Contribution of amyloid deposition from oligodendrocytes in a mouse model of Alzheimer’s disease

We crossed Bace1 conditional knockout (Bace1^fl/fl) mice [22] with Olig2-Cre mice (JAX:025567) [65] to obtain Bace1^fl/fl;Olig2-Cre mice. We also crossed Bace1^fl/fl mice with heterozygous App^{NL−G−F/wt} knock-in AD mice (RIKEN Center for Brain Science, Japan) to generate Bace1^fl/fl;App^{NL−G−F/wt} mice. Finally, we crossed Bace1^fl/fl;Olig2-Cre with Bace1^fl/fl; App^{NL−G−F/wt} mice to obtain Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice. All mice were maintained on a C57/Bl6J background and housed on a 12-hour light/12-hour dark cycle with access to food and water ad libitum and both sexes were used. All animal use and procedures were performed according to the Institutional Animal Care and Use protocols at UConn Health, Farmington, and in compliance with the guidelines established by the Guide for the Care and Use of Laboratory Animals, as adopted by the National Institutes of Health.

Mice brains were surgically removed and cut mid-sagittally into equal halves. One half of the brain was fixed in 4% paraformaldehyde for 24 h and then immersed in 20% sucrose overnight at 4 °C and then embedded with optimal cutting temperature compound (OCT). The other half was used for western and ELISA analysis. Brains were sectioned sagittal into 16 μm-thick sections on a cryostat microtome (Thermo HM525 NX). Sections on slides were washed in PBS 3× for 5 min to remove OCT and then permeabilized with 0.03% H₂O₂/0.3% Triton X-100 in PBS for 30 min, followed by washing with PBS (3× for 5 min). Antigen retrieval was performed by microwaving the sections in 0.05 M citrate-buffered saline (pH 6.0) for 3 min. The sections were blocked with 5% normal goat serum and incubated with the primary antibody 6E10 (1:1000 dilution, AB_2564652, BioLegend). After washing with PBS (3× for 5 min), the sections were incubated with universal biotinylated anti-mouse/rabbit IgG (1:200, Vector Laboratories) at room temperature for 2 h. Washing with PBS (3× for 5 min), the sections were incubated with avidin-biotin peroxidase complex (1:200, Vector Laboratories) at room temperature for 1 h and developed with 0.05% DAB (Sigma) with 0.01% H₂O₂ in PBS for 5 min. Then, the slides were mounted on a coverslip with 60% glycerol followed by standard immunohistochemical staining procedures. Antibodies for amyloid plaques and glial cells are marked by 6E10 (1:1000, Covance Research Products Inc Cat# SIG-39330-200, RRID: AB_662804), IBA1 (1:1000, AB_839504, Fuji Wako Chemical USA), GFAP (1:1000, AB_2631098, Cell Signaling). Dystrophic neurites are marked by RTN3 R458 antibody [25].

O4⁺ immature and mature oligodendrocytes were isolated from the forebrain in Postnatal day 13 (P13) pups (5 pups per group in Bace1^fl/fl, Bace1^fl/fl;Olig2-Cre, Bace1^fl/fl; App^{NL−G−F/wt}, and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}) using anti-O4 microbeads (MiltenyiBiotec, catalog no. 130-094-543) and adult brain dissociation kits (MiltenyiBiotec, catalog no. 130-107-677). Briefly, forebrains were dissociated into a single-cell suspension; myelin, cell debris, and erythrocytes were removed subsequently; and cells were immunolabeled with anti-O4 microbeads. The cell suspension was allowed to pass though the magnetic column and retained O4⁺ cells from the column, which were flushed out and washed with PBS for western blotting.

