iKO mouse brains (Fig. 1T). To further address the phagocytic changes and increased Aβ plaque clearance, we first examined the expression of CD68, a marker for activated phagocytic monocytic cells (Fernando et al., 2006; Simpson et al., 2007; Chistiakov et al., 2017), and the lysosome-associated membrane protein, LAMP1, a major endosomal/lysosomal membrane protein thought to be responsible for lysosomal function in phagocytosis (Colombo et al., 2021; Wu et al., 2021), in mouse brains from two-month-old (one month after TAM injection) control, Tsc1^iKO, Trem2^iKO, and Dual^iKO mice. Immunostaining analyses revealed that both CD68 and LAMP1 were expressed at very low levels in the microglia of WT mouse brains (Fig. 6Ab,c, co-stained against Iba1), but increased significantly in Tsc1-deficient microglia, specifically in microglia phagosomes (Bodea et al., 2014; Chistiakov et al., 2017; Fig. 6Bb,c). The increase in CD68 staining is consistent with increased mRNA levels of CD68 in microglia from Tsc1-deficient (Tsc1^cKO) mouse brains, as observed in RNA-seq data (Fig. 1T). In microglia from Trem2^iKO mouse brains, CD68 and LAMP1 levels were similar to those observed in the WT microglia (Fig. 6Cb,c, compare with 6Ab,c). Quantification of the immunofluorescence intensities of CD68 and LAMP1 is presented in Figure 6E. These data reveal that Tsc1 loss-dependent constitutive mTOR activation in microglia causes upregulation of CD68 and LAMP1, which is not observed in the Trem2 knock-out microglia. Interestingly, in Dual^iKO mouse brains, microglia maintained increased levels of CD68 and LAMP1, ∼50–60% of what was observed in Tsc1-deficient microglia, but significantly higher than the ones in WT or Trem2^iKO microglia, suggesting that Trem2 function may be required in the increased lysosomal biogenesis caused by the activation of mTOR in microglia.

Figure 6.

mTOR activation in microglia increases Aβ phagocytosis and clearance. A–D, Representative immunostaining of brain sections from control (+/+), Tsc1^iKO and Trem2^iKO and Dual^iKO mice, using anti-Iba1 (blue, a), anti-LAMP1 (red, b), anti-CD68 (green, c), and merged images showed in d. Scale bar: 20 μm. E, Quantification of immunofluorescence intensities of CD68 and LAMP1 in microglia from two-month-old +/+, Tsc1^iKO, Trem2^iKO, and Dual^iKO mice (n = 3M/3F for each genotype). F–I, LysoTracker live-cell imaging of primary microglia culture derived from +/+, Tsc1^cKO, Trem2^cKO, and Dual^cKO mice (Fx;Cx3cr1-Cre lines) pups, noted as +/+, Tsc1^−/−, Trem2 ^−/−, and Dual ^−/−, respectively (scale bar: 10 μm). J, Quantification of fluorescence intensities of the LysoTracker Green DND-26 (50 nm) indicator in +/+, Tsc1^−/−, Trem2^−/−, and Dual^−/− primary microglia. A total of 40–50 cells were quantified for each genotype, and experiments were repeated three times. K–N, Cultured microglia from F–I, incubated with FAM-Aβ containing medium for 1 h, followed by removal of FAM-Aβ. Cells were then imaged at 0 min, 30 min, 1 h, and 2 h. Representative images of microglia loaded with FAM-Aβ (green) taken at different time points. Cell nuclei were counterstained with DAPI (blue; scale bar: 10 μm). O, Quantification of FAM-Aβ dye intensities per cell in primary +/+, Tsc1^−/−, Trem2^−/−, and Dual^−/− microglia incubated for time intervals as indicated. A total of 40–50 cells were quantified for each genotype, and experiments were repeated three times. P, Quantification of FAM-Aβ dye intensities per cell in primary +/+, Tsc1^−/−, Trem2^−/−, and Dual^−/− microglia after removal of FAM-Aβ dye to monitor fluorescence clearance over indicated times. A total of 40–50 cells were quantified for each genotype/time point, and experiments were repeated three times. Q–T, Images of microglia with FAM-Aβ dye uptake (green) at 0 min or 1 h after removal of dye, treated with vehicle (DMSO) or CQ for 1 h at 37°C CO₂ incubator. Cell nuclei were counterstained with DAPI (blue; scale bar: 10 μm). U, Quantification of FAM-Aβ dye intensities per cell under conditions (Q–T) for primary +/+, Tsc1^−/−, Trem2^−/−, and Dual^−/− microglia cells. For F–I, K–N, and Q–T, microglia cultures were prepared by mixing brain tissues from four to five pups to obtain single-cell suspensions for each genotype. Single-cell suspension was dispensed into four-chamber wells (n = 4) for each time point/condition, and 8–10 random pictures were taken for each well. Comparisons between genotypes were performed by one-way ANOVA with Tukey's post hoc test; * or #p < 0.05, ** or ##p < 0.01, *** or ###p < 0.001. All data are presented as mean ± SEM.

