Trem2 deficiency differentially affects phenotype and transcriptome of human APOE3 and APOE4 mice

This study adhered to the guidelines outlined in the Guide for the Care and Use of Laboratory Animals from the United States Department of Health and Human Services and was approved by the University of Pittsburgh Institutional Animal Care and Use Committee. APP/PS1dE9 (B6.Cg-Tg (APPswe, PSEN1dE9)85Dbo/J) and Trem2^em2ADiuj/J mice were purchased from The Jackson Laboratory (USA) and human APOE3 (B6.129P2-Apoe^{tm3(APOE*3)Mae}N8) and APOE4 (B6.129P2-Apoe^{tm3(APOE*4)Mae}N8) targeted replacement mice from Taconic (USA) [29]. APP/PS1dE9 mice express mutant familial variants of human amyloid precursor protein (APP) with Swedish mutation, and human presenilin 1 carrying the exon-9-deleted variant (PSEN1dE9). All purchased mice were on a C57BL/6 genetic background and crossbred for at least 10 generations in our laboratory.

APP/PS1dE9 mice were bred to human APOE3^+/+ or APOE4^+/+ targeted replacement mice [3, 30] to generate APP/PS1dE9/APOE3^+/+ (referred to as APP/E3, APP/PS1dE9/APOE4^+/+ (APP/E4), APOE3^+/+ (E3), APOE4^+/+ (E4) mice expressing human APOE isoforms and wild-type Trem2. Trem2^−/− mice were bred to human APP/PS1dE9/APOE3^+/+ or APP/PS1dE9/APOE4^+/+ targeted replacement mice to generate APP/PS1dE9/APOE3^+/+/Trem2^−/− (referred to as APP/E3/Trem2^ko), APP/PS1dE9/APOE4^+/+/Trem2^−/− (APP/E4/Trem2^ko), APOE3^+/+/Trem2^−/− (E3/Trem2^ko); and APOE4^+/+/Trem2^−/− (E4/Trem2^ko) mice. All APOE3 or APOE4 mice were littermates and fed normal mouse chow diet ad libitum. Mice had water accessible at all times and were kept on a 12-h light-dark cycle. Male and female mice from each genotype were used for this study at an average age of 6.5 months.

Novel object recognition (NOR) was performed as previously described [31] with minor modifications. The NOR task assesses recognition memory and is based on the spontaneous tendency of mice to explore a novel object over a familiar one. Mice were placed in individual containers before any testing then returned to their housing cages after the daily testing was completed. Each mouse was handled for 3 min. for three successive days before testing to reduce anxiety. The NOR task was performed over three consecutive days, each pertaining to a unique phase. On Day 1, habituation phase, each animal was allowed to freely explore an open arena (40 cm X 40 cm X 30 cm tall white plastic box) for two 5 min. trials with a 5 min. inter-trial interval. On Day 2, familiarization phase, each animal was returned to the same arena for two 5 min. trials separated by a 5 min. intertrial interval with two identical objects (tower of LEGO® bricks 8 cm X 3.2 cm, built using white, blue, yellow, red, and green bricks) located in opposite diagonal corners of the arena. After a 24-h retention period, the testing phase was initiated on Day 3. The animal was returned into the arena with two objects in the same positions as the previous day, but one object was replaced with a novel object (metal bolt and nut of similar size). Mice were allowed to explore for one 10 min. interval. The exploration of both objects was recorded and scored with ANY-maze software (Stoelting Co., USA). The exploration by the software was defined as the mouse sniffing, climbing on, or interacting while facing an object within 3 cm. Mice were consistently placed into the middle of the arena facing the posterior wall to prevent any object preference. The arena and objects were cleaned with 70% ethanol between animals to prevent any olfactory cues. Animals that failed to have a total exploration time of 10 s for the objects during the novel phase were excluded from the analysis. The total distance traveled by each mouse was recorded during the habituation phase to assess locomotor activity. The percent exploration was determined by dividing the time exploring the novel object by the total time exploring both objects. This calculated value provides an indicator of recognition memory, with less time spent exploring the novel object signifying memory deficits.

