Effects of microglial depletion and TREM2 deficiency on Aβ plaque burden and neuritic plaque tau pathology in 5XFAD mice

The key pathological hallmarks of the AD brain are extracellular senile plaques comprising Aβ peptides and intracellular inclusions of misfolded tau protein, which normally binds microtubules (MTs) [9, 17]. Tau pathology can be found in Aβ plaque-associated neuronal processes as NP tau, within the soma as neurofibrillary tangles or in dendrites as neuropil threads. Aβ plaques develop first in the neocortex [5], whereas the initial distribution of tau pathology appears to be restricted to distinct brain areas including the locus coeruleus and trans-entorhinal cortex [3, 5] before progressing to the hippocampus and later to higher cortical regions. The “amyloid cascade” hypothesis [18, 41] suggests that cortical Aβ pathology initiates still poorly understood events that ultimately culminate in the cortical spread of tau pathology. Tau inclusions have been strongly implicated in the neurodegeneration of AD, as multiple immunohistochemical studies reveal a significant correlation between the degree of tau pathology and cognitive deficits in AD patients [1, 4, 48], a relationship that is not observed with Aβ plaque burden. Moreover, recent tau PET imaging studies further confirm a tight linkage between tau pathological burden and AD patient cognitive status [7, 42], as well as brain atrophy [45]. Thus, a key research objective is identifying mechanisms by which the formation of Aβ plaques promote the spread of neocortical tau pathology that is linked to AD dementia.

Plaque-associated neuronal processes may represent a key node in the amyloid cascade hypothesis, as this is a site where NP tau pathology forms in proximity to Aβ deposits. A model of coincident Aβ plaque and NP tau has recently been described [19] in which plaque-forming Tg mice (e.g., 5XFAD [31] or APP-knock-in [34]) develop widespread NP tau in the absence of tau overexpression after the intracerebral injection of a small amount of AD-tau. Endogenous mouse tau that accumulates within plaque-associated dystrophic processes [19], likely as a result of local disruption of MTs [11, 33, 50], is converted to insoluble NP tau aggregates that are recognized by multiple phospho-tau antibodies as well as antibodies to misfolded tau upon seeding by internalized AD-tau. Over time, this NP tau can lead to the subsequent formation of neurofibrillary tangles and neuropil thread tau pathology [19]. This unique model of combined Aβ and tau pathology provides a platform to investigate how Aβ pathology might affect the onset and spread of tau pathology.

When considering how Aβ plaques and/or oligomers might alter nearby neuronal processes, it is also important to consider the potential effects of plaque-associated microglia as found in both AD brain and in mouse models [6, 39]. A role of microglia in AD pathobiology is supported by a large body of data, including genome-wide association studies that have identified several microglial gene variants as risk factors for AD [29], including TREM2 variants [15, 23] that appear to affect microglial function [44]. Many studies have examined the potential influence of microglia in mouse models of Aβ plaque formation. Microglia can be quantitatively eliminated with CSF1R inhibitors [10], and whereas Aβ plaque burden was unchanged in older mice after CSF1R inhibitor treatment, there was a rescue of dendritic spines, improvement of cognitive function and reductions of neuron loss [32, 40]. More recent studies [37, 38] have examined the effect of depleting microglia from younger Aβ plaque-forming mice, where Aβ plaque burden was diminished after CSFR1 inhibitor treatment. Thus, microglia may play a role in the initial deposition of Aβ plaques, whereas they do not appear to affect later plaque development but nonetheless contribute to neuronal dysfunction.

As noted, TREM2 variants confer increased risk of AD with the R47H variant exerting the largest effect through a partial TREM2 loss-of-function with reduced binding of natural TREM2 ligands [2, 46, 51]. The R47H TREM2 substitution also causes a significant reduction of TREM2 protein expression in mice, although this is due to generation of a mis-spliced TREM2 transcript that is not formed in humans with the R47H allele [49]. There have been numerous studies [14, 22, 30, 46, 47, 52] investigating the consequences of TREM2 loss-of-function via genetic knockout (KO) in mouse models with Aβ pathology. Although there have been differences in the findings among these studies, particularly with regard to the effect of TREM2 KO on Aβ plaques [14, 22, 30, 46], there is an emerging consensus that TREM2 deficiency results in reduced microglial recruitment to and consequent containment of plaques, resulting in increased Aβ-mediated damage to nearby neuronal processes [13, 24]. Although TREM2 KO studies are informative, they may not fully model the heterozygous TREM2 variants linked to AD. There have been fewer studies examining plaque-forming mice harboring TREM2^+/− microglia, but it appears they exhibit some reduction in the number of plaque-associated microglia [43, 52] although less than that observed in TREM2^−/− mice. There may still be changes in plaque morphology in mice with TREM2 haploinsufficiency that leads to increased plaque-associated neuritic dystrophy [52], although a recent report [27] indicates that heterozygous TREM2 microglia show normal plaque interaction without the evidence of increased neuritic damage observed in TREM2^−/− plaque-bearing mice.