Furthermore, we examined the lysosomal activity by live-staining the primary microglia cells with the LysoTracker indicator (Cell Signaling). Primary microglia cells were obtained from control, Tsc1^cKO, Trem2^cKO, or Dual^cKO mouse brains (all microglia are deficient in either Tsc1, Trem2, or both from birth in these mice, cells are labeled as +/+, Tsc1^−/−, Trem2^−/−, Dual^−/−) and after 30-min incubation with fluorescent LysoTracker indicator, live imaging of all four genotypes of microglia revealed that microglia deficient in Tsc1 showed the greatest fluorescence signals (Fig. 6G), and such signal intensities in cells deficient in Trem2 (Fig. 6H) or both Tsc1/Trem2 (Fig. 6I) only remained at half strength of that in Tsc1^−/− cells; in cells prepared from WT (Fig. 6F) or Trem2^−/− mouse brains (Fig. 6H), the LysoTracker indicator only showed ∼10% intensity compared with the signals in Tsc1^−/− cells (quantified in Fig. 6J). These data demonstrate that lysosomal function is significantly augmented with elevated mTOR signaling and that Trem2 is required for achieving enhanced lysosomal function in mTOR-activated microglia.

Since we observed significantly increased expression levels of CD68 and LAMP1 in Tsc1-deficient microglia, which could result in potentially enhanced phagocytotic activity, we evaluated whether mTOR activation affects the catabolic activity of cultured primary microglia by using FAM-labeled Aβ fluorescent dye uptake assay (Zhao et al., 2018b). We first monitored FAM-Aβ dye uptake at 15 min, 30 min, 1 h, and 2 h which revealed that FAM-Aβ uptake peaked with saturated levels at 2 h and very minimal dye uptake occurred at 15–30 min (quantified in Fig. 6O). Hence, we used the 1 h incubation time to follow the FAM-Aβ dye clearance. After incubation with FAM-Aβ dye for 1 h, microglial cultures were then washed and processed for immunostaining immediately (0 min), after 30 min, 1 h, and 2 h to follow the fluorescence intensity levels of FAM-Aβ (Aβ clearance) overtime across genotypes. Aβ dye uptake was significantly increased in Tsc1-deficient microglia compared with WT (+/+) microglia (Fig. 6La compare to Ka) after 1 h of incubation with FAM-Aβ dye. In either Trem2^−/− (Fig. 6Ma) or Dual^−/− microglia (Fig. 6Na), Aβ dye uptake was modestly reduced compared with WT microglia (Fig. 6Ka) but was significantly reduced compared with Tsc1-deficient microglia (Fig. 6Ma,Na, compare to La). The FAM-Aβ clearance rate was similar in Trem2^−/− microglia as compared with WT cells, indicating that the Aβ degradation after uptake is not significantly affected in Trem2 mutant microglia (Doens and Fernández, 2014; Ulland and Colonna, 2018; Zhao et al., 2018b). More importantly, we observed that the rate of Aβ degradation was significantly increased in the Tsc1-deficient microglia cells compared with the WT microglia, particularly in the first 30 min after removal of the FAM-Aβ dye, indicating a potentially enhanced lysosomal function (quantified in Fig. 6P). While enhanced lysosomal function was also observed in Dual^iKO microglia (Fig. 6D), the degradation of Aβ in Dual^−/− microglia was indistinguishable from +/+ or Trem2^−/− (Fig. 6M,N compare with L), most likely because of low uptake of FAM-Aβ in the absence of Trem2. Together, these data show that mTOR activation in microglia promoted FAM-Aβ uptake as well as its clearance, while loss of Trem2 mostly affected FAM-Aβ uptake but not its rate of clearance.