Contextual and Cued Fear Conditioning (CCFC) was performed as previously described [31]. CCFC provides a measure of memory in relation to receiving a mild foot shock to a particular environment (context) or stimulus (cue). CCFC testing was initiated 24 h following completion of NOR and was performed for three consecutive days. On Day 1, training phase, mice were placed in a conditioning chamber (Stoelting Co., USA) for 5.5 min. The first 2 min. was silent, allowing the mouse to acclimate to the chamber. This was followed by a 30 s tone (2800 Hz; y 85 dB, conditioned stimulus (CS)) ending in a 2 s foot shock (0.7 mA, unconditioned stimulus (US)) through the floor of the conditioning chamber. The process was repeated one more time (learning phase) and ended with 30 s of reacclimation. On Day 2, contextual phase, mice were placed in the same conditioning chamber for 5 min. with no tone or shock administered, to measure contextual fear conditioning. For the final day, the gray walls of the chamber were covered with black and white striped walls to introduce a novel environment for assessing cued fear conditioning. Mice were placed in the conditioning chamber for 5 min. After the first 2 min. of silence (novel phase), the tone was administered for 3 min. (cued phase). Testing was performed at the same time of the day to ensure 24-h between phases. The chamber was cleaned with 70% ethanol between each animal. Freezing time was defined as the absence of movement except for respiration and recorded using ANY-maze software. Animals that had below 30 s total freezing time during the contextual phase were excluded from the analysis. Freezing time was calculated as percent freezing of the total time in the chamber during each phase of testing. Since freezing behavior is a fear characteristic in rodents, memory deficits were defined as diminished freezing when reintroduced to the context or cue from the training phase.

Two days post behavior, mice were anesthetized by intraperitoneal injection using Avertin (250 mg/kg of body weight). Blood was collected from the right ventricle through a cardiac puncture, followed by transcardial perfusion through the left ventricle with 20 mL of cold 0.1 M phosphate buffered saline (PBS), pH 7.4. The brain was removed and divided into hemispheres. One hemisphere was dissected into the cerebellum, subcortical, hippocampus and cortex regions and flash frozen on dry ice. A separate section of the anterior cortex was removed for whole-brain RNA-seq. The other hemisphere was drop fixed in 4% phosphate-buffered paraformaldehyde at 4 °C for 48 h and stored in 30% sucrose until sectioning [3]. Hemibrains were mounted in O.C.T. and cut in the coronal plane at 30 μm sections using a frozen cryotome (Thermo Scientific, USA). Six serial sections were collected with each section 450 μm apart, starting approximately 150 μm caudal to the first appearance of the dentate gyrus; covering an area in the brain from bregma − 1.25 mm and ending at bregma − 3.95 mm. Sections were stored in glycol-based cryoprotectant at − 20 °C until histological staining [3].

Methoxy-X34 (X34) 1,4-bis (3-carboxy-4-hydroxyphenylethenyl)-benzene, was provided by William E Klunk, MD, PhD, (University of Pittsburgh). For in vivo labeling of dense plaques, we used Methoxy-X04 (X04) 1,4-bis (4′-hydroxystyryl)-2-methoxybenzene, synthesized in W. Klunk’s lab (University of Pittsburgh). X04 readily crosses the blood-brain barrier [32] and remains bound to plaques for at least 90 days [33]. One mg/mL X04 stock solution was made by dissolving X04 in a vehicle consisting of 4% DMSO and 7.7% Cremophore EL in phosphate buffered saline (PBS).

