BACE1 controls Clusterin levels in support cells to help clear beta-amyloid peptides

To understand the role of BACE1 in astrocytes, we enriched astrocytic cell population from 3 brains of wild-type (WT) and Bace1 KO mice at 2-months of age using astrocyte-cell specific antigen 2 (ACSA2) immunomagnetic beads. Purified astrocytes were subjected to 10 × Genomic single cell RNA sequencing and initial read count analysis by Cell Ranger (10 × Genomics). Overall gene number, read counts, and mitochondrial RNA appeared to be similar between all samples (Supplemental Figure S1A). Approximately 17,000 cells were identified for each genotype. Using Seurat R package (Version 4, Satija lab), we analyzed the sequencing results and clustered these cells into UMAP-defined clusters (Fig. 1A). Different cell clusters were arranged based on the unique gene signature as previously described [23, 24]: 1) major populations of non-reactive astrocytes [indicated by high Atp1b2 expression and low Vimentin (Vim) expression], 2) reactive astrocytes (high Atp1b2 and high Vim), 3) microglia [Cluster of differentiation 68 (CD68)], 4) oligodendrocytes [Myelin and lymphocyte protein (Mal)], 5) oligodendrocyte precursor cells (OPC) [Platelet-derived growth factor receptor A 2 (Pdgfr2) or Chondroitin sulfate proteoglycan 4 (CSPG4)], and 6) endothelial cells [Platelet endothelial cell adhesion molecule (PECAM1)] (Fig. 1B). From UMAP-defined clusters, it appeared that Cluster 0, 2, and 6 were non-reactive astrocytes while Clusters 3, 4, 8, and 10 were reactive astrocytes (Fig. 1B and Supplemental Figure S1B). As expected, astrocytes represented the majority of cells, and approximately 20% of remaining cells were identified as oligodendrocytes (ochre; Clusters 1 and 5), and OPCs (teal; Cluster 7 and 12). Although in low abundance, microglia (in cluster 9 in blue) and endothelial cells (cluster 7 in fuchsia) were present (Fig. S1B).

After using known brain cell type markers to filter out non-astrocyte cells [23, 25], we re-clustered and compared gene profiles derived from Bace1-null and WT astrocytes. Quantification showed that reactive astrocytes comprised approximately 20% of total cells in Bace1^−/− brains compared to only ~ 10% reactive astrocytes in Bace1^+/+ controls (Fig. 1C). Correspondingly, the proportion of non-reactive astrocytes in WT controls was greater, while other cell populations were not significantly altered under this immune-panning condition (Fig. 1C). Furthermore, BACE1 deficiency visibly increased numbers of reactive astrocytes (labeled as R Astrocytes in teal, Fig. 1D).

Consistent with this unbiased result, we found that GFAP protein levels in lysates from BACE1 deficient primary astrocyte cultures were significantly higher compared to WT controls (Fig. 1E and F), further indicating that BACE1 deficiency promotes astrocytes in the reactive states.

Comparing the transcriptomes of pan reactive astrocytes from Bace1 KO and WT mice using Seurat V4, we identified 37,658 unique gene signatures, with about 610 significantly differentially expressed genes (DEGs) (adjusted p-value < 0.05, log2FC > 0.2; Supplemental Table 1). A volcano plot revealed many significant DEGs (Fig. 2A). Among these genes, we found several genes known to have roles in synaptic maintenance, clearance of β-amyloid peptides (Aβ), glutamate homeostasis, metabolism, hippocampal synaptogenesis, insulin and insulin growth factor signaling and AP-1 transcription factor family (Table 1 and Supplemental Table 1). Expectedly, Bace1 from Bace1 KO reactive astrocytes was also significantly reduced.

Genes thought to regulate clearance of Aβ, marked in green triangles, received our particular attention. We showed that Clu and C-X-C Motif Chemokine Ligand 14 (Cxcl14) expression levels were significantly elevated in Bace1-null reactive astrocyte gene clusters (Fig. 2B and C). These changes were then further validated on the protein level. Western blot analysis showed that both CLU and CXCL14 protein levels were increased in Bace1-null primary astrocytes, both basally and when stimulated with Aβ, compared to WT astrocytes (Fig. 2D-E).

