Animals

Mice were cared for by the Yale Animal Resource Center and all experiments, including animal husbandry, genotyping, behavioral testing, and euthanasia, were approved by Yale's Institutional Animal Care and Use Committee (IACUC, protocol #07281). Animals were housed in groups of 1–5 mice per cage with access to food and water ad libitum. 12-h light/dark cycles were maintained throughout the duration of animal housing. All strains were maintained on a pure C57Bl6J background after more than 10 backcrosses. The AppNL−G−F/NL−G−F, MapthMAPT/hMAPT mice were provided by Drs. Saito and Saido, RIKEN Center for Brain Science [19, 24, 25]. The Prnp−/− mice, Edinburgh strain, were obtained from Dr. Chesebro of the Rocky Mountain Laboratories [12, 26]. We utilized the following abbreviated strain designations:

$$\begin\mathrm~\mathrm~^,\,^,\,^\\\mathrm~\mathrm~^,\,^,\,^\\ ^\mathrm\;~^,^,^\\\mathrm;\,^\mathrm\;~^,^,^\end$$

DKI; Prnp−/− mice were generated by first crossbreeding DKI mice with Prnp−/− mice to generate triple-heterozygous AppNL−G−F/+, MapthMAPT/+, Prnp± mice. The triple-heterozygous mice and their offspring were interbred to create a colony of triple-homozygous DKI; Prnp−/− mice. Mice were genotyped by Transnetyx (Cordova, Tennessee, USA).

PET imaging of synaptic density

[18F]SynVesT-1 was synthesized at the Yale PET Center as utilized for WT and DKI mouse imaging as described [19, 27]. PET measurements were collected on a Focus 220 (Siemens Medical Solutions). PET data were acquired over the period of 30–60 min post intramuscular injection of [18F]SynVesT-1, followed by a transmission scan using 57Co. Anesthesia was with isoflurane.

The cerebellum was the reference region for BPND values and calculating SUVRCB. Differences in SUVRCB between groups were assessed on a voxel-by-voxel basis in Matlab R2018 (Mathworks) with Statistical Parametric Mapping (SPM12). The Z-score for the difference of two groups was calculated in the framework of the general linear model per voxel. Resulting Z-score maps were projected onto the T2-weighted Ma-Benviste-Mirrione mouse brain template.

Novel object recognition

Mice were tested for novel object recognition as described [13, 20, 21, 28, 29]. Briefly, mice were handled for 5 min a day for 3 days prior to testing to familiarize with the experiment and reduce anxiety. Mice were placed in clean, empty, rectangular, covered rat cages to habituate for 1 h. During acquisition, mice were briefly removed, and 2 identical objects were placed 1 inch from either edge along the long axis of the cage. The objects included: (a wrapped 5 mL plastic syringe or single 15 mL conical tube with an orange cap (3 months), and a large, black binder clip or large glue stick (9 months)). Familiar object choice was pseudorandom. Mice were placed facing perpendicular to the long axis of the cage and a 10-min timer was started. Mice were allowed to explore the two identical objects for the duration of the 10 min, and the time to accrue 30 total seconds of orofacial object exploration was recorded. The objects were removed and either discarded (15 mL conical tube or 5 mL wrapped syringe) or wiped down with 70% ethanol. Mice were left in their cages for 1 h while other mice underwent acquisition.

During testing, one each of the novel and familiar objects were placed on pseudorandom sides of the cage. Orofacial object exploration was timed until 30 total seconds was reached. After testing trials, rat cages were cleaned to eliminate scent cues. The experimenter was blind to novel object identity and genotype. Mice that did not explore both objects or failed to accrue 30 total seconds of orofacial object exploration in less than 6 min during either acquisition or testing were removed from the analysis.

Morris water maze

Spatial learning and memory were analyzed using the Morris water maze as previously described [20]. Briefly, testing was performed in a circular, open pool ~ 1 m in diameter with the water kept at room temperature (~ 23 °C). Throughout testing and analyses, the experimenter was blinded to genotype. Animals were randomly divided into two cohorts of approximately 40 subjects and swims were conducted in 2 sets of 4 days for 7 consecutive days (day 4 of the first cohort overlapping with day 1 of the second cohort). Acquisition trials consisted of 8 attempts per day for 3 consecutive days, divided into one morning and one afternoon training session of four swims each. For each trial, mice were placed in one of four drop zones facing away from the pool’s center in the quadrant opposite the target quadrant, which contained a submerged plexiglass platform. The order of the drop locations varied across each session. The location of the target quadrant was varied in data acquired at 3 months and 9 months of age. Attempts were considered complete once a mouse located and spent 1 s on the platform. On the first day only, mice were allowed to spend 15 s on the platform until removal from the pool. Additionally, on the first day only, if a mouse did not find the platform within 60 s, it was guided to the platform and allowed to spend 15 s on the platform until removal from the pool.