For the OC-X34 dual stain, free-floating sections were washed followed by 30 min incubation in 50 mM sodium citrate buffer at 80 °C for antigen retrieval. Sections were then washed, blocked with 5% normal horse serum made in 0.1% triton-X 100 PBS for 1 h and finally incubated in anti-OC (1:1000, Millipore) primary antibody overnight at 4 °C [34]. The following morning, the sections were washed, incubated in horse anti-rabbit Dylight 594 (Di-1094 Vector Labs) for 2 h, and washed before mounting on positively charged glass slides. The slides were treated with 100 μM X34 followed by two 5 min destaining steps in 50% ethanol and coverslipped. Fluorescent images were taken using a Nikon Eclipse 90i microscope at 10X magnification. OC-X34 images were analyzed using Nikon NIS elements software and thresholding for the detection of plaques. The ratio between X34 positive (compact plaque) and OC positive areas (protofibular Aβ) was used to determine plaque compaction. Plaques with increased OC / X34 ratio are less compact than plaques with a ratio closer to one. We also assessed the area of OC not associated with X34 positive plaques (non-core bound OC) as a percentage of total detectable OC area (total OC).

A second series of brain tissue was used for 6E10 immunostaining as previously described [3] with some modifications. Antigen retrieval was performed on free-floating sections using 70% formic acid, followed by quenching of endogenous peroxidases with 0.3% hydrogen peroxide. The tissue was incubated in 3% normal goat serum (Vector, USA) then blocked for endogenous avidin and biotin. Sections were incubated in 6E10 biotinylated antibody (1:1000 Biolegend, USA) for 2 h and subsequently developed using the Vector ABC kit and DAB substrate kit (Vector, USA). Sections were mounted onto superfrost plus slides (Fisher Scientific, USA) and coverslipped. Bright-field images were taken using a Nikon Eclipse 90i microscope at 4X magnification.

Thioflavin S (ThioS) staining was performed on a third series of brain sections. Sections were mounted onto slides, washed in PBS, and stained with 0.02% ThioS (Sigma, USA) in PBS for 10 min. Next, sections were differentiated in 50% ethanol for 2 min. After a final wash in PBS, slides were coverslipped. Fluorescent images were taken using a Nikon Eclipse 90i microscope at 4X magnification.

To quantify plaque pathology, two separate regions of interest (ROI) were drawn around the cortex and hippocampus for each section and an image intensity threshold was established to detect the stained plaques compared to the background using NIS Elements software (Nikon Instruments Inc., USA). OC, X34, 6E10, and ThioS staining values were represented as the area of staining normalized to ROI area or percentage of area covered by 6E10 or ThioS stain.

A forth series of brain sections were immunostained with anti-IBA1 antibody (WAKO, USA) and anti-GFAP antibody (Agilent, USA). Free-floating sections were washed, then antigen retrieval performed in sodium citrate buffer at 80 °C for 60 min, blocked in normal donkey serum (Jackson Lab, USA) for 1 h, and finally incubated in IBA1 antibody (1:1000) overnight at 4 °C. Sections were washed and transferred into secondary donkey anti-rabbit Alexa 594 (Invitrogen, USA) for 1 h, before being washed and transferred to GFAP antibody (1:1000) overnight at 4 °C. Again, sections were washed and transferred into secondary donkey anti-rabbit Alexa 488 (Invitrogen, USA) for 1 h, before being washed and mounted onto slides. Slides were stained with X34 as documented for ThioS, followed by DAPI staining, and coverslipped. Fluorescent images of individual plaques were taken using a Nikon Eclipse 90i microscope at 20X magnification. Plaques were chosen with an average area of 300 μm². The number of IBA1 positive microglia and GFAP positive astrocytes were counted in circular radiating regions of interest with a diameter of 10, 20, 40, and 60 μm from the edge of the X34 positive plaque.