O4⁺ cells and brain tissues were homogenized in radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris–HCl, pH 7.4, 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM NaF, 1 mM Na₃VO₄, and a protease inhibitor cocktail [Roche]) and centrifuged at 13,200 rpm for 90 min. Protein concentrations were determined using a bicinchoninic acid (BCA) assay kit. Equal amounts of protein were loaded and resolved on 4 to 12% SDS-polyacrylamide gels (NuPAGE system, Life Technologies). Subsequently, blots were transferred to nitrocellulose membranes at 100 V for 2 h. The membranes were blocked with 5% BSA for 1 h at RT. The membranes were probed with the following primary antibodies at the noted dilutions: 1:1,000 APP-C (AB_258409, Sigma); 1:1,000 BACE1(gift from Huaibin Cai, NIH); 1:500 Olig2 (AB_2157541, Proteintech); 1:1,000 MBP (AB_2799920, Cell Signaling); 1:5,000 PLP (gift from Bruce Trapp, Cleveland Clinic);1:1000 NeuN (AB_2298772, Millipore); 1:50,000 actin (AB_476744, Sigma); 1:5,00 ADAM10 (AB_2242320, Millipore) Aβ42 (1:1000, AB_2798671, Cell Signaling). After 24 h primary incubation at 4^oC, blots were washed extensively, incubated with HRP-conjugated secondary antibodies, and visualized using enhanced chemiluminescence (Thermo Scientific). The antibody-bound protein blots were detected by an iBright 1500 imaging system (Invitrogen). For quantification purposes, band intensities of immunoblots were analyzed using ImageJ software (National Institutes of Health). Original blot images can be found in the supplemental file.

In situ hybridization was performed using the RNAscope Multiplex Fluorescent v2 kit (ACD Bio, Cat. No. 323100) and the kit-described procedures. Briefly, fixed-frozen brains were sectioned sagittal into 16 μm-thick sections on a cryostat microtome (Thermo HM525 NX). The sections were post-fixed in 4% paraformaldehyde and dehydrated in 50%, 70%, and 100% ethanol. After treatment with hydrogen peroxide and the target retrieval, sections were digested with Protease III for 30 min at 40 °C. After washing with wash buffer, sections were hybridized for 2 h at 40 °C with the following probes: mBace1-C2, mSyp-C3 (neuronal marker), and mMbp-C3 (oligodendrocyte marker). The probes were then amplified by sequentially incubating with the kit reagents AMP1, AMP2, and AMP3. Finally, the sections were developed by sequentially incubating with an appropriate HRP linked to the specific probe channel, fluorophore dye, and HRP blocker. Images were captured using Zeiss LSM800 confocal microscopy.

Serial sagittal brain sections starting from the beginning of the hippocampus were selected at 10-section intervals. Sections were probed with Aβ monoclonal antibody 6E10, which recognizes the first 16 residues of Aβ, and stained with DAB as described above. Images were captured with a Keyence fluorescence microscope (Keyence, BZ-X810). Plaque counting in the cortex and hippocampus was conducted using ImageJ software (National Institutes of Health).

Insoluble Aβ_1–42 was prepared from frozen hippocampal tissues by the guanidine hydrochloride method [22]. Levels of insoluble Aβ_1–42 and Aβ_1–40 were quantified by the human Aβ₄₂ ultrasensitive ELISA kit (ThermoFisher, cat # KHB3544) and human Aβ₄₀ ELISA kit (ThermoFisher, cat # KHB3481) according to kit instructions. Results were obtained from 12-month-old male (6 Bace1^fl/fl;App^{NL−G−F/wt} and 8 Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}) and female (3 Bace1^fl/fl;App^{NL−G−F/wt} and 6 Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}) mice. Numbers and size of GFAP⁺-reactive astrocytes and IBA1⁺-microglia in surrounding amyloid plaques were chosen for quantification by utilizing ImageJ software (NIH).

Myelin sheath thickness was examined according to the procedure described in our previous study [24]. The myelinated axon circumference was measured by digitally tracing the inner and outer layers of the myelinated fiber using Photoshop CS6 with measurement tools. (Adobe). The g-ratio was calculated as the inner-to-outer diameter of a myelinated axon.