The existing literature suggests there are differences between the effects of microglial depletion and TREM2 knockout on early plaque deposition. In particular, whereas early depletion of microglia in 5XFAD mice appears to cause a lowering of plaque burden[37, 38] and plaque volume [38], TREM2 deficiency seems to have a more variable effect on plaque deposition that may depend on the mouse model. Moreover, the improved neuronal outcomes observed after microglial depletion in older plaque-bearing mice, where no changes in plaque burden were observed [32, 40], suggests that reducing microglial interaction with Aβ plaques can be beneficial. This contrasts with observations from TREM2 knockout studies, where reduced microglial plaque association results in enhanced neuritic damage.

Here, we have examined the consequences of both microglial depletion and TREM2 deficiency on the formation of NP tau pathology, utilizing our previously described model of concurrent Aβ plaque and NP tau pathology after intracerebral seeding with AD brain-derived tau [19]. Although a recent report has revealed that TREM2 KO, or expression of human R47H TREM2 on a mouse TREM2 null background, results in exacerbation of NP tau pathology after AD-tau injection of APP/PS1 mice [28], we extend these studies by comparing the effects of both TREM2^−/− and TREM2^−/+ microglia in AD-tau-injected 5XFAD mice. In addition, we also assessed the effect of microglial depletion via CSFR1 inhibitor treatment. Depletion of microglia after CSFR1 inhibitor treatment in young 5XFAD mice resulted in decreased Aβ plaque burden and a corresponding reduction of NP tau pathology, including a trend toward reduced neuritic dystrophy in plaques that remained after microglial depletion that likely related to smaller plaque size. Moreover, although 5XFAD × TREM2^+/− mice had significantly more plaque-associated microglia than 5XFAD × TREM2^−/− mice, the amount of NP tau pathology was equal to or greater in 5XFAD × TREM2^+/− than 5XFAD × TREM2^−/− mice. Gene expression analyses revealed that 5XFAD × TREM2^+/− mice had DAM gene expression levels that were generally intermediate between 5XFAD × TREM2^+/+ and 5XFAD × TREM2^−/− mice, and microarray expression profiling showed differences in gene ontology pathways, including those linked to microglial function, between 5XFAD × TREM2^+/− and 5XFAD × TREM2^−/− mice. Our data suggest that the amount of NP tau pathology observed in AD-tau-injected 5XFAD mice is not entirely predicted by the extent of microglial interaction with plaques, nor is development of appreciable tau pathology dependent on the complete loss of a DAM phenotype. The observed differences between TREM2 heterozygous and KO mice suggest that the former may better model the single copy TREM2 risk alleles in AD.

Mouse Breeding: 5XFAD mice were kindly provided by Dr. Robert Vassar and were bred as previously described [31]. The generation of 5XFAD mice with differing TREM2 genotypes is summarized in Additional file 1: Fig. S1, resulting in 5XFAD^+/− × TREM2^+/+, 5XFAD^+/− × TREM2^+/− and 5XFAD^+/− × TREM2^−/− mice on identical genetic backgrounds. The genotype of all mice was confirmed by PCR, with verification of the mutated APP and PS1 transgenes of 5XFAD mice as previously described [19], and TREM2 genotyping performed utilizing the following primers:

TREM2 common forward primer: TCA GGG AGT CAG TCA TTA ACC A.

TREM2 wild-type reverse primer: AGT GCT TCA AGG CGT CAT AAG T.

TREM2 KO reverse primer: CAA TAA GAC CTG GCA CAA GGA.

Expected TREM2 amplification products: KO = 396 bp; Wild type = 254 bp.