To further validate that Tsc1-deficient microglia accelerated FAM-Aβ clearance mainly through the lysosome pathway, we treated microglia with lysosomal inhibitor, chloroquine (CQ; Ofengeim et al., 2017; Zhao et al., 2018b). While CQ only partially inhibited Aβ degradation in WT (Fig. 6Q), Trem2^−/− (Fig. 6S), or Dual^−/− (Fig. 6T) microglia, CQ treatment of Tsc1-deficient microglia severely impaired Aβ clearance (Fig. 6Rc; data quantified in Fig. 6U). These results indicate that activation of mTOR signaling pathway in microglia increases Aβ catabolism largely through highly activated lysosomal pathways. Taken together, our data show that activation of mTOR in Tsc1-deficient microglia causes increased uptake of Aβ and accelerated Aβ degradation via enhanced lysosomal function, and that loss of Trem2 reduces Aβ uptake and but does not affect the rate of clearance by lysosomal pathways in microglia.

mTOR inactivation by rapamycin decreases Aβ plaque clearance

Since mTOR activation enhanced Aβ clearance and rapamycin treatment of microglia affected Aβ uptake, we wanted to determine whether treatment of mice with rapamycin will cause an increase or decrease in Aβ clearance in AD mouse models. First, to confirm the status of mTOR activation in microglia, we isolated microglia from six-month-old WT and 5XFAD mice; we did not observe any significant upregulation of mTOR signaling in the 5XFAD mice compared with controls, as revealed by immunoblotting for several mTOR downstream effectors (Fig. 7A). Next, we fed 5XFAD and 5XFAD;Tsc1^iKO mice with a customized diet containing mTOR inhibitor, rapamycin (144 ppm, active rapamycin at 14 mg/kg body weight). As 5XFAD mice start to show Aβ deposition from three months of age (Oakley et al., 2006), we injected TAM at P30 followed by rapamycin treatment at different times so that 5XFAD (Fig. 7B, black) and 5XFAD;Tsc1^iKO (Fig. 7B, red) without rapamycin served as controls or mice received rapamycin after one month of TAM injections for one month beginning at two to three months (Fig. 7B, green group 1), or after three months of TAM injections for two months beginning at four to six months (Fig. 7B, blue group 2), or after two months of TAM injections for three months beginning at three to six months (Fig. 7B, yellow group 3; refer schematic Fig. 7B). All treated or untreated mice of the relevant genotypes were further subjected to immunohistochemical or biochemical analyses. We isolated microglia from these experimental animals, and immunoblots showed that p4EBP1 was significantly inhibited in long-term (two and three months) rapamycin-treated groups, as compared with nontreated or short-term treated (one month) group (Fig. 7C). We conducted immunostaining of coronal brain sections for Aβ plaque burden in the cortex (Fig. 7D) and hippocampus (Fig. 7E) of 5XFAD and 5XFAD;Tsc1^iKO using anti-Aβ antibody and NeuN as a neuronal marker. As quantified in Figure 7F, the 5XFAD;Tsc1^iKO brains without rapamycin showed a substantial reduction in Aβ plaque numbers (red bars) which was not significantly affected with one-month rapamycin treatment (green bars). The two-month (blue bars) and three-month (yellow bars) rapamycin treatment caused a significant increase in Aβ plaque burden compared with 5XFAD;Tsc1^iKO untreated animals indicating that the benefit of mTOR activation for enhanced Aβ clearance is significantly blunted by mTOR inhibition through rapamycin in 5XFAD animals. The Aβ plaques across all genotypes were divided based on the sizes of the plaques (Fig. 7G). Importantly, the Aβ plaques in 5XFAD;Tsc1^iKO mice were generally of smaller sizes than those seen in 5XFAD mice and the number of medium and large plaques was also reduced than in 5XFAD. These changes were more noticeable when considering the plaque size distribution on rapamycin treatment for two and three months which remarkably reduced the number of smaller size plaques (<10 μm) and accordingly increased the number of larger plaques (>40 μm; Fig. 7G). Next, we conducted immunostaining of the whole-brain sections from relevant genotypes using anti-Aβ antibody and MX04 followed by confocal microscopy (Fig. 7H,I). We analyzed Aβ plaque conformational changes which is a reflection of the types of Aβ plaque aggregations in the cortex regions of all 5XFAD;Tsc1^iKO mice that were treated with rapamycin. Consistently, we observed that in 5XFAD;Tsc1^iKO mice, Aβ plaques became more inert and less filamentous, compared with 5XFAD mice (Fig. 7J). The one-month treatment of rapamycin did not significantly change the confirmational profiles of Aβ plaques. However, after longer periods of rapamycin treatment for two and three months, plaques became larger, less inert, more compact and more filamentous, as compared with 5XFAD;Tsc1^iKO mice (Fig. 7I; quantified in Fig. 7J), indicating that the long-term rapamycin treatment of 5XFAD;Tsc1^iKO mice augmented the Aβ phenotypes, similar to those observed in the 5XFAD mice further underscoring that rapamycin treatment alters Aβ plaque aggregation and clearance by microglia. To further confirm these results, we employed Aβ ELISA assay for more quantitative measurement of Aβ levels, which again showed a strong reduction of Aβ levels in 5XFAD;Tsc1^iKO mice without rapamycin or one-month rapamycin treatment compared with 5XFAD; however, rapamycin treatment for two and three months significantly increased the Aβ levels in both the 5XFAD (Fig. 7K) and 5XFAD;Tsc1^iKO mice (Fig. 7L) indicating that rapamycin treatment increased the plaque burden. Together, these data show that prolonged in vivo mTOR inhibition by rapamycin causes an increase in Aβ plaque accumulation and exacerbates Aβ pathology in 5XFAD mice.