LTP recordings on hippocampal slices were performed according to previously described procedures using a MED-A64HE1S head amplifier and a MED-A64MD1 main amplifier, run by Mobius software [9, 22]. Upon obtaining horizontal hippocampal slices from three female 12-month-old Bace1^fl/fl and Bace1^fl/fl;Olig2-Cre mice, the prepared slices were then placed onto the center of an MED probe (MED-P515A; AutoMate Scientific) with continuous perfusion of artificial cerebrospinal fluid consisting of (in mM): 115 NaCl, 2 KCl, 1.25 KH2PO4, 1.0 MgSO4, 2.0 CaCl2, 26 NaHCO3, 10 glucose, and 1.0 L-Ascorbic acid and bubbling of 95% O₂/5% CO₂. The device has an array arranged in an 8 × 8 pattern of 64 planar microelectrodes across a hippocampal slice. Each electrode used for data acquisition and analysis was 20 μm × 20 μm with an interelectrode distance of 150 μm. Schaffer collateral-to-CA1 synapses were analyzed for LTP assays. Field excitatory postsynaptic potentials (fEPSPs) caused by theta burst stimulation were recorded at a 20-kHz sampling rate within the CA1 subregion of the hippocampus. Control fEPSPs were recorded for at least 10 min before the conditioning stimulation using a response ~ 50% of the maximum. After a stable baseline was established, LTP was induced with three trains of 100 Hz for 1 s with an intertrain interval of 20 s. Field potential amplitudes were then measured. Synaptic strength was evaluated by measuring changes in the fEPSP amplitude relative to baseline.

Hippocampi and cortices were extracted from 8 female mice at the age of ~ 5 months, with 2 mice each in Bace1^fl/fl, Bace1^fl/fl;Olig2-Cre, Bace1^fl/fl;App^{NL−G−F/wt}, and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}, and were snap-frozen on dry ice. Tissues were then homogenized and lysed in buffer containing 0.01% tween/0.01% NP-40 on ice. A debris removal (Miltenyi Biotec, Germany) step was included to remove myelin debris, and nuclei were washed and suspended in PBS containing 0.04% BSA and RNase inhibitor (Invitrogen) and immediately processed as follows. Nuclei count and viability were assessed on a LUNA FX7 automated cell counter (Logos Biosystems), and up to 32,000 nuclei from each suspension were loaded onto one lane of a 10x Genomic Chip M. Single cell capture, barcoding and library preparation were performed using the 10x Genomics Chromium X platform [66] version 3.1 NEXTGEM chemistry and according to the manufacturer’s protocol (#CG000416↗). cDNA and libraries were checked for quality by Tapestation 4200 (Agilent) and Qubit Fluorometer (ThermoFisher), quantified by KAPA qPCR, and sequenced on an Illumina NovaSeq X Plus 100 cycle 10B flow cell, with a 28-10-10-90 asymmetric read configuration, targeting 12,000 barcoded nuclei with an average sequencing depth of 70,000 read pairs per nucleus. Illumina base call files for all libraries were converted to FASTQs using bcl2fastq v2.20.0.422 (Illumina) and FASTQ files associated with the gene expression libraries were aligned to the GRCm38 reference assembly with vM23 annotations from GENCODE (10x Genomics mm10 reference 2020-A) using the version 7.2.0 CellRanger count pipeline (10x Genomics), producing a digital cell by gene counts matrix corresponding to each input suspension.