Stereotaxic Injection of AD-Tau or Control Brain Extract: AD brain-derived pathological tau and control brain extracts were prepared as previously described [16]. 5XFAD mice were injected with a total of 2 µg of AD-tau (typically 10–20% of total protein) or an equal volume of control brain extract in the hippocampus and overlying cortex of one hemisphere (2.5 µl of 0.4 mg/mg of AD-tau/site), at coordinates bregma: − 2.5 mm; lateral: 2 mm; depth: − 2.4 mm and − 1.4 mm from the skull, as previously described [16]. In one study, 5XFAD mice that had received either control or PLX3397-containing chow beginning at 1.5 months of age received stereotaxic injections at 3 months of age. There were 7 mice in each treatment group injected with AD-tau (Control chow, 4 males and 3 females; PLX3397 chow, 3 males and 4 females). In the control extract-injected 5XFAD mice, there were 6 mice in the PLX3397 and control treatment groups (3 males and 3 females). In another study, a total of 8 5XFAD mice (4 males and 4 females) from each TREM2 genotype received stereotaxic injections of AD-tau at 4 months of age. In both studies, mice survived an additional 3 months after intracerebral injections before sacrifice and post-mortem analyses.

PLX3397 Dosing of 5XFAD mice: PLX3397 (1Click Chemistry, Inc.) was formulated in chow at 1000 mg/kg or 290 mg/kg (Open Standard Diet; Research Diets, Inc.). Control chow was of the same composition but without added PLX3397. As summarized in Additional file 1: Fig. S2, groups of 1.5 month old 5XFAD mice received either control or 1000 mg/kg of PLX3397-containing chow for 1 week, after which the mice receiving PLX3397 were switched to chow containing 290 mg/kg of PLX3397. After 5 additional weeks on PLX3397 (290 mg/kg) or control chow, all mice received intracerebral injections of either AD-tau or control brain extract and were then maintained on control or PLX3397 (290 mg/kg) chow for an additional 3 months.

Preparation of Brain Sections: All study mice were sacrificed and perfused with phosphate-buffered saline (PBS) using protocols approved by the University of Pennsylvania IACUC. Brains were collected and fixed in formalin for 24 h at 4 °C, followed by transfer to 20% sucrose solution at 4 °C for 48 h. Coronal slices (30 mm) were obtained by using a brain mount, and these were cassetted and frozen in dry ice. A small hole was punctured on the left side of the midbrain to ensure proper identification of the hemisphere that had received AD-tau or control brain extract injections. Cassetted brain slices were stored at − 80 °C until they were sectioned (40 µm) and placed on slides, using a Thermo Shandon cryotome.

Immunofluorescence (IF): Frozen sections were stained with selected antibodies and fluorescently-labeled secondary antibodies to visualize microglia, phosphorylated tau, Aβ plaques, and APP positive dystrophic processes. Sections from 3 or 4 bregma levels from each study mouse were stained for analyses and utilized for quantification unless there was a technical flaw with a section, with a minimum of two evaluable bregma quantified from each mouse. The following antibodies and dyes were utilized: Aβ: NAB228 (1:20 K) monoclonal antibody (mouse anti-Aβ1-11; in-house); APP: 22C11 (1:1 K) mouse monoclonal antibody (APP N-terminus; Millipore, Inc.); microglia: Iba1 (1:1 K) rabbit monoclonal antibody (FujiFilm Wako Pure Chemical Industries, Ltd.); pTau: AT8 (1:1 K) mouse monoclonal antibody (tau pSer202/pThr205; Thermo-Fisher) and AT180 (0.33 µg/ml) mouse monoclonal antibody (tau pThr231; Thermo-Fisher); and misfolded tau: Alz50 (1:1000) monoclonal antibody (kind gift from Dr. Peter Davies). Sections were placed in wells of a 12-well plate containing 1 ml of Dulbecco’s phosphate-buffered saline (PBS), pH 7.4, with a maximum of 12 sections per well. Sections were washed (3 × 10 min) in PBS. In sections to be stained with 22C11 antibody, antigen retrieval was performed in citrate buffer (1:100 citric acid antigen unmasking solution; Vector Laboratories) for 1 h at 60 °C. All sections were blocked for 1 h at room temperature in 3% fetal bovine serum, 3% bovine serum albumin (w/v), 0.1% Triton X-100 in PBS solution. Primary antibodies were prepared in blocking buffer without Triton X-100 and incubated overnight at 4 °C, with the exception of 22C11 antibody which was incubated overnight at room temperature. Bound primary antibodies were detected with one of the following secondary antibodies: goat anti-mouse IgG2a conjugated with Alexa Fluor 647 (Thermo-Fisher); goat anti-mouse IgG1 conjugated with Alexa Fluor 488 (Thermo-Fisher); or goat anti-rabbit IgG conjugated with Alexa Fluor 568 (Millipore). A code was generated for all study mice to allow for masked processing of sections. Sections were imaged with a Nikon eclipse Ni microscope using either a 4X,10X, or 20X microscopic objective.