A fifth series of brain sections were immunostained with anti-APOE antibody (Invitrogen, USA) and Thiazine Red (TR, Sigma Aldrich). Free-floating sections were washed, blocked in normal donkey serum (Jackson Lab, USA) for 1 h, and incubated in APOE antibody (1:100) overnight at 4 °C. Sections were transferred into secondary donkey anti-rabbit Alexa 488 (Invitrogen, USA) for 1 h, before being washed and mounted onto slides. Mounted slides were stained with 2 μM TR in PBS for 15 min. After a final wash, slides were dried and coverslipped. Fluorescent images of plaques were taken using a Nikon Eclipse 90i microscope at 10X magnification. Staining was defined by threshold analysis using NIS Elements software, and the area of APOE staining associated with TR positive plaque area was assessed.

A sixth series of brain sections were immunostained with anti-LAMP1 antibody (Abcam, USA) and X34. Free-floating sections were washed, blocked in normal goat serum (Jackson Lab, USA) for 1 h, and incubated in LAMP1 antibody (1:500) overnight at 4 °C. Sections were transferred into secondary goat anti-rat Cy5 (Vector, USA) for 2 h, before being washed and mounted onto slides. Mounted slides were stained with X34 as above. Fluorescent images of plaques were taken using a Nikon Eclipse 90i microscope at 10X magnification. Staining was defined by threshold analysis using NIS Elements software, and the area of LAMP1 staining associated with X34 positive plaques was assessed. For all plaque specific imaging, plaques were selected so they were at least 50 μm from the edge of the tissue, and at least 50 μm away from other plaques, with an even representation of all plaque sizes and composition across all groups to account for any bias introduced by differences in plaque stage, composition, or size.

X04 was administered intraperitoneally (i.p.) at a concentration of 10 mg/kg to mice at either 3.5 or 5.5 months of age and sacrificed after 30 days for the collection of brain tissue for in vivo labeling of dense core amyloid plaques. Tissues used for the X04-TR-IBA1 triple stain allowed for the assessment of plaque growth dynamics and microglia plaque barrier. Free-floating sections were washed, incubated in 0.5 uM TR in PBS for 20 min followed by a final PBS wash. Sections were then incubated in 50 mM sodium citrate buffer for 30 min at 80 °C to perform antigen retrieval, washed, incubated in 5% normal goat serum made in 0.1% triton-X 100 PBS for 1 h, and finally incubated in anti-IBA1 (1:1000, Wako) primary overnight at 4 °C. The following morning the sections were washed, incubated in goat anti-rabbit Dylight 488 (Di-1488 Vector Labs) for 2 h, and washed before mounting on positively charged glass slides. X04-TR-IBA1 triple stained tissues were imaged on all three channels using an Olympus FV1000 confocal microscope at 60x, with 1.5 μm step size. For confocal imaging, plaques were selected if they were at least 50 μm from the edge of the tissue, and at least 50 μm away from other plaques, with an even representation of all plaque sizes and composition across all groups.

To assess the size of β-amyloid plaques at each age group (4.5 and 6.5 months), we analyzed plaque volume using Imaris on an independent set of APP/E3, APP/E3/Trem2^ko, APP/E4 and APP/E4/Trem2^ko mice to extract quantitative data from the high-resolution three-dimensional confocal images. In the 48 h controls, plaques of all sizes and compositions were intentionally imaged in order to account variance when thresholding near detection limits. Greater than 94% of all the plaques imaged in the experimental groups fall within the minimun and maximum range of TR plaque volume analyzed in the 48 h control group. Additionally, when comparing the TR volume to X04 volume in the 48 h control plaques we saw a very high correlation (R² = 0.9579) and no departure from the linear regression line in either the extremely small, or extremely large plaques (see supplemental figure 3). Briefly, images were loaded into the Imaris (v9.3.1) environment and voxels less than 500 intensity were removed from all channels to reduce background noise. Surfaces were then generated for the X04 and TR channels and the volume of each surface created was analyzed. Surfaces were created to assess IBA1 coverage and, using the “surface-surface contact area” XTension, the percent of IBA1 / TR surface contact was calculated. A 1 voxel shell is generated surrounding the TR labeled amyloid plaque and any time IBA1 signal is colocalized with the TR shell it is counted as surface contact. The sum of the colocalized voxels divided by the total surface area of TR generated the percent surface contact. The surface area contacted by microglia is subtracted from the total surface area giving the exposed surface area of each plaque (surface area not covered by microglia). The change in plaque volume was calculated by subtracting the volume of the plaque at the time of sacrifice (TR) from the volume of the plaque at the time of in vivo labeling (X04). In the 48-h control mice, we found no significant difference in the volume of X04 and TR labeling, an average growth volume (TR-X04) near 0 and an average fold change near 1 (TR/X04).