As previously mentioned, conditional knockout of BACE1 in the adult reverses previously formed amyloid plaques in adult 5xFAD;Bace1^fl/fl; UBC-creER compared to non-BACE1 deleted 5xFAD;Bace1^fl/fl adult mice [15]. To understand the role that adult BACE1 deficiency might play on astrocyte transcriptomes in adult mice with the 5xFAD background, we compared scRNA-Seq of astrocytes from 14-month old 5xFAD;Bace1^fl/fl; UBC-creER and 5xFAD;Bace1^fl/fl mice. Overall gene number, read counts, and mitochondrial RNA appeared to be similar between all samples (See Supplemental Figure S2). About 7,000 astrocytes were recovered from these mice after magnetic activated cell sorting (MACS) by ACSA2 immunomagnetic beads and cell type filtration. When comparing reactive astrocyte population transcriptomes from 5xFAD;Bace1fl/fl; UBC-creER and 5xFAD;Bace1^fl/fl mice, 1,873 DEGs (P < 0.05) were noted (Supplemental Table 2). Among these DEGs, insulin degrading enzyme (IDE), klotho (Kl) and lysosomal associated membrane protein 2 (LAMP2) were upregulated in the 5xFAD;Bace1^fl/fl; UBC-creER compared to 5xFAD;Bace1^fl/fl mice. Furthermore, we once again found significantly elevated Clu and Cxcl14 gene expression in the reactive astrocyte population (Fig. 3A), and an increase in the number of Clu^high and Cxcl14^high astrocytes in the case of 5xFAD;Bace1^fl/fl; UBC-creER mice (Fig. 3B).

Our data suggest that both germline and adult BACE1 deficiency increases a group of astrocytes that commonly express more Cxcl14 and Clu. These Clu^high and Cxcl14^high reactive astrocytes are likely capable of clearing more β-amyloid peptides (Aβ) as higher levels of CLU are known to increase Aβ clearance [26, 27].

Since BACE1 deficiency altered astrocytic transcriptomic profiles, we then asked whether astrocytes with BACE1 deficiency would enhance Aβ clearance, which might contribute to the reversal of pre-formed amyloid plaques in 5xFAD mice with deletion of Bace1 in the adult as previously reported [15]. We cultured primary astrocytes from WT and Bace1-null mice, and then treated these cultured astrocytes with aggregated oligomeric human Aβ₄₂ according to the published procedure [28]. We found that BACE1 deficient astrocytes had visibly increased uptake of HiLyte Fluor-555-labeled Aβ₄₂ (Anaspec, Fremont, CA) after 12 hrs incubation compared to WT controls (Fig. 4A). High magnification images showed that the majority of the HiLyte Fluor-555-labeled Aβ₄₂ signal was found within astrocyte cell bodies (Fig. 4A, inset). Quantification showed that enhanced uptake of aggregated Aβ₄₂ began at around 6 hrs post incubation and plateaued at around 12 hrs, while it took about 36 hrs for WT astrocytes to uptake about the same amount of HiLyte Fluor-555-labeled Aβ₄₂ (Fig. 4B).

We also examined Aβ levels by Western blot. Although the levels of oligomeric Aβ₄₂ in Bace1-null astrocytes were higher than that in WT controls during initial incubation (12 to 36 h), significantly less Aβ₄₂ was found at the time of 72 hrs (Fig. 4C). This was further confirmed by multiple replication experiments (Fig. 4D), indicating that BACE1 deletion likely also enhanced degradation of oligomeric Aβ₄₂ after initial uptake. In line with previous reports [20, 21, 29], astrocytes treated with Aβ₄₂ appeared to enhance expression of BACE1 (Fig. 4C), suggesting that elevated BACE1 in AD brains might have vicious inhibitory effect on astrocytic Aβ clearance.