On day 4, the platform was removed for the probe trial assessment. Mice were placed facing away from the center of the pool between the drop zones furthest from the target quadrant and tracked over 60 s of swim time. Latency to platform during the training swims and percent time spent in the target quadrant during probe trials were recorded on a JVC Everio G-series camcorder and tracked by Panlab’s Smart software. After completion of the probe trial, a visible platform was placed in the target quadrant and trials were administered until latencies stabilized over a maximum number of 15 trials. The latencies for the final 3 trials were averaged.

Immunohistology of mouse brain sections

Mice were euthanized by CO2 inhalation and perfused with ice-cold PBS before decapitation and rapid brain dissection. Right brain hemispheres were drop-fixed in 4% paraformaldehyde for 48 h at 4 °C before transfer to PBS at 4 °C. Right brain hemispheres were sliced into 40 μm cortical brain sections using a Leica WT1000S vibratome and free-floating sections transferred to PBS with 0.05% sodium azide for long-term storage.

For AT8 staining, slices were stained using Alexa Fluor 594 Tyramide SuperBoost kit, Streptavidin (Thermofisher; B40935) according to manufacturer instructions. An antigen retrieval step was performed for pThr217 and PSD-95 single staining prior to the blocking step by incubating slices in 1 × Reveal Decloaker buffer (Biocare Medical; RV 1000 M) for 15 min at 90 °C in an incubator and then 15 min at room temperature. For pThr217 staining, sections were incubated in blocking buffer (1% BSA + 1% Triton X-100 in PBS) for 1 h and then incubated in primary antibodies overnight (~ 24 h) at 4 °C in blocking buffer. For all other stains, sections were blocked in 10% normal horse serum (Jackson ImmunoResearch Laboratories) in PBS + 0.2% Triton X-100 for 1 h and then incubated with primary antibodies overnight (~ 24 h) in 1% normal horse serum in PBS at 4 °C.

The following primary antibodies were used: anti-AT8 (Phospho-TAU (Ser202, Thr205), Biotin; Invitrogen MN1020B; 1:100), anti-LAMP-1 (1D4B) (lysosome-associated membrane protein 1; Santa Cruz sc-19992; 1:500) anti‐IBA1 (ionized calcium‐binding adapter molecule 1; Wako 019‐19,741; 1:250), anti-PSD-95 (for PSD-95 single stain) (postsynaptic density protein 95; Invitrogen 51–6900; 1:250), anti-PSD-95 (for PSD-95/C1Q co-stain) (Millipore MAB1596; 1:200), anti-SV2A (synaptic vesicle glycoprotein 2A; Abcam 32,942; 1:250), anti-NEUN (Fox-3; Millipore ABN91; 1:500), anti-C1Q (Abcam ab182451; 1:1000), anti-pThr217 (phospho-TAU threonine 217; Invitrogen 44–744; 1:200), anti-pS396 (phospho-TAU serine 396; Invitrogen 44-752G; 1:250), anti-PrP (8H4) (prion protein; Abcam ab61409; 1:1000), anti-GFAP (glial fibrillary acidic protein; Abcam ab4674; 1:2000).

Sections were washed in PBS three times for 5 min each and incubated in secondary antibodies in PBS (goat or donkey anti-rabbit, anti-mouse, or anti-rat fluorescent antibodies; Invitrogen Alexa Fluor; 1:500) for 1 h at room temperature in the dark. Sections were washed in PBS three times for 5 min each and then mounted onto glass slides (Superfrost Plus, Fisher Scientific) and cover-slipped with Vectashield (Vector Laboratories H‐1200) antifade aqueous mounting medium.