Figure 7.

Rapamycin treatment increases Aβ plaque deposition and exacerbates Aβ-related AD pathologies. A, Representative immunoblots showing the expression of mTOR signaling molecules, including pS6K, p4EBP1, p-mTOR, p-AKT, in isolated microglia from WT (n = 1 M/2F) and 5XFAD mouse brain (n = 2 M/1F). APP protein expression was examined (with 6e10 antibody) in brain lysates from the same cohort of WT and 5XFAD mice. Each lane represents lysates from individual animals. B, Experimental design and timeline showing color-coded genotypes with TAM injection at one month followed by rapamycin treatments for one, two, or three months and then phenotypic analysis at six months. C, Representative immunoblots showing the expression of p4EBP1 in isolated microglia from 5XFAD, and 5XFAD;Tsc1^iKO mice with various treatment time of rapamycin. D, E, Immunostaining of cortical (D) and hippocampal areas (E) using anti-Aβ (green) and anti-NeuN (red) from 5XFAD (no rapamycin, a), 5XFAD;Tsc1^iKO (no rapamycin, b), and 5XFAD;Tsc1^iKO with one month (c), two months (d), or three months (e) of rapamycin treatment. Scale bar: 50 μm. F, Number of plaques quantified for cortical and hippocampal areas in genotypes for D, E. G, Aβ plaque size distribution quantified from respective genotypes with or without rapamycin grouped by the size of plaques at <10, 10–40, and >40 μm. H, I, Immunostaining of brain sections from 5XFAD and 5XFAD;Tsc1^iKO with the various periods of rapamycin treatment with anti-Aβ antibody (red) and MX04 dye (green), respectively (scale bar: 25 μm). J, Quantification of plaque types as inert, compact, or filamentous in 5XFAD and 5XFAD;Tsc1^iKO without or with various periods of rapamycin treatment. K, L, Protein levels of Aβ₄₂ in insoluble, guanidine-HCl-extracted fraction from cortex and hippocampus of 5XFAD (K) or 5XFAD;Tsc1^iKO (L) without or with various periods of rapamycin treatment (normalized to total protein levels). For B–L, 5XFAD mice without Rapa (n = 3M/4F), 5XFAD + 1mo Rapa (n = 3M/2F), 5XFAD + 2mo Rapa (n = 3M/3F), 5XFAD + 3mo Rapa (n = 2 M/3F); 5XFAD;Tsc1^iKO mice without Rapa (n = 3M/4F), 5XFAD;Tsc1^iKO + one month Rapa (n = 3M/3F), 5XFAD;Tsc1^iKO + two months Rapa (n = 4 M/4F), 5XFAD;Tsc1^iKO + three months Rapa (n = 4 M/5F). Comparisons between genotypes/rapamycin treatment conditions were performed by one-way ANOVA with Tukey's post hoc test; *p < 0.05, **p < 0.01, ***p < 0.001. All data are presented as mean ± SEM.