Because CLU is known to regulate Aβ clearance and degradation and elevation of the CLU protein level was evident, we chose to examine whether elevated CLU levels would contribute to enhanced Aβ uptake and degradation in Bace1-null astrocytes. To this end, we identified that one well-characterized siRNA (Santa Cruz Biotechnology) was able to reduce CLU protein levels in primary astrocytes in a dose dependent manner, with astrocytes treated with 80 pmol of Clu siRNA down-regulated about 48.6% expression of CLU (Supplemental Figure 3A and B).

We then pretreated Bace1-null and WT astrocytes with 80 pmol of Clu or control siRNA before treating astrocytes with aggregated Aβ₄₂ and examined intracellular Aβ levels by either Western blot (Fig. 5A and B), or fluorescently tagged Aβ by confocal imaging (Fig. 5C). We showed that knocking down Clu in Bace1-null astrocytes reduced the uptake of Aβ₄₂, when compared to Bace1-null astrocytes treated with control siRNA (Fig. 5).

Further quantification showed a clear shift in the uptake pattern (Fig. 5B): slowed uptake of Aβ₄₂ by astrocytes when CLU levels were reduced, and treatment with Clu siRNA reduced total levels of Aβ₄₂ in both Bace1-null and WT astrocytes from 0.5 to 36 hrs compared to control siRNA treated astrocyte (Fig. 5B). Confocal images also showed delaying uptake of Aβ₄₂ when comparing Bace1-null astrocytes with and without specific Clu siRNA treatment (Fig. 5C).

Altogether, these results add evidence that CLU indeed plays a direct role in astrocytic clearance of Aβ. Furthermore, it suggests that increased CLU levels underlie the increased uptake of Aβ in Bace1-null astrocytes.

In order to investigate signaling pathways responsible for the increased expression of Clu and other aforementioned DEGs in BACE1 deficient astrocytes, we performed a database search using BioMart to determine common transcription factors for these DEGs. Members of the activated protein 1 (AP-1) transcription family were the most commonly found elements to be related to DEGs of BACE1 deficient astrocytes, including Cxcl14 and Clu. Furthermore, some members of the AP-1 family themselves were DEGs upregulated in BACE1 deficient reactive astrocytes (Fig. 2B, Table 1, and Supplemental Table 1). AP-1 activation is known to be modulated by upstream molecules such as P38, JNK1/2, and ERK1/2 [30, 31]. We found that the levels of JNK were not significantly changed, but levels of phosphorylated P38, Jun, and ERK1/2 were significantly elevated in Bace1-null astrocyte cultures (Fig. 6A); total Jun and P38 levels were not obviously changed. Quantification of replicated results confirmed this in Fig. 6B.

Our results imply that BACE1 deficiency in astrocytes results in an increase of pP38 and pERK1/2 activation, which then phosphorylates and activates downstream Jun and AP-1 mediated transcription of its downstream molecules such as Clu, which is related to enhanced astrocytic clearance of Aβ.

We further explored how BACE1 deficiency would activate the aforementioned signaling molecules in astrocytes. It was previously reported that insulin receptor (IR) is a BACE1 cleavage substrate in the liver [32], and insulin signaling regulates both P38 and ERK1/2 MAPK activity. Furthermore, some DEGs of the insulin and insulin growth factor families were altered in Bace1-null astrocytes (Table 1 and Supplemental Table 1). IR is composed of a heterodimer dimer, α and β subunits, and the membrane anchored β subunit is identified as a BACE1 substrate. When Bace1 is deleted, the IRβ subunit is no longer cleaved and more preserved IRβ subunit might be available for enhancing P38 and ERK1/2 activity. We therefore investigated whether BACE1 deficiency might also enhance astrocytic insulin receptor availability.

Western blot analysis revealed that BACE1 deficiency significantly reduced levels of BACE1-cleaved IRβ C-terminal fragment (CTF) in Bace1-null astrocytes compared to their WT controls (Fig. 7A), consistent with prior results that BACE1 deficiency abrogated cleavage of IRβ in astrocytes. An increase of total mature IRα and IRβ levels was also noted, while as well as a much larger increase in the phosphorylated IRβ (pIRβ) subunit (Fig. 7A), which is required for downstream activation for pP38 and ERK1/2 MAPK. Quantification of replicated results confirmed changes of these proteins (Fig. 7B). Altogether, this suggests that astrocytic BACE1 regulated IR receptor availability by cleavage in a manner similar to liver BACE1 and IR.