For Thioflavin S (Sigma; T1892) staining, slices were washed twice for 5 min with 70% ethanol, stained with 0.1% ThioS in 70% ethanol for 15 min at room temperature, and washed twice for 5 min with 70% ethanol before being mounted as previously described. For pThr217 staining, after the three post-secondary PBS washes, slices were briefly dipped in ddH2O and then incubated in CuSO4 buffer (10 mM CuSO 4 in 50 mM ammonium acetate buffer, pH 5) for 15 min at room temperature. Slices were then briefly dipped again in ddH2O, returned to PBS, and mounted as previously described.

Imaging and analysis of immunohistochemistry

Staining for synaptic proteins with anti-SV2A and anti-PSD-95 antibodies was imaged using a Zeiss LSM 800 confocal microscope with a 63X 1.4 NA oil-immersion lens. The percent area occupied by immunoreactive synaptic puncta from the polymorphic layer of the dentate gyrus was measured as described [12, 19]. For imaging of the tissue stained for AT8, NeuN, GFAP, IBA1, CD68, pS396, pThr217, Lamp1, Aβ, and Thioflavin S, a Zeiss LSM 800 confocal microscope with a 20X 0.8 NA air-objective lens was used. Imaged regions included hippocampus, retrosplenial cortex, and auditory cortex. pS396 was imaged in retrosplenial cortex medially, and auditory cortex laterally. GFAP, Iba1, CD68, Thioflavin S, and Aβ were imaged using a 5 × 3 tile and z‐stack through the full slice in the center of the hippocampus with maximum intensity projection. pThr217 was imaged using a 2 × 2 tile and z-stack through the full slice in CA1 with maximum intensity projection. AT8 and Lamp1 were imaged using a 2 × 2 tile and z-stack through the full slice in medial cortex with maximum intensity projection. The percent area occupied by immunoreactive signal was quantified using ImageJ software. Statistical analysis was based on separate mice, with two slices averaged per mouse [19].

C1q synaptic targeting and PrP/C1q colocalization experiments were completed as described [19]. Briefly, images were acquired using a Zeiss LSM 800 confocal microscope with Airyscan detector and a 63X 1.4 NA oil‐immersion lens. 5-layer z-stacks were acquired from three adjacent locations in CA1 per slice. Images were acquired using a magnification factor of 3 × and optimized XY pixel size and Z pixel dimension per Zeiss Zen software (~ 34 nm and 200 nm, respectively). Following image capture, images were 3D Airyscan processed in Zeiss Zen Blue with an Airyscan filter strength of 6. C1q synaptic targeting and PrP/C1q colocalization were calculated using the JACoP plug-in in ImageJ [30].

Immunofluorescent staining of brainstem tissue utilized tissue from 20-month-old mice. After 24 h of post-fixation, tissue was cut through the midline and sagittally sectioned into 50 μm free-floating sections by vibratome. Sections were collected and stored in serial order. Sections were blocked with PBS containing 0.1% Triton X-100 (American Bio; AB02025) and 1% BSA for 1 h at room temperature. Sections were incubated in rabbit anti-Tyrosine Hydroxylase (Millipore AB152; 1:500) primary antibody diluted in blocking buffer overnight at 4 °C. Slices were washed with PBS and then incubated with goat anti-rabbit Alexa Fluor488 (Thermo Fisher Scientific #A-11008, 1:500) secondary antibody for 1 h at room temperature in the dark. After washing in PBS, sections were dipped into distilled water and then incubated in ammonium acetate with 10 mM CuSO4 (pH 5) for 15 min at room temperature to minimize lipofuscin autofluorescence. Samples were then washed in PBS and incubated with DAPI (1:1000) for 1 h at room temperature. After further washing in PBS, sections were mounted onto glass slides (Superfrost Plus, Fisher Scientific) and coverslipped with Prolong Diamond (ThermoFisher Scientific; P36965) antifade aqueous mounting medium. Tyrosine hydroxylase sections were imaged with a Leica DMi8 confocal microscope with a 10X 0.4 numerical aperture air-objective lens. Seven µm z-stacks were acquired from locus coeruleus. Six image slices were z-stack projected via maximum orthogonal projection.

For immunohistological analysis of AT8, CD68, D54D2, GFAP, Iba1, Lamp1, ThioS, pS396, bAT8, and pThr217, one field from each of two stained brain sections for each mouse was imaged in the designated areas. Measurements from different sections were averaged to create one value for “n” mice. For synaptic PSD-95 and SV2a staining, and for the C1q/PSD-95 colocalization study, two brain sections were used with two images taken per section in the designated area. For tyrosine hydroxylase staining in the locus coeruleus, five consecutive brain sections containing the nucleus were used with one image per section. All images were acquired, processed, and analyzed by an experimenter blinded to animal genotypes by chip number. Target marker’s fluorescence intensity, and immunoreactive area were quantified using Image J with macros. Positive TH cell numbers were counted blindly to animal genotypes chip number.