Image Quantification: Image J-Fiji software was used for all IF analyses in this study. Iba1-, AT8- and NAB228-stained sections within the cortex and subiculum were quantified by manually annotating each region of interest in 4X images after staining. Images for these analyses were obtained from bregma levels that captured the majority of tau pathology after AD-tau injection (bregma – 2.40, – 2.18, – 3.08, and – 3.52 for the PLX3397 study, and – 2.52, – 3.08, and – 2.70 for the TREM2 study). AT8 images were utilized from bregma – 2.40 in the PLX3397 study to avoid potential exogenous signal from AD-tau at the site of AD-tau injection. Images containing the injection site were not used for quantification. Once regions of interest were annotated, the image was converted to 8-bit and the remaining area outside of the annotated region was cleared. The image was converted to a grey scale and a threshold was applied to eliminate background signal, and integrated density (area × mean intensity) values were obtained. For visualization, the thresholded image was applied back to the original image to display the quantified IF signals (e.g., Additional file 1: Figs. S3 and S4). For NAB228 quantification, a particle size setting of 40-infinity was applied to exclude objects < 40 µM². A particle size analysis (setting of 150-infinity) was utilized to quantify cortical plaque-associated APP at bregma levels – 2.70, – 2.46 and – 3.08 to avoid confounding signal from APP-positive neuronal cell bodies (see Fig. 5a). To closely examine cortical APP-positive dystrophic processes associated with plaques, higher power 20X images were obtained. Any APP-positive staining that was co-localized with or touching NAB228-positive plaques was annotated (see Fig. 6a). The manual selection was transferred to the 22C11 fluorescence channel, and the integrated density (area × mean density) values were obtained, which were normalized to the number of annotated plaques to give the APP integrated signal per plaque. In addition, the Aβ-positive plaque area was separately annotated in the NAB228 channel to allow calculation of the integrated Aβ signal. Similarly, to examine the plaque-associated cortical microglia, Iba1 staining that was co-localized with or touching NAB228-positive plaques was annotated in 20X images (Fig. 2b). The manual selection was transferred to the Iba1 fluorescence channel, and the Iba1 area values were obtained, which was normalized to the number of annotated plaques to give the relative Iba1 area per plaque.

IF Statistics: All IF quantifications were normalized to the appropriate control group (either control chow in the PLX3397 study or 5XFAD × TREM2^+/+ in the TREM2 genotype study). As there are known differences between female and male 5XFAD mice (e.g., Aβ production and plaque load [31, 50]), IF values were normalized to the mean of the control group by sex. Statistical outliers within a treatment group (by sex) were identified by a two-sided Grubb’s or Dixon’s test and excluded from the analysis if p < 0.05. Statistical analyses with two study groups were done by a two-tailed unpaired T-test, and analyses of > 2 groups were done by one-way ANOVA with Tukey’s post-hoc comparisons. For the qPCR results, statistical analysis was done with a two-way ANOVA with Tukey's multiple comparisons test. All analyses were performed in GraphPad Prism. Error bars on all graphs represent the standard error of the mean.

RNA extraction: All study mice were sacrificed and perfused with PBS using protocols approved by the University of Pennsylvania IACUC. Brains were dissected, and the cortex and the hippocampus (with subiculum) from both the left and right hemisphere were collected, flash-frozen in liquid nitrogen and stored at − 80 °C until further use. Brains were obtained from 4 mice of genotype 5XFAD^+/− × TREM2^+/+ (3 males and 1 female), 6 mice of genotype 5XFAD^+/− × TREM2^+/− (3 males and 3 females) and 6 mice of genotype 5XFAD^+/− × TREM2^−/− (3 males and 3 females). Tissues were transferred into Lysing Matrix D tubes (MP Biomedicals) and homogenized in 750 µl TRIzol™ reagent (Ambion) by using the MP Fast Prep24 homogenizer (MP Biomedicals) with 2 × 30 s at a speed of 6 m/s. Chloroform (150 µl) was added to the Lysing Matrix D tubes and thoroughly mixed with the homogenized samples. After centrifugation (15 min, 14,000 g, at 4 °C), 300 µl of the aqueous phase of each sample was transferred to a deep well plate and mixed with 300 µl of 70% ethanol. 600 µl of this mixed solution was transferred onto a RNeasy 96-well plate provided with the RNeasy 96-kit (Qiagen). The RNA was extracted according to the manufacturer’s instructions. RNA was eluted into 80 µl of RNase free water by means of two centrifugation steps (2 × 4 min, 6000 rpm at room temperature) and stored at − 80 °C until further use.