The frozen cortices were homogenized according to previously published work [35]. Individual cortices were weighed and transferred into a glass Dounce containing the appropriate amount of tissue homogenization buffer (1 M TRIS base, 0.5 M EDTA, and 0.2 M EGTA) and protease inhibitor (Sigma-Aldrich, USA). Cortices were homogenized in 1 mL of tissue homogenization buffer and protease inhibitor per each 100 mg of tissue. Once homogenized, the tissue was spun in a centrifuge at 16,000 rcf for 1 h. The supernatant was kept for future use for the determination of soluble Aβ. 70% formic acid was then added to the pellet and sonicated for two fifteen sec intervals before being spun again at 16,000 rcf for 1 h. The resulting supernatant was kept and used for the determination of insoluble Aβ.

Aβ ELISA was performed according to previously published work [35] with modifications. The wells of MaxiSorp plates (Nunc) were coated by adding 100 μl/well of 6E10 antibody (Biolegend, USA) diluted to 5 μg/ml in Coating buffer (0.1 M NaHCO₃, 0.1 M Na₂ CO₃, pH 9.6) and incubated overnight at 4 °C while rocking. The next day, wells were washed with PBS and 200 μL/well of Block Ace Solution (1% Block Ace (Bio Rad, USA) in PBS, 0.05% NaN₃) was added. Plates were incubated for 4 h at room temperature with rocking to block non-specific binding. Once the Block Ace Solution was removed 50 μl/well of Buffer EC (20 mM sodium phosphate, 2 mM EDTA, 400 mM NaCl, 0.2% BSA, 0.05% CHAPS, 0.4% Block Ace, 0.05% NaN₃, pH 7.0) was added to the wells to prevent drying while adding samples. 100 μL of standards and samples were added to each well, and high-range samples were diluted with Buffer EC where necessary. For the insoluble Aβ fraction, samples were also neutralized with FA neutralization solution (1 M TRIS base, 0.5 M Na₂HPO₄, 0.05% NaN₃) before dilution. For standards, ranging from 0 to 0.8pM, equal parts of 8 pM stocks of Aβ40 and Aβ42 were used. Once standards and samples were loaded, the plates were incubated overnight at 4 °C with rocking. Plates were washed and 100 μL detection antibody HJ5.1 (1:3300, a gift from John Cirrito) diluted in 0.05%-PBSTween was added to each well and incubated for 90 min. at room temperature with rocking. After washing, 100 μL HRP40 secondary (1:16000, Fisher Scientific, USA) diluted in 1% BSA-PBSTween, was added to each well and incubated for 90 min. at room temperature with rocking. Wells were washed and 100 μL of prepared TMB Substrate/H₂O₂ Solution (Thermo Sci., USA) was added. Absorbance was read at 650 nm on a plate reader (Molecular Devices, USA). All samples were run on the same day in duplicates. The concentration of total Aβ in pM was interpolated using linear regression on GraphPad Prism 7.0 and multiplied by the dilution factor for each sample.

Bradford Assay. Protein concentrations were determined according to previously published work [36]. The Bradford assay was used to determine the protein concentration of all samples. Bovine Serum Albumin (BSA, Fisher Scientific, USA) standards ranging from 0 to 100 μg/ml were used. A 40% Bio-Rad Protein Assay Dye Reagent (Bio-Rad, USA) was prepared with a 1:1 volume of diluted samples and absorbance was read at 595 nm. Total Protein (μg/ml) was interpolated using linear regression on GraphPad Prism 7.0 and multiplied by the dilution factor for each sample. To normalize the data, total protein concentration from the Bradford assay (μg/ml) was divided by the pM concentration of total Aβ for each sample from ELISA.