mTOR inhibition by rapamycin reprograms microglia and downregulates Trem2

As rapamycin treatment caused significant changes in microglial morphologies and Aβ plaque size distribution, we wanted to determine whether rapamycin caused any reprogramming in microglial in 5XFAD;Tsc1^iKO mice, as microglia are known to respond to both intrinsic and extrinsic signals (Hansen et al., 2018). As shown in Figure 8A–E, immunostaining of brain sections of relevant genotypes using Iba1, Aβ, and F4/80 antibodies revealed highly activated microglia in the 5XFAD;Tsc1^iKO mice as shown by enhanced expression of F4/80 (Fig. 8Bc compare to control 5XFAD, Ac). Interestingly, the one-month treatment of rapamycin did not affect the overall number of microglia (Fig. 8F) and change the programming status as reflected by F4/80 immunostaining of microglia in the 5XFAD;Tsc1^iKO mice (Fig. 8Cc); however, the two-month (Fig. 8Dc) and three-month (Fig. 8Ec) treatment of rapamycin greatly reduced F4/80 immunostaining showing a change in microglial programming. The long-term rapamycin treatment significantly reduced the microglia numbers (Fig. 8F), and also altered the overall profiles of microglia in the 5XFAD;Tsc1^iKO mice, showing a decrease in cell volume (Fig. 8G), an increase in the length of microglial processes (Fig. 8H) with a marked increase in process branch points per microglia (Fig. 8I), and significantly reduced immunofluorescence intensity of F4/80 for each microglia (quantified in Fig. 8J). These data show that prolonged mTOR inhibition by rapamycin treatment alters microglial morphology and properties.

Figure 8.

mTOR inhibition by rapamycin alters the activation status of microglia and downregulates Trem2. A–E, Representative immunostaining pictures of brain sections for Iba1 (green), Aβ (red), and F4/80 (blue) from 5XFAD (without rapamycin, A) and 5XFAD;Tsc1^iKO without (B) or one month (C), two months (D), and three months (E) of rapamycin treatment (scale bar: 20 μm). F, Quantification of Iba1+ cell numbers in cortex and hippocampus regions in mice shown in A–E color-coded 5XFAD (without rapamycin, black) and 5XFAD;Tsc1^iKO without (red) or one month (green), two months (blue), and three months (yellow) of rapamycin treatment. G, Quantification of Iba1+ cell volume in genotypes shown in A–E color-coded 5XFAD (without rapamycin, black) and 5XFAD;Tsc1^iKO without (red) or one month (green), two months (blue), and three months (yellow) of rapamycin treatment. H, I, Quantification of the morphologic features showing the length of microglial processes (G) and the number of branch points per microglia (H) in genotypes as in F. J, Percentage of F-480⁺/Iba1⁺ cells in genotypes as in F. K, Quantification of microglia associated with plaques, grouped by the sizes of plaques (<10 μm, or between 10 and 40 μm, or >40 μm) in genotypes as in F. L–P, Immunostaining of brain sections for Trem2 (green) and Iba1 (red) from 5XFAD (without rapamycin, L) and 5XFAD;Tsc1^iKO without (M) or one month (N), two months (O), and three months (P) of rapamycin treatment. Scale bar: 20 μm. Q, Quantification of Trem2 immunofluorescence against Iba1 in microglia from 5XFAD (without rapamycin) and 5XFAD;Tsc1^iKO without or one month, two months, and three months of rapamycin treatment. R, Representative immunoblots of protein levels of Trem2 in microglia from brain lysates of 5XFAD (without rapamycin) and 5XFAD;Tsc1^iKO without or one month, two months, and three months of rapamycin treatment. β-Actin was used as the protein loading control. All mice were from the same cohort experiments described in Figure 7. For quantification, 40–50 microglia for each animal were quantified for the morphology analysis and protein fluorescence intensities. For R, microglia were isolated by CD11b+ beads from the same cohort of animals. Comparisons between groups were performed by one-way ANOVA with Tukey's post hoc test; * or #p < 0.05, ** or ##p < 0.01, *** or ###p < 0.001. All data are presented as mean ± SEM.