Data were then analyzed in an R (v4.3.0) environment using Seurat v4.4.0. We created Seurat objects for each library after removing nuclei with less than 200 features and features occurring in fewer than three nuclei. The raw counts matrix was filtered using cutoff values of mitochondrial transcripts below 5% and between 100 and 7500 unique features. The expression profiles of each cell using the 2000 most variable genes as measured by dispersion [49, 66] were used for neighborhood graph generation and dimensionality reduction with UMAP [4, 58]. Clustering with Louvain algorithm, cell type annotation, and differential expression was performed ad hoc on a per-cluster basis using the Seurat v4.0 R toolkit [20]. Cell types in each cluster were assigned with the marker genes, excitatory neurons (Slc17a7), inhibitory neurons (Gad2), astrocytes (Aqp4, Clu), microglia (Cx3cr1, Hexb), endothelial (Cldn5, Vtn), macrophage (Marc1), pericytes (Atp13a5), ependymal (Kl), OPC (Pdgfra) and OLs (Mog) (Supplemental Figure S1).

Downstream analysis of single-nuclei RNA sequencing data was performed using Scanpy (v1.9.3). Filtered data output from cellranger from eight samples were read into Scanpy, annotated, and concatenated into a single AnnData object. Quality control involved filtering out cells with fewer than 50 genes and mitochondrial content above 5%. Genes expressed in fewer than 5 cells were also removed, followed by normalization and log transformation. Doublets were identified using Scrublet with a batch-specific threshold of 0.15 and removed. Highly variable genes were selected with specific cutoffs for mean and dispersion. Principal component analysis (PCA) was performed using these genes, and batch effects were corrected using Harmony [29]. Dimensionality reduction was achieved with UMAP, and clustering was performed using the Leiden algorithm at a resolution of 0.8. Differential gene expression tables were generated using rank_genes_groups function in scanpy using Wilcoxon test.

Gene ontology term enrichment was performed using the R package clusterProfiler 4.6.2 [62]. The mapIds() function in the org.Mm.eg.db package was used to convert gene symbols to Ensmbl IDs. The functions enrichGO from clusterProfiler were then used to enrich for gene ontology terms from GO databases (biological process). A p-value cutoff of 0.05 was used to determine significant GO terms.

All results are expressed as means ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 6.0 software (GraphPad Software, San Diego). Welch’s t tests were used in the case of a significant F-test. Two-tailed, unpaired Student’s t tests were used for all other comparisons. Differences with *p < 0.05, **p < 0.01, and ***p < 0.001 were considered significant.

Global and neuron-specific Bace1 deletion models exhibit significant synaptic side effects and deficits in synaptogenesis and maturation of oligodendrocytes [5]. However, the impact of Bace1 deletion specifically in oligodendrocytes and how it impacts their cellular function is unknown. To evaluate this, we crossed Olig2-Cre mice [65], JAX:025567) with Bace1 conditional knockout (Bace1^fl/fl) mice [22] to obtain Bace1^fl/fl;Olig2-Cre mice. To evaluate deletion of Bace1 in oligodendrocytes, we performed RNA in situ hybridization fluorescent assays to examine Bace1 expression in Bace1^fl/fl control and Bace1^fl/fl;Olig2-Cre mouse brains. As shown, Bace1 was highly expressed in both Syn⁺ neurons and Mbp⁺ oligodendrocytes in the cerebral cortex of 4-month-old Bace1^fl/fl control mice (Fig. 1A). In Bace1^fl/fl;Olig2-Cre brains, Bace1 was barely detectable in oligodendrocytes, while neuronal Bace1 in both Bace1^fl/fl;Olig2-Cre and Bace1^fl/fl mice was comparable (Fig. 1A).