In a separate cohort, mice were perfused, and tissue fixed and sectioned as documented above. RNAscope experiments were performed using the Multiplex Fluorescent Reagent kit v2 (Advanced Cell Diagnostics, USA) following the manufacturer’s recommendations with minor adjustments. Six freshly sectioned tissues per animal were mounted onto superfrost plus slides (Fisher Scientific, USA) within a 0.75″ × 0.75″ square, and baked at 60 °C for 60 min. Slides were incubated in X34 for 10 min. before being dehydrated using a series of ethanol dilution steps, then submerged in target retrieval reagent at 100 °C for 10 min. Protease digestion was performed at 40 °C for 30 min. using Protease III, and probe hybridization was carried out at 40 °C for 2 h. We used probe sets available from ACD for Clec7a, Csf1r, Apoe, and Tmem119. Following the amplification steps, the sections were counterstained with DAPI and coverslipped. Imaging was carried out using a Nikon Eclipse 90i microscope at 20X magnification with imaging of individual plaques and analyzed using NIS Elements software (Nikon Instruments Inc., USA). One circular ROI that extends 50 μm from the center of the X-34 positive plaque was drawn, and a threshold established for each probe to determine the area of puncta coverage. Four ROI’s of the same area were randomly selected from areas away from the plaque on the same image and averaged to determine the area of puncta coverage greater than 50 μm from plaque center. Apoe FISH analysis was performed on Tmem119-positive microglia within the same 50 μm ROI. The nuclei and surrounding area of cells with positive Tmem119 signal were outlined and identified as microglia. The intensity of Apoe FISH signal was normalized to the number of Tmem119 positive cells.

RNA was isolated from the frontal cortex and purified using RNeasy mini kit (Qiagen, Germany). RNA quality was assessed using 2100 Bioanalyzer (Agilent Technologies, USA) and only samples with RIN > 8 were used for library construction. Library generation was performed by Novogene Co. Ltd. and sequenced using an Illumina HiSeq 2500 instrument. Following initial processing and quality control, the sequencing data was aligned to the mouse genome mm10 using Subread/featureCounts (v1.5.3; https://sourceforge.net/projects/subread/files/subread-1.5.3/↗) with an average read depth of 50 million successfully aligned reads. Statistical analysis was carried out using Rsubread (v1.34.2; https://bioconductor.org/packages/release/bioc/html/Rsubread.html↗), DEseq2 (1.24.0; https://bioconductor.org/packages/release/bioc/html/DESeq2.html↗), EdgeR (v3.26.5; https://bioconductor.org/packages/release/bioc/html/edgeR.html↗), and WGCNA (v1.68; https://cran.r-project.org/web/packages/WGCNA/index.html↗) all in the R environment (v3.6.0; https://www.r-project.org/↗). For network analysis using WGCNA, samples were clustered by gene expression profile which enabled the detection of outliers that were removed from the downstream analysis. Modules were generated automatically using a soft thresholding power, β = 10, a minimum module size of 44 genes and a minimum module merge cut height of 0.25. To account for bias introduced by sequencing batch, we implemented empirical Bayes-moderated linear regression which removes variation in the data due to unwanted covariates while preserving variation due to retained covariates. Networks were built using the top 5 genes identified as hub genes from any given module (gene significance > 0.2, and module membership > 0.8). Following hub gene selection, all other connections generated from the top 5 genes were visualized using Cytoscape (v3.7.1). Functional annotation clustering was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID v6.8, https://david.ncifcrf.gov↗). All GO terms are considered significant if p < 0.05 following multiplicity correction using the Benjamini-Hochberg method to control the FDR.