Next, we conducted a more detailed analysis of whether the number of microglia associated with the Aβ plaques was altered by rapamycin treatment, as previous studies have shown the specific association of microglia with Aβ plaques in disease states (Keren-Shaul et al., 2017). As shown in Figure 8K, more extended periods of rapamycin treatment significantly reduced the number of microglia that were associated with Aβ plaques across the plaque sizes. Interestingly, the number of microglia associated with larger plaques (>40 μm) showed no difference between 5XFAD mice and 5XFAD;Tsc1^iKO mice, but the number was dramatically reduced in mice treated with rapamycin for two and three months, suggesting that prolonged treatment with rapamycin significantly impacted the overall microglia association with Aβ plaques, and accordingly reduced their ability to clear Aβ plaques in the 5XFAD mice.

Since the loss of Tsc1 in microglia led to upregulation of Trem2 expression as revealed by both mRNA and protein analyses (Fig. 2A–H), we wanted to determine whether rapamycin treatment of 5XFAD;Tsc1^iKO mice will alter the in vivo Trem2 expression in microglia. We conducted immunostaining of brain sections using anti-Trem2 and Iba1 antibodies. As shown in Figure 8L–P (quantified in Fig. 8Q), microglial Trem2 levels in 5XFAD;Tsc1^iKO were significantly reduced by the two-month (Fig. 8O) and three-month (Fig. 8P) treatment of rapamycin, as compared with one-month (Fig. 8N) rapamycin treatment. The two and three months of rapamycin treatment dramatically reduced the levels of Trem2 to that seen in the untreated 5XFAD mice (Fig. 8L). These data (all quantified in Fig. 8Q) indicate that mTOR signaling regulates microglial Trem2 expression. Immunoblotting of protein lysates prepared from purified microglia from 5XFAD and 5XFAD;Tsc1^iKO mice further confirmed a strong reduction in Trem2 protein expression on prolonged rapamycin treatment of two and three months (Fig. 8R). Together, these data demonstrate that mTOR inhibition by rapamycin fundamentally changes the programming status of microglia in 5XFAD mice by suppressing Trem2 expression and thereby diminishing the ability of microglia to robustly uptake and clear Aβ aggregates resulting in the elevated Aβ plaque accumulation and enhanced AD pathologies.

Discussion

In the current study, we examined the impact of elevated mTOR signaling in microglia in the 5XFAD mice, a well-established AD disease model. mTOR-activated microglia adopted a hyper-active phenotype with an amoeboid morphology, augmented phagocytosis and clearance of Aβ. These phenotypic changes significantly altered the pathologies in 5XFAD mice, leading to a pronounced reduction in Aβ deposition and recovery of memory retention. We uncovered a novel regulatory mechanism in which Trem2 expression was augmented in mTOR-activated microglia, and inhibition of mTOR signaling by rapamycin exacerbated Aβ-related AD pathologies.