Since neuronal BACE1 was not obviously affected, BACE1 protein levels in Bace1^fl/fl;Olig2-Cre hippocampal tissues, which contained BACE1 from all hippocampal cells, were not discernibly altered compared to that in Bace1^fl/fl controls (Fig. 1B-C). Consistently, full-length APP (APP-fl) levels were not altered. For a more specific comparison of proteins levels in oligodendrocytes, we isolated O4⁺ immature and mature oligodendrocytes from postnatal day 13 (P13) mouse forebrains. BACE1 was significantly reduced while APP-fl was visibly increased in Bace1^fl/fl;Olig2-Cre mice compared to Bace1^fl/fl controls (Fig. 1D-E), indicating abrogated cleavage of APP. Interestingly, α secretase-cleaved APP cleavage product, CTF-83, was not different between the groups (Fig. 1D-E). Since APP in WT mice is predominantly cleaved by α secretase, a small increase in CTF83 was likely not notable. We also found OLIG2 levels were reduced by ~ 50% in Bace1^fl/fl;Olig2-Cre compared to Bace1^fl/fl, likely because of one copy of Cre recombinase being inserted into the Olig2 gene, leading to the reduction of OLIG2 levels [65].

Since genes in amyloidogenic pathways are expressed in oligodendrocytes, we asked whether Bace1 deletion in oligodendrocytes would affect amyloid deposition in AD mouse brains. To this end, we crossed Bace1^fl/fl;Olig2-Cre mice with Bace1^fl/fl;App^{NL−G−F/wt} mice [47] to generate an AD mouse line lacking oligodendrocyte Bace1 (Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}). O4⁺ immature and mature oligodendrocytes were similarly isolated from P13 mouse forebrains to evaluate APP processing with or without Bace1 deletion. We showed that BACE1 levels were significantly reduced in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} cells and correlated with decreased the levels of CTF-99, the BACE1-cleaved APP cleavage product, in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} cells, while CTF83 levels were higher (Fig. 2A-B). OLIG2 levels were reduced by ~ 50% in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} compared to Bace1^fl/fl;App^{NL−G−F/wt} controls.

Next, we stained amyloid plaques in 12-month-old Bace1^fl/fl;App^{NL−G−F/wt} and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice. We found a visible reduction of plaque deposition in both the cerebral cortex and hippocampus of Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice (Fig. 2C). In Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice, we noted that the plaque number was reduced by ~ 29% in the cerebral cortex (Figs. 2D and 14.27 ± 2.79 plaques per mm² in Bace1^fl/fl;App^{NL−G−F/wt}, N = 9, vs. 10.14 ± 1.81 in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}, N = 15, P < 0.001), and by ~ 33% in the hippocampus (Figs. 2D and 10.74 ± 2.42 in Bace1^fl/fl;App^{NL−G−F/wt} vs. 7.17 ± 1.61 in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}, P < 0.001). In line with plaque reduction, both insoluble Aβ_1–40 (Figs. 2E and 5092.78 ± 1360.73 tissue in Bace1^fl/fl;App^{NL−G−F/wt}, N = 9 vs. 3563.66 ± 631.04 pg/g tissue in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}, N = 14, P < 0.01) and Aβ_1–42 levels were reduced by ~ 30% in hippocampal tissues of 12-month-old Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice when compared to age-matched Bace1^fl/fl;App^{NL−G−F/wt} littermates (Figs. 2F and 1009.22 ± 269.19 ng/g tissue in Bace1^fl/fl;App^{NL−G−F/wt}, N = 9, vs. 704.19 ± 88.66 ng/g tissue in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}, N = 14, P < 0.01). The ratio of Aβ_1–42 over Aβ_1–40 was not significantly altered (Figs. 2G and 199.15 ± 24.33 in Bace1^fl/fl;App^{NL−G−F/wt} vs. 207.23 ± 20.14 in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}; P = 0.3962). There was no detectable gender effect on Aβ production or amyloid plaque load in both Bace1^fl/fl;App^{NL−G−F/wt} and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mouse brains (Supplemental Fig. 1A-B).