We further tested whether insulin signaling might be responsible for the elevated CLU and CXCL14 levels in Bace1-null astrocytes. Since insulin is already present in the G-5 supplemented astrocyte media, we inhibited insulin signaling by using BMS-754807 (SelleckChem), a potent IR tyrosine kinase phosphorylation inhibitor (Fig. 7C and D). BMS-754807 treatment resulted in a potent reduction of both CLU and CXCL14 levels in Bace1-null astrocytes (Fig. 7C). Quantification of replicated results confirmed changes of these proteins (Fig. 7D). This suggests that elevated CLU and CXCL14 levels in Bace1-null astrocytes is dependent on IR signaling.

Altogether, this suggests that astrocytic BACE1 regulated IR availability by cleavage in a matter similar to liver BACE1 and IR and that IR signaling may play a role in reactive astrocyte amyloid clearance.

To investigate whether in vivo astrocytic BACE1 deficiency results in increased IR, CLU and CXCL14, we bred Bace1^fl/fl mice with Gfap-cre mice [Tg-Gfap-cre 73.12Mvs, the Jackson lab] and examined Bace1^fl/fl;Gfap-cre mice, which conditionally deleted Bace1 mostly in astrocytes due to expression of Cre recombinase by the mouse Gfap promoter [33]. To avoid potential effects from other Gfap-expressing lineage cells, we isolated astrocytes from 2-month old Bace1^fl/fl;Gfap-cre mice using ACSA2⁺ immunobeads and the purified ACSA2 + astrocytes were used for Western blot (Fig. 8).

We found that astrocytes from Bace1^fl/fl;Gfap-cre mice had an increased amount of mature IRβ and reduced IRβ CTF compared to Bace1^fl/fl astrocytes in a manner similar to above in vitro conditions (comparing Fig. 7 and 8). We also found a significant increase in CLU and CXCL14 protein levels (Fig. 8). Altogether, this suggests that BACE1 deficiency increases astrocytic IR signaling and its downstream molecule CLU and CXCL14 in a mostly cell autonomous manner both in vivo and in vitro.

To further understand the effect of astrocytic Bace1 deletion on amyloid pathology, we bred Bace1^fl/fl;Gfap-cre with 5xFAD;Bace1^fl/fl mice to generate 5xFAD;Bace1^fl/fl;Gfap-cre and 5xFAD;Bace1^fl/fl mice for comparison. We used P75 female mice from each genotype, as female 5xFAD mice show less variations in amyloid pathology [34]. Thioflavin-S staining of fixed brain sections revealed an overall less amyloid depositions in 5xFAD;Bace1^fl/fl;Gfap-cre mouse brains compared to 5xFAD;Bace1^fl/fl mouse controls (Supplemental Figure 4). In an enlarged view, it was clear that less Aβ plaques were formed in the 5xFAD;Bace1^fl/fl;Gfap-cre cortex compared to controls (Fig. 9A), and was confirmed by quantification (Fig. 9B). Although, levels of Aβ plaques appeared to be reduced in the subiculum and hippocampus of 5xFAD;Bace1^fl/fl;Gfap-cre mice, further quantification showed that this decrease in hippocampus was, overall, not significant (Fig. 9B).

We also examined the effect of astrocytic BACE1 on total amyloid processing and generation by Western blot (Fig. 9C). We found that total levels of full length APP, C99, and C83 cleavage substrates appeared to be unchanged across cortical samples. This most likely indicates that brain Aβ levels are mainly resulted from neuronal sources while the contribution of astrocytic Aβ levels are minimal. It is noted that neurons have higher levels of BACE1 and APP than glial cells. Most likely, astrocytic BACE1 deletion induced IR signaling and increased CLU levels, which contribute to the enhanced glial Aβ clearance.