mTOR activation augments microglia phagocytosis and enhances Aβ clearance

mTOR signaling pathways have been shown to be vital in many physiological and pathologic processes. Here, we observed dramatic morphologic alternations when mTOR was selectively activated in microglia, with remarkably enhanced expression of lysosomal genes, C1q complement components, and anti-inflammatory factors. Such reprogramming is consistent with previous reports using noninducible Tsc1 ablation in microglia (Zhao et al., 2018a). In some studies, mTOR signaling has been reported to act as a negative regulator of lysosomal biogenesis genes, by maintaining TFEB cytosolic localization through phosphorylation (Martina et al., 2012; Roczniak-Ferguson et al., 2012; Napolitano et al., 2018). However, this model of mTOR/TFEB regulation cannot explain the sustained lysosomal function under persistent mTORC1 signaling during some physiological states, such as starvation recovery and physical exercise. Strikingly, our study and other studies (Peña-Llopis et al., 2011; Betschinger et al., 2013; Kawano et al., 2015) consistently showed constitutively activated mTORC1 signaling maintained a strong lysosomal biogenesis, most likely via AKT feedback loop through activating EGFR and HER2 (Asrani et al., 2019). In addition, mTOR activation downregulates genes specific to homeostatic microglia including purinergic receptors P2ry12 and Tmem119, which indicate a loss of induced motility (Haynes et al., 2006; van Wageningen et al., 2019). Phenotypically, mTOR-activated microglia shared many similar features with DAM, which display a protective role with upregulated lysosomal, phagocytic and lipid metabolism pathways, as well as downregulated homeostatic genes (Keren-Shaul et al., 2017; Deczkowska et al., 2018) suggesting that cell-autonomous mTOR activation in microglia changes microglial properties.

More importantly, mTOR activation in microglia significantly increased the mRNA levels of AD-risk genes such as Trem2, CD33, Ms4a, Plcg2 genes. The exact role of mTOR signaling in AD models is still being debated (McQuade and Blurton-Jones, 2019) and it has been hypothesized that acute Aβ presence directly stimulates microglia inflammation; while, chronic stimulation of Aβ re-programs microglia into a tolerant state with defects in cellular metabolism and reduced responses to inflammatory stimuli (Baik et al., 2019). As the constitutive activation of mTOR increases lysosomal biogenesis through the Akt pathway, we would also expect to see the inhibitory effects of the rapamycin treatment on lysosomal function. Therefore, it is not surprising that Aβ clearance by microglia seems to be weakened by the inactivation of the mTOR pathway in chronic Aβ-stimulated cellular models or AD mouse models, as uncovered in the current study and also by others (Lafay-Chebassier et al., 2005; Son et al., 2012; Li et al., 2013; Bai et al., 2020). However, the mechanism of how Aβ inhibits mTOR signaling in the late stage of AD is still unclear. A recent proteomics study provided further evidence that in 5XFAD mouse brains, phospho-S6K and phospho-4EBP1 are reduced or not altered at the whole-protein levels, indicating the mTOR activation is either unchanged or compromised (Bai et al., 2020). As presented here, induced loss of Tsc1 in microglia provided continued mTOR activation throughout the experimental time course (six to seven months of age) and no animal deaths were observed within this timeframe suggesting that postnatal activation of mTOR does not lead to lethality. mTOR signaling activation augmented microglial priming at the earlier stage of Aβ stimulation, and subsequently sustained the phagocytic ability of microglia despite chronic Aβ stimulation. Our data reveal that loss of Tsc1 in microglia significantly increased Aβ uptake as a consequence of Trem2 upregulation, a known Aβ receptor (Udeochu et al., 2018; Zhao et al., 2018b) and also accelerated Aβ degradation in cultured microglia cells, through upregulation of lysosomal functions (Feng et al., 2020; Meilandt et al., 2020). Thus, mTOR activation in microglia not only reduced the Aβ plaque burden, but also significantly improved cognitive deficits in the water maze test. Furthermore, mTOR-activated microglia exhibited an augmented anti-inflammatory state and were more clustered at or surrounding the Aβ plaque areas in the 5XFAD;Tsc1^iKO mice causing a reduction in both the number and size of the Aβ plaques. Together, our data highlight the beneficial impact of mTOR activation in microglia in the 5XFAD AD mouse model.