One intriguing observation was that a higher number of astrocytes were found to surround neuritic plaques in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} compared to that in Bace1^fl/fl;App^{NL−G−F/wt} mice (Fig. 3A; quantified in 3B: 3.21 ± 0.67 vs. 2.47 ± 0.52 per plaque, *P < 0.05). The overall astrocyte size near neuritic plaques was also larger (Figs. 3C and 416.32 ± 103.90 µm² in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} vs. 301.96 ± 97.87 µm² in Bace1^fl/fl;App^{NL−G−F/wt} mouse brain sections; N = 79 plaques in Bace1^fl/fl;App^{NL−G−F/wt} and 80 in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} samples, P = 0.058). Similarly, the number of microglia in surrounding neuritic plaque was also higher in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mouse brains (4.87 ± 0.73) compared to that in Bace1^fl/fl;App^{NL−G−F/wt} (4.06 ± 0.41) (Fig. 3D-E, P = 0.032). However, the size of microglia appeared to not differ (Figs. 3F and 82.17 ± 19.80 µm² in Bace1^fl/fl;App^{NL−G−F/wt} vs. 85.14 ± 14.63 µm² in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} brain sections; N = 79 plaques in Bace1^fl/fl;App^{NL−G−F/wt} vs. 80 in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt}; P = 0.752). The number of dystrophic neurites, labeled by RTN3 antibody as previously discussed [50], was slightly less but not statistically significant (Supplemental Figure S2).

Since plaque load was significantly reduced in the cerebral cortex and hippocampus in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice, we determined the molecular changes transcriptionally occurring in oligodendrocytes with Bace1 deletion in the App^{NL−G−F/wt} mice. By performing unbiased snRNA-Seq using nuclei samples isolated from the cortex and hippocampus of ~ 5-month-old female mice from four groups (Bace1^fl/fl, Bace1^fl/fl;Olig2-Cre, Bace1^fl/fl;App^{NL−G−F/wt}, and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice), we were able to divide the nuclei into 37 clusters based on cell types clustered via the marker genes such as excitatory neurons (Slc17a7), inhibitory neurons (Gad2), astrocytes (Aqp4, Clu), microglia (Cx3cr1, Hexb), endothelial (Cldn5, Vtn), macrophage (Marc1), pericytes (Atp13a5), ependymal (Kl), OPC (Pdgfra) and OLs (Mog) (Supplemental Figure S3A-D). UMAP visualization of cellular populations in each genotype group did not reveal an obvious change when all cellular populations were included (Supplemental Figure S3E). We then focused only on oligodendrocytes as Bace1 was deleted mainly in this cell population. We identified the differentially expressed genes in oligodendrocytes between: (1) Bace1^fl/fl;Olig2-Cre mice and Bace1^fl/fl mice, (2) Bace1^fl/fl;App^{NL−G−F/wt} mice and Bace1^fl/fl mice, (3) Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice and Bace1^fl/fl;App^{NL−G−F/wt} mice, and (4) Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice and Bace1^fl/fl mice (see volcano plots in Fig. 4A and B; Supplemental Table 1). When Bace1 was deleted in oligodendrocytes, a modest increased expression of myelin genes such as Mbp, Plp1, Mog, Mal, Sirt2 (marked in red in Fig. 4A and B) were commonly seen regardless of whether it was in WT (Bace1^fl/fl;Olig2-Cre vs. Bace1^fl/fl) or AD (Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} vs. Bace1^fl/fl;App^{NL−G−F/wt}) mouse background, indicating that BACE1 regulates expression of these myelin genes. Elevation of these myelin genes was also seen in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} vs. Bace1^fl/fl mice (Supplemental Figure S4C). The differentially expressed genes in oligodendrocytes were further analyzed by Gene Ontology (GO) gene set enrichment analysis (GSEA) (Fig. 4C and D; Supplemental Figure S4B, D). The GO enrichment analysis showed that pathways for axon ensheathment, gliogenesis, and oligodendrocyte differentiation were upregulated in oligodendrocytes with Bace1 deletion irrespective of presence of App^NL−G−F gene (Fig. 4C and D, Supplemental Figure S4B, 4D, Supplemental Table 1). Interestingly, pathways associated with synapse organization, axonogenesis, regulation of synapse organization, regulation of synapse structure or activity were reduced in the oligodendrocytes from both Bace1^fl/fl;Olig2-Cre mice and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice when compared to Bace1^fl/fl mice (Fig. 4C and D, Supplemental Figure S4B, D, Supplemental Table 1).