Astrocyte specific Bace-1 deletion was achieved by breeding Tg-Gfap-cre, Gfap specific Cre expressing mice (73.12Mvs, JAX stock#012,886, the Jackson Laboratory) with Bace-1 conditional mice (Bace-1 ^fl/fl), carrying loxP-flanked genes as previously described [15]. The resulting Bace1^fl/+;Gfap-cre mice were crossed with Bace1^fl/fl mice to obtain a colony with the following genotype: Bace1^fl/fl;Gfap-cre. Bace1^fl/fl;Gfap-cre mice were maintained by breeding with Bace-1 ^fl/fl. Female Bace1^fl/fl;Gfap-cre mice were also crossed with male 5xFAD;Bace1 ^fl/fl resulting in 5xFAD;Bace1^fl/fl;Gfap-cre mice. Various age groups of 5xFAD;Bace1^fl/fl;Gfap-cre and 5xFAD;Bace1^fl/fl females were used for downstream Western blot and amyloid staining due to reported inconsistencies in male 5xFAD amyloid pathology [34].

Adult conditional knockout Bace1^fl/fl;UBC-CreER and 5xFAD;Bace1^fl/fl;UBC-CreER were bred and maintained as previously described [15]. All lines were routinely backcrossed with C57BL/6 J mice for at least five generations to ensure consistent genetic background for phenotypic analyses. BACE1-null mice were obtained by heterozygous breeding of Bace1^+/- males and females (Jax stock# 004,714, Jackson Lab) as previously described.

All experimental protocols were approved by the Institutional Animal Care and Use Committee of the Lerner Research Institute and University of Connecticut Health Center in compliance with the guidelines established by the Public Health Service Guide for the Care and Use of Laboratory Animals.

Astrocytes were isolated from mouse brains using adult brain dissociation kit (MiltenyiBiotec, cat.no. 130–107-677) for single-RNA sequencing (see below), as well as isolating astrocytes from Bace1^fl/fl;Gfap-cre and Bace1^fl/fl. Briefly, mice were perfused with 25 mL of PBS and forebrains were isolated and transferred to gentleMACS Dissosciator (Milltenyi) to form a single-cell suspension. Subsequently, myelin, cell debris, and erythrocytes were removed and remaining cells were immunolabelled with ACSA2 immunomagnetic beads (cat. No. 130–097-678). The cell suspension was passed though magnetic columns and QuadroMACS separators, which retained ACSA2-positive cells on the column. This ACSA2 immunomagnetic bead MACS method is reported to be less disruptive to glial cell processes and reactivity than microfluidic systems such as FACS [58].

For single cell RNA-sequencing, ACSA2 positive astrocytes from littermate controlled 2-month-old Bace1-null and WT mice. 3 separate samples per group (2 males, 1 female) were submitted for downstream analysis. Astrocytes from 3 pooled brains (2 males, 1 female) from 14-month old 5xFAD;Bace1^{fl/fl;UBC−creER} and 5xFAD;Bace1^fl/fl mice were also submitted for scRNA seq. Astrocytes were checked for cell number and viability. Samples with little debris and > 80% viability were then used for single cell RNA sequencing (University of Connecticut Single Cell Genomics). About 12,000 cells of each sample type were loaded onto 10X Chromium Single Cell RNA sequencing chips. Sample mRNAs were barcoded and subsequently converted into cDNA. After cDNA library quality inspection, libraries were sequenced using an Illumina Novaseq 6000 sequencer. Custom CellRanger pipeline (University of Connecticut Single Cell Genomics) provided initial UMI counts and tSNE clustering. Raw scRNA-seq data will be provided to the NIH Gene expression Omnibus (GEO) for open-source use.

The data analysis was taken into consideration of a published guidance [24]. For secondary analysis, the Seurat version 4 packaged in R (Satija lab) was used to determine differential gene expression using normalized read counts for identifiable genes, reclustering, displaying violin plots of gene expression, and mapping gene expression in cells. For violin and dimension plots of gene expression, scaling excludes outlier cells below 5^th or above 95^th percentile of range.