Immunoblotting

Mice at 20-months age were sacrificed via rapid decapitation and hemispheres separated medially on ice using a sharp razor blade. Cortex were microdissected and homogenized on ice with polypropylene pellet pestles in three times the brain tissue weight in RIPA lysis buffer (Millipore 20–188) containing PhosSTOP (Roche; 4,906,845,001) and cOmplete-mini protease inhibitor cocktail (Roche; 11,836,153,001) to extract cytosolic proteins. After centrifugation for 45 min at 100,00 × g at 4 °C, the supernatants were collected and boiled in Laemmli sample buffer with 5% β-mercaptoethanol and 1% SDS at 95 °C for 5 min. Sample volumes were normalized to total protein concentration by BCA assay (Thermo Scientific; 23,225).

Supernatants were then subjected to SDS-PAGE through 4–20% Tris–glycine gels (Bio-Rad; 5,671,095) and transferred onto nitrocellulose membranes (Invitrogen #IB23001) using an iBlot 2 Gel Transfer Device (Invitrogen; IB21001). Actin was run on the same gel as the loading control. Nitrocellulose membranes were blocked in blocking buffer for fluorescent western blotting (Rockland; MB-070–010) for 1 h at room temperature and incubated overnight in primary antibodies at 4 °C. Anti-pTau S202/ T205 (AT8) (Thermo Fisher Scientific #MN1020, 1:700), anti-C1Q (Dako A0136; 1:1000), anti-beta Actin (D6A8) Rabbit mAb (Cell Singnaling 8457S; 1:2000) and anti-beta Actin (Abcam ab8226; 1:5000) were used as primaries in this experiment.

Membranes were washed 3 times 5 min each with TBST after primary antibody incubation and then incubated with appropriate secondary antibodies for 1 h at room temperature. Anti-Mouse IRDye 800CW (LI-COR Biosciences #926–32,212, 1:8000) and α-rabbit IRDye 680CW (LI-COR Biosciences; 926–68,023, 1:8000) were used as secondary antibodies. Immunoblots were visualized with a LI-COR Odyssey infrared imaging system. Quantification of band intensities was performed within a linear range of exposure and by ImageJ software. Protein levels were normalized to beta-actin level.

Single nuclei 10x  genomic sequencing and analysis

Mouse brain tissue collection and mouse brain nuclei isolation for single nuclei RNA-seq were completed as previously described [19]. Briefly, WT, Prnp−/−, DKI, and DKI; Prnp−/− mice were aged to 40.3 ± 0.7 weeks (10 months) or 143.9 ± 3.5 weeks (20 months) then sacrificed via rapid dissection. Cortical and hippocampal regions from the left-brain hemisphere were microdissected, pooled and immediately frozen on dry ice then stored at -80 °C until nuclei isolation. Sample tissue (50–100 mg wet weight) was homogenized, placed on top of homogenization buffer, and centrifuged for 1 h. Nuclei pellets were obtained, resuspended, and counted on a hemocytometer to inform subsequent dilution or concentration to 700–1200 nuclei/µl for generating single nuclei cDNA libraries.

Barcode incorporated single-nucleus cDNA libraries were constructed using the Chromium Single Cell 3’ Reagents Kit v3 (10 × Genomics) following the manufacturer’s guidelines. Sample libraries, for both 10- and 20-month-aged cohorts were then pooled and underwent batch sequencing on an Illumina NovaSeq 5000 using single indexed paired-end HiSeq sequencing. A sequencing depth of  > 300 million reads was achieved for all samples with an average read depth of 30,000 reads per nuclei. The resulting sequencer BCL files were demultiplexed into FASTQ files. Sequenced samples were then aligned to the mm10-2020-A Mus musculus reference genome using the Cell Ranger Count software (pipeline version 6.01, 10 × Genomics), generating barcoded sparse matrices of gene-nuclei raw UMI counts.