cDNA conversion and Real-time qPCR of DAM gene mRNA: The RNA concentration of the cortex and hippocampus samples from 5XFAD mice with different TREM2 genotypes was measured on the Nanodrop 8000 Spectrophotometer (Thermo Scientific). RNA (2 µg) was converted into cDNA (100 µl reaction volume) by using the High-Capacity cDNA reverse transcription kit (Thermo Fisher) and then diluted 1/9 into RNase-free water. For the real-time qPCR, 4.5 µl/well of diluted cDNA was combined with 0.5 µl/well of 20 × PrimeTime Std qPCR Taqman assay (IDT Integrated DNA Technologies) and 5.0 µl/well of Prime Time Gene Expression 2 × Master Mix with ROX reference dye (IDT Integrated DNA Technologies) into MicroAmp™ Optical 384-Well Reaction Plate with Barcode (Applied Biosystems). Reactions were evaluated on the QuantStudio 7 Pro Real-Time PCR system, using 384-well (Applied biosystems) default settings with an initial denaturation step at 95 °C for 10 min, followed by 40 cycles of 15 s at 95 °C for denaturation and 1 min at 60 °C for hybridization. For each assay all the samples were run in triplicates. The following 20 × PrimeTime Std qPCR Taqman assays were used:

The raw data files extracted from the QuantStudio 7 Pro Real-Time PCR system were uploaded into qBase + software version 2.4 (Biogazelle) for quality control and quantification of the Calibrated Normalized Relative Quantities (CNRQ). Only samples with at least two out of the three replicates with a replicate variability (difference in Cq) smaller than 0.5 were included. Reference genes used for normalization were determined by geNorm analysis. The optimal number of reference genes in our experimental setup was two (geNorm V < 0.15) and the most stable reference genes for both cortex samples and hippocampus samples were Hprt and Gapdh. Per tissue type scaling was done to the average of the control group 5xFAD Trem2 ^{+ /+} . For left and right hemisphere samples the CNRQ values were pooled per genotype group. CNRQ values were visualized using GraphPad Prism.

Microarray analysis: RNA from cortex and hippocampus was prepared as described above. Amplification and labelling of total RNA were performed using the GeneChip® PICO Reagent Kit following the manufacturer’s protocol (ThermoFisher 2016, P/N 902,790). Biotin-labeled target samples were hybridized to the GeneChip® Clariom S Mouse HT containing probes for over 20,000 genes. Target hybridization was processed on the GeneTitan® Instrument according to manufacturer’s instructions provided for Expression Array Plates (P/N 702,933). Images were analyzed using the GeneChip® Command Console Software (GCC) (ThermoFisher). Microarray data were processed using the statistical computing R-program (R version 3.6.1) and Bioconductor tools [12]. The gene expression values were normalized using Robust Multi-array Average (RMA) [21]. Individual probes were grouped into gene-specific probe sets based on Entrez Gene using the metadata package clariomsmousehtmmentrezg (version 22.0.0) [8].

The genetic linkage of microglial TREM2 loss-of-function variants with increased risk of AD has led to numerous studies [14, 22, 30, 46, 47, 52] characterizing the consequences of reduced TREM2 expression on Aβ pathology and plaque-associated neuritic damage. There have been differences in the findings among these studies, particularly with regard to the effect of TREM2 KO on Aβ plaques, but a general consensus has emerged that TREM2 deficiency results in reduced microglial containment of plaques, leading to increased Aβ-mediated damage to nearby neuronal processes [13, 24]. The differing reports on the effect of TREM2 deficiency on Aβ plaque burden may be due to the use of differing mouse models and mice of different ages.