Quality control parameters were set to only include cells with greater than 1,000 identifiable genes, less than 25,000 read counts, and less than 20% mitochondrial RNA. Filtered samples were then grouped by genotype and read counts were normalized and transformed using SCTransform (Supplementary Figuresand) PCA dimensional reduction was followed by UMAP clustering. Identified reactive astrocyte cluster transcriptomes were then compared with a log normalized scaling factor of 100,000 in order to generate a number of differentially expressed genes, with low overall expression. DEG adjusted p-values are presented as p-values for analysis. 1 2

Primary astrocytes cultures were prepared from littermate controlled perinatal mouse pups resulting from Bace1^+/- X Bace1^+/- breeding. Pups were genotyped at P0 (see below) and then primary mixed glia were prepared from P1-3 as described previously [59, 60]. Briefly, brain tissue was isolated from skulls, bisected, and then incubated with 0.25% trypsin at 37 °C for 15 min, with occasional gentle swirling. Afterwards, tissue solution was gently triturated with fire-polished glass pipettes and passed through a 70 micron filter to form a single-cell suspension. The mixed glia culture was cultured in a T75 flask in DMEM-F12 containing 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, and 1% penicillin/streptomycin (Life Technologies) for 12–14 days. Mixed glia cultures were split twice with 4–5 day intervals to achieve a pure mature astrocyte culture.

Astrocytes were seeded 1 × 10⁶ onto 6 well plates for Western blot purpose, and 1 × 10⁴ for 8 well chamber sides for confocal staining. Media was switched to DMEM-F12 medium containing 1% serum-free G5 astrocyte media supplement (Gibco) for 12–24 hrs before treatment. Cells were maintained and grown in a humidified atmosphere of 5% CO₂ at 37 °C.

Non-tagged Aβ₄₂ peptide (ThermoFisher) and fluorescent Hilyte 555 tagged- Aβ₄₂ (Anaspec) were prepared according to previous reports [28]. Briefly, peptides were solubilized in 0.1% NH₄OH containing 0.01% (w/v) NaN₃, and further re-suspended in DMEMF12 (pH 7.4) and stored at -20 °C. Prior to use, peptides were oligomerized by incubation with constant rotation for 24 hrs at 4 °C. Formation of beta-sheets was detected by Thioflavin-T fluorescence indicating the formation of aggregated Aβ₄₂ (data not shown). Cells were treated with 2 μM of aggregated Aβ₄₂ for 36 hrs for most experiments, except for Western blotting of phosphorylated signaling pathways, in which cells were treated with Aβ₄₂ for 4 hrs or otherwise indicated.

Knockdown of Clu in astrocyte cultures was attained by using Clusterin siRNA reagents (Santa Cruz Biotechnology, catalog no. sc-43689) or control scrambled siRNA (Santa Cruz Biotechnology, catalog no. sc-37007) according to manufacturer’s instructions. Briefly, siRNA plasmids were combined with siRNA transfection medium and duplexed with siRNA transfection media. Astrocyte specific G-5 media was removed and replaced with siRNA duplexed transfection solution for 6 hrs. Afterwards transfection solution was replaced with astrocyte specific G-5 media.

For determining correct dosage, cells were incubated with 10, 20, 40 or 80 pmol of Clu siRNA or 80 pmol of Ctrl siRNA, before switching to astrocyte media. After 72 hrs in astrocyte media, cells were lysed and probed for CLU expression. After optimization, cells were treated with 80 pmol of Clu or Ctrl siRNA before treatment with aggregated Aβ₄₂ either untagged (Western blot) or tagged with HiLyte 555 (Cytochemistry imaging).