Trem2 is downstream of mTOR signaling in microglia

The activation of mTOR in microglia significantly elevated the expression of several AD risk genes, among which most notable is Trem2. Trem2 has been extensively studied in various AD disease models and loss of Trem2 or overexpression of mutant Trem2^R47H in AD mouse model has been shown to be detrimental for AD pathogenesis (Jay et al., 2015; Xiang et al., 2016; Yuan et al., 2016; Ulland et al., 2017). On the other hand, increasing the levels of endogenous expression of functional Trem2 (CYD Lee et al., 2018a), or soluble fraction of Trem2 in AD mouse models greatly reduced the amyloid burden and led to the improvement in cognitive defects (Zhong et al., 2017). In the current study, selective loss of Trem2 in microglia from Tsc1^iKO mouse brains caused microglia to become “desensitized,” and led to significant morphologic changes, including reduced cell volumes, enhanced process length, and reduced expression levels of F4/80. Double knock-out of Trem2 and Tsc1 in microglia (Dual^iKO) abolished the beneficial effects of Tsc1^iKO microglia in 5XFAD mice, with increased Aβ burden and further deteriorated Aβ pathology underscoring the critical role of Trem2 in AD pathogenesis. Thus, our results are consistent with previous studies that the dosage of Trem2 is a determinable factor for AD progression and pathogenesis, and further highlight that the mTOR signaling pathway could be exploited for therapeutic benefits in Aβ-associated AD pathologies.

Rapamycin desensitizes primed microglia and accelerates Aβ-associated AD pathology

Rapamycin, a highly specific allosteric partial inhibitor for mTOR signaling pathway, has been extensively studied in various disease models and suggested to be beneficial in various pathologies (Lin et al., 2013, 2017; Talboom et al., 2015). In the hAPP-J20 AD mouse model, rapamycin has been shown to reduce Aβ plaque deposition and Tau protein aggregation when it was applied to the mice before the disease onset (Spilman et al., 2010; Van Skike et al., 2021). The J20 mouse model overexpresses human APP carrying two mutations (the Swedish and Indiana mutations) linked to familial AD (FAD) by the PDGF-β promoter with increased levels of Aβ than mice expressing equivalent amounts of the WT human APP (Mucke et al., 2000). The 5XFAD transgenic mice use the Thy1-promoter to express human amyloid β (A4) precursor protein 695 (APP) with three FAD-related mutations (the Swedish K670N/M671L, Florida I716V, and London V717I) along with human Presenilin 1 (PS1) harboring two mutations (M146L and L286V). These mouse models may have unique differences that need to be further characterized in terms of the effect of mTOR inhibition and whether in the J20 mouse model rapamycin treatment affects Trem2 expression. It is well established that rapamycin inhibits mTOR signaling, and in turn transfers proteostasis to catabolism, mostly through the lysosomal system (Laplante and Sabatini, 2012). In AD patients or mouse models, lysosomal system cargos (mostly in microglia) remarkably accumulate around amyloid plaques and fail to clear Aβ plaques because of their toxicity (McBrayer and Nixon, 2013). Given the fact that in human subjects, plaque deposition begins decades before a precise clinical diagnosis of dementia (Roberts et al., 2017), it is unlikely an AD patient would benefit from a therapy that interferes with a poorly functional lysosomal system. Here, for two to three months of rapamycin treatment starting in three- to four-month-old 5XFAD mice, we observed an increase in Aβ plaque deposition, and exacerbated Aβ-associated pathologies. We also observed changes in both the morphology and the priming status of microglia in 5XFAD;Tsc1^iKO mice on rapamycin treatment. These observations lead us to conclude that rapamycin treatment of the 5XFAD model in fact accelerated Aβ pathologies through downregulation of Trem2.

Collectively, through in vivo and in vitro analyses, we provide a mechanistic rationale for potential therapeutic consideration for Aβ-associated AD pathologies that may impinge on mTOR activation rather than inhibition. Our data highlight that mTOR activation reprograms microglia, and changes their priming status through upregulation of Trem2, thereby reducing the Aβ plaque burden and causing improvements in cognitive deficits in the 5XFAD AD model. We provided in vivo evidence that feeding mice with a rapamycin diet posed additional risks to the AD pathogenesis by further desensitizing microglia and increasing Aβ plaque deposition. Our results are consistent with recent suggestions (Carosi and Sargeant, 2019) that rapamycin may have detrimental effects in the treatment of AD considering the disrupted mTOR signaling, and treating dementia patients with rapamycin may cause further damage to microglia and promote Aβ plaque accumulation leading to enhanced Aβ-associated AD pathologies.