Although there is experimental support for the hypothesis that reduced microglial interaction with Aβ plaques in TREM2 KO mice leads to increased damage to nearby neuronal processes, reports of microglial depletion with CSFR1 inhibitors in Aβ plaque-forming mice suggest that reduced microglial containment of plaques is not always harmful to neurons. For example, treatment of 10-month old 5XFAD mice [31] with the CSFR1 inhibitor, PLX3397, for one-month resulted in the elimination of ~ 80% of microglia [40] and although this treatment did not affect Aβ plaque levels, it did rescue dendritic spines and reduce neuronal loss in the subiculum. This suggests a detrimental role of microglia on neuronal function that is independent of effects on Aβ plaques and that a reduction in plaque- (and non-plaque)-associated microglia can be beneficial. This conclusion is further support by a report with another CSFR1 inhibitor in APP/PS1 transgenic mice, where dosing from 6- to 9-months of age led to improvements in cognitive function without significant changes in Aβ plaque load [32].

We have further investigated the consequences of reduced microglial interaction with Aβ plaques utilizing our previously described model of Aβ plaque and NP tau pathology that is facilitated by intracerebral seeding of 5XFAD mice with AD brain-derived tau [19]. Initial studies assessed the consequences of microglial depletion in this model utilizing PLX3397, with dosing initiated at 1.5 months of age when Aβ plaques have not yet formed in the brains of the 5XFAD mice. Dosing continued until 3 months of age, when the mice received intracerebral injection of AD-tau, with PLX3397 exposure maintained until the mice reached 6 months of age. As expected, a large reduction of microglia was observed in the brain, including plaque-associated microglia that was particularly evident in the cortex. In agreement with recent studies in young 5XFAD mice [37, 38], treatment with the CSFR1 inhibitor led to a reduction in Aβ plaques that was significant in the cortex but not in the subiculum. Moreover, plaques that did form in the PLX3397-treated 5XFAD cortex had a smaller overall size. The reduction of cortical plaque burden in the 5XFAD mice treated with PLX3397 led to a decrease in plaque-associated APP-positive dystrophic processes, with a corresponding reduction of AT8-positive NP pathology. Moreover, there was a trend toward decreased neuritic dystrophy in proximity to remaining cortical plaques in the CSFR1 inhibitor-treated mice, with a significant correlation between the amount of plaque Aβ and neuritic dystrophy in both PLX3397- and vehicle-treated 5XFAD mice. This suggests that the degree of neuritic damage was driven by the amount of plaque Aβ irrespective of the extent of microglial-plaque interaction.

In contrast to CSFR1 inhibitor-mediated microglial ablation, neither partial or complete reduction of TREM2 expression led to significant changes in Aβ plaque burden in combined groups of male and female 5XFAD mice, although we did note sex-dependent differences in which TREM2 deficiency resulted in increased plaque burden in female 7-month old 5XFAD mice and a trend toward reduced plaque load in male 5XFAD mice. Although prior studies [22, 46] have examined the effect of TREM2 KO on Aβ plaque formation, with evidence of increased plaque burden in older 5XFAD × TREM2 KO mice relative to control 5XFAD mice, these studies did not differentiate male and female mice. However, a recent study [30] found results similar to ours in that plaque load was significantly increased in 6–7- month old female PS2APP × TREM2^−/− mice compared to TREM2-sufficient PS2APP mice, whereas this change was not observed in the corresponding male mice. The authors did not speculate on the cause of this sex-dependent difference and we are unsure of the mechanism leading to these differences, although it may possibly relate to greater Aβ production by young female than male 5XFAD mice [31, 50]. For example, if TREM2 deficiency affects microglial phagocytosis or catabolism of Aβ, then elevated Aβ release by young female 5XFAD mice may result in enhanced plaque deposition when microglia are TREM2-deficient, whereas this would not be as apparent in young male 5XFAD mice with lower basal Aβ release. However, this is only a hypothesis and further investigation would be required to gain a better understanding of these observations.

A notable difference between CSFR1 inhibitor treatment and TREM2 deficiency is that the former depletes brain microglia, whereas loss of TREM2 affects microglial phenotype. Thus, TREM2 heterozygous or KO microglia may still contribute to the production of Aβ plaques and plaque-associated neuritic damage, and our data suggest that decreased TREM2 expression may in fact facilitate plaque development in female 5XFAD mice. Notably, a very recent publication reveals that microglial phagocytosis of Aβ promotes the formation of mature plaques, and that KO of the TAM receptor kinases Axl and Mer results in fewer plaques in APP/PS1 mice [20]. This group showed that TREM2^−/− microglial are capable of greater Aβ phagocytosis than the Axl and Mer KO mice, suggesting that they could still contribute to plaque biogenesis. However, the ability of TREM2-deficient microglia to contribute to plaque formation may wane with age, as another recent study revealed that anti-sense oligonucleotide knockdown of TREM2 expression in 10-month old APP/PS1 mice led to a reduction of Aβ plaque burden, whereas TREM2 KO had no effect on plaque levels in younger APP/PS1 mice at 4- or 7-months of age [36].