Astrocytes from perinatal culture or acutely isolated from adult mice were washed with ice-cold PBS and then were lysed in RIPA lysis buffer containing 50 mM Tris–HCl (pH 7.4), 1 mM EDTA, 100 mM NaCl, 0.1% SDS, 1 mM PMSF, 1 mM sodium orthovanadate, 1 μg/ml leupeptin, 1 μg/ml pepstatin, and 10 μg/ml aprotinin, for 10–30 min at 4 °C with rotation. The lysate was collected and further bath sonicated on ice for 30 sec on and off cycles at 60 Hz for 5 min and then centrifuged at 15,000 × g for 30 min at 4 °C. Total brain lysates were sonicated by drill sonicator in RIPA buffer before centrifugation. Protein concentrations were determined using a BCA assay kit (Pierce). Equal amounts of protein from each sample were loaded and electrophoretically resolved on 4%–12% SDS-PAGE (NuPAGE system, Life Technologies) gels with MOPS or MES running buffer (Thermofisher). After electrophoresis, proteins were transferred to nitrocellulose membranes at 100 V for 2 hrs. The membranes were washed with PBS with 1:2,000 20% Tween-20 and blocked with 1% bovine serum albumin (BSA) and 0.5% evaporated milk for 1 hr at room temperature. The membranes were probed with primary antibody (see table for antibody list), followed by incubation with appropriate secondary HRP-conjugated antibody (1:2,000). The antibody-bound proteins were detected by an iBright 1500 imaging system (Invitrogen). To ensure equal loading, the blots were reprobed with monoclonal anti-actin (1:10,000) or calnexin (1:1,000). For quantification purposes, band intensities of immunoblots were analyzed using imageJ software. For phosphorylated antibodies, membranes were washed and blocked in Tris-Buffered Saline with 1:2,000 of 20% Tween-20 and blocked in 5% BSA.

After treatments with aggregrated HiLyte Fluor-555-labeled Aβ₄₂ (Anaspec, Fremont, CA), astrocytes were washed with 3 times with PBS for 5 min, before fixation with 4% formaldehyde for 10 min at room temperature, and permeabilization with 0.1% TritonX-100 for 5 min. After careful washing, cells were incubated with 100 nM of Alexaflour 488 phalloidin (Thermofisher A12379) and Topro3 (Thermofisher T3605) diluted in 1% BSA/PBS for 10 min at room temperature. Cells were washed then mounted on coverslips with 90% glycerol, 20 mM Tris–HCL (1 M, pH 8.8), and 0.5% (w/v) p-Phenylenediamine (CSHB Protocol). Images were captured with either a Leica TCS-SP8-AOBS or Zeiss LSM 800 Confocal microscope.

Images were quantified with ImageJ software. Integrated density of HiLyte Fluor-555-labeled Aβ₄₂ fluorescence intensity with the bounds of a mask based upon phalloidin defined maximum threshold was measured and normalized to phalloidin mask area and nuclei marked by ToPro3.

Quantification of amyloid plaques was conducted with serial sagittal sections, which were selected at 10-section intervals. Amyloid plaques were labeled with 0.05% Thioflavin-S. Images were captured by BZ-X800 microscope (Keyence) using a 2 × objective and stitched together with the Keyence stitching module. Total plaque numbers in the cerebral cortex and hippocampus (including subiculum) were counted using ImageJ software (National Institutes of Health). Seven female 5xFAD;Bace1^fl/fl ;Gfap-cre and six female 5xFAD;Bace1^fl/fl mice in each age group were used.

Bar graph results are expressed as mean ± SD. Fold change values and p-values for scRNA seq comparison were based on Wilcoxon Ranked Sum test generated by Seurat. The statistical analyses were performed using GraphPad Prism 6.0 software (GraphPad Software, San Diego). Student’s t-tests were used to compare between 2 groups. Two-way ANOVAs were used to compare multiple groups, post-test of Sidak, Bonferonni, or Tukey was used to compare between different groups. Differences with *p < 0.05, **p < 0.01, ***p < 0.001 were considered significant. (Table 1).

All experimental protocols were approved by the Institutional Animal Care and Use Committee of the University of Connecticut School of Medicine in compliance with the guidelines established by the Public Health Service Guide for the Care and Use of Laboratory Animals.

All authors have read and approved the final manuscript.

All authors declare no conflict of interests.

All original data presented in the paper will be made available for reviews when needed. Research materials will be also made available when it is required. The GEO entry for the scRNA seq data: GSE230116↗.