Sample gene count matrices were converted and combined into a single anndata object for quality control (QC) and downstream processing with Scanpy (version 1.8.2) [31], a Python-based gene expression analysis toolkit. Genes detected in less than 10 nuclei were discarded. Nuclei with over 5% of UMI counts mapped to mitochondria genes as well as nuclei with less than 50 genes or more than 8000 genes detected were considered outliers and also discarded. In order to perform comparative gene expression analysis, the retained UMI counts were normalized to their library size by scaling the total number of transcripts to 10,000 per nuclei then Log-transformed. In total, across all experimental groups, 206,575 (10 months, n = 17 samples) and 219,900 (20 months, n = 23 samples) nuclei and 27,479 genes were retained post QC processing and clustering.

Sample snRNA-seq data were clustered using Scanpy’s implementation of Seurat_v3 [32]. This process first finds nuclei expression state ‘integration anchors’ across sample datasets by identifying common highly variable genes (HVGs) using the highly_variable_genes function of Scanpy. In brief, using raw UMI counts, HVGs were identified by ranking the computed normalized variance of each gene across all nuclei. The set of HVGs identified within each sample dataset were then merged to remove batch-specific genes, after which the top 2000 HGVs were annotated within the anndata object. Next, the regression of total library size and mitochondrial transcripts per nuclei was performed with sample-based batch correction using the Combat function in Scanpy. The Scale function was used to scale the corrected expression matrix to a max unit variance of 10 standard deviations. A neighborhoods graph of the corrected matrix was constructed using the Neighborhoods function, with a knn restriction of 10 and the first 25 principal components, computed from HVGs, as input parameters. Integrated Leiden clustering of nuclei into cell-type subgroups was performed from the neighborhoods graph and dimensionally reduced in UMAP space for visualization. To refine the clustering, nuclei doublets were removed from the dataset and integrated Leiden clustering was reiterated. Nuclei were deemed doublets if associated with clusters enriched for cell-specific marker genes of multiple cell types within the same cluster, indicating mixed-cell expression profiles.

Nuclei clusters enriched with a particular set of marker genes were considered to be of the corresponding primary cell type. Using the rank_genes_groups function, the top highly differentially expressed marker genes for individual Leiden clusters were tested against the rest using the Wilcoxon Ranked-Sum test. The highest-ranked genes with singular cluster enrichment were deemed marker genes and their cell specificity was varied by literature to determine cluster cell types.

The rank_genes_groups function, with Wilcoxon Ranked-Sum test, was also used for differential expressed gene (DEG) analysis between experimental groups for each identified cell type. Only genes expressed in a minimum of 10% of all nuclei post-ranking were considered differentially expressed. Significant DEGs were defined as genes with an adjusted p-value (false discovery rate) less than 0.005 and Log-transformed fold change greater than ± 0.25 (Supplemental Table S1).

Gene set enrichment analysis of DEGs from specific cell types was performed using the network visualization software, Cytoscape (version 3.9.1) [33], with application extensions ClueGo (version v2.5.9) [34] and STRING (version 2.0.0) [35]. DEG lists were queried in ClueGo for the overrepresentation of functional and pathway term associations in Gene Ontology (GO) Molecular Functions, Reactome and KEGG ontology databases. A two-sided hypergeometric test with BH-adjusted p-value of 0.0005 and 4-gene threshold was used for term enrichment selection. GO fusion of represented terms was used to collapse redundant and related biological themes with more than 50% similarity of associated genes. Filtered term associations were organized as degree-sorted networks for optimal visualization. The same gene sets were separately tested for cellular localization enrichment using GO cellular compartments and protein–protein interaction (PPI) networks using STRING. Highly enriched PPI networks were selected using a confidence score of 0.4 for strength of gene associations with MCL clustering.

Experimental design and statistical analysis

Each animal was implanted with an implantable electronic ID transponder (Biomedic Data System Inc, NC1043806) encoding a unique chip number. Each experiment was performed and analyzed using the chip Id number with experimenters unaware of genotype.

One-way ANOVA with post hoc Tukey’s multiple comparisons tests, One-way ANOVA with post hoc Dunnett’s multiple comparisons tests, and unpaired two-tailed t-tests were performed using GraphPad Prism software, version 9. Group means ± SEM and sample sizes (n) are reported in each figure legend. Data were considered statistically significant if p < 0.05. For all figures, all statistically significant group differences are labeled. For any given group comparison, the absence of any indication of significant difference implies lack of significance by the applied statistical test.