Consistent with prior reports, the 5XFAD × TREM2^−/− mice in our study had a significant reduction of microglial clustering around Aβ plaques. 5XFAD × TREM2^+/− mice had a modest, non-significant decrease of plaque-associated microglia, with significantly more of these microglia than 5XFAD × TREM2^−/− mice. In this regard, a recent study found that TREM2 KO caused a significant reduction of plaque-associated microglia in bigenic mice expressing both tau and APP transgenes, whereas microglia expressing a single copy of TREM2 showed normal plaque interaction [27]. It is thus notable that AD-tau injected 5XFAD × TREM2^+/− mice showed an increase of NP tau pathology relative to 5XFAD × TREM2^+/+ mice, including a trend toward greater NP tau than was observed in 5XFAD × TREM2^−/− mice. These data, like those from the PLX3397-treated 5XFAD mice, suggest that the extent of microglia clustering around plaques is not entirely predictive of plaque-associated neuritic damage. DAM gene expression profiling revealed that the 5XFAD × TREM2^+/− mice had a reduction of Stage 2 DAM markers that was roughly half the magnitude observed in the 5XFAD × TREM2^−/− mice. Moreover, microarray analyses of brain mRNA revealed significant differences between the 5XFAD × TREM2^−/− and 5XFAD × TREM2^+/− mice, including changes in pathways associated with microglial function and activation. Thus, although the TREM2^+/− microglia clearly show differences compared to TREM2^−/− microglia in gene expression and degree of plaque engagement, the extent of NP tau formation in the 5XFAD × TREM2^+/− mice is equal to or somewhat greater than in 5XFAD × TREM2^−/− mice. A somewhat similar trend was previously observed, where TREM2 haploinsufficiency was found to be more detrimental than TREM2 KO in the development of tau pathology in a mouse model with mutant tau over-expression [35]. The TREM2^+/− microglia in these tau transgenic mice had increased expression of inflammatory proteins and complement transcripts relative to TREM2^−/− microglia. Gene ontology analyses of microarray data from our studies revealed significant differences between 5XFAD × TREM2^−/− and 5XFAD × TREM2^+/− mice in pathways related to microglial activation in the cortex and hippocampus, although cytokines and complement proteins were not among the most significant differentially expressed mRNA in our study.

The data reported here are generally aligned with another recent study that reported an increase of NP tau pathology after intracerebral AD-tau injection into APP/PS1 × TREM2^−/− mice, as well as in APP/PS1 mice expressing human R47H TREM2 in the absence of endogenous mouse TREM2 [28]. APP × TREM2^+/− mice were not examined in this study, and in this regard R47H TREM2 microglia may more closely resemble TREM2^−/− than TREM2^+/− microglia since the R47H TREM2 transcripts undergo atypical mRNA splicing in mice that results in a premature stop codon and reduced TREM2 protein expression [49]. Thus, heterozygous TREM2 mice may serve as a representative model of the single copy loss-of-function TREM2 variants that cause increased risk of AD and may lead to somewhat different results and interpretations than would be obtained with TREM2 KO mice, as disclosed here.

We thank Janssen Pharmaceutica NV for providing funding, Dr. Diederik Moechars for his support and review of these studies, and Yuemang (Mandy) Yao for technical assistance.

TD, data acquisition and analysis; AN, data acquisition, analysis and interpretation, manuscript preparation; WT, data acquisition; AD, data acquisition, analysis and interpretation, manuscript preparation; IV, data acquisition and analysis; BZ, data acquisition; EM, data acquisition; MFO, data acquisition; PK, data acquisition; SM, study design; JQT, data interpretation and manuscript editing; VML, data interpretation and manuscript editing; KRB, study design, data analysis and interpretation, manuscript preparation.

Funding was provided by Janssen Pharmaceutica, NV to Dr. Brunden to support these studies.

Study data will be made available upon reasonable request.