Strikingly, several genes involved in Aβ production or clearance such as Adam10, Ano4, ApoE, Il33, Sort1, and Sort1 were increased in the oligodendrocytes of Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice when compared to the other three groups of mice (Fig. 4E). Since ADAM10 can function as α-secretase to mediate production of CTF83, we prepared the cortex lysate from 1-year-old Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice and Bace1^fl/fl;App^{NL−G−F/wt} mice and performed the immunoblot of ADAM10. We observed a significant increase of pro-ADAM10 levels in Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice compared to Bace1^fl/fl;App^{NL−G−F/wt} mice while mature ADAM10 was slightly elevated (Fig. 4F and G). The slight increase in ADAM10 protein levels may also contribute to the reduction of Aβ in the Bace1-deleted APP knock-in mice.

Recent single-cell analysis identified four subclusters of oligodendrocytes in AD mice [28]. We pooled a total 9795 nuclei of oligodendrocytes to perform sub-clustering of oligodendrocyte nuclei and identified 4 subclusters based on specific marker genes (Fig. 5A-C). Interestingly, Bace1 deletion or APP mutations in the knock-in mice did not cause significant changes in the cellular composition or proportion in each cluster (Fig. 5B). Compared to other clusters, cells in the cluster 1 express higher levels of genes involved in Aβ production or clearance or critical components of the gamma-secretase complex. This list of genes included ADAM10, Apoe, Il33, Bace1, App, Ncstn, Psen1, Psen2, and Aph1a (Fig. 5D), suggesting that cells in this cluster are likely the major source of oligodendrocyte Aβ.

Since global germline or neuronal Bace1 deletion causes hypomyelination in the PNS [24, 60] and impaired maturation of oligodendrocytes [5, 24], we examined whether the loss of Bace1 from oligodendrocytes would affect levels of myelin proteins and the thickness of myelin sheath. The above snRNA-Seq results in Fig. 4A and B suggested an elevation of myelin gene expression. We isolated hippocampal protein lysates from 4-month-old Bace1^fl/fland Bace1^fl/fl;Olig2-Cre mice and performed immunoblot experiments. We showed that MBP and PLP levels were not significantly elevated in Bace1^fl/fl;Olig2-Cre mice compared to controls (Fig. 6A, B). Consistent with this, we found no significant change in myelin sheath thickness in corpus callosum axons from Bace1^fl/fl, Bace1^fl/fl;Olig2-Cre, Bace1^fl/fl;App^{NL−G−F/wt}, and Bace1^fl/fl;Olig2-Cre; App^{NL−G−F/wt} mice at 4 months of age by EM analysis (Fig. 6C, g-ratio comparisons plotted in 6D-E). However, we found that the axons in the optic nerves were wrapped by thinner myelin sheaths in the Bace1^fl/fl;Olig2-Cre mice compared with Bace1^fl/fl mice (Fig. 6G). Morphometric quantification of myelin thickness by g-ratio analysis (ratio of individual axon diameters to myelinated fiber diameters) in the optic nerve confirmed the relative decrease in myelin thickness (higher g-ratio shown in Fig. 6G). When Bace1^fl/fl;Olig2-Cre mice were compared with littermate controls, the most significant differences were observed in axons ranging between 0.5 and 1.0 μm [average g-ratios: 0.71 ± 0.007 in Bace1^fl/fl (N = 69 axons) vs. 0.74 ± 0.005 in Bace1^fl/fl;Olig2-Cre (N = 65 axons); P = 5.3 × 10^{− 7}], but not the other size groups. However, this significance could not be calculated if only 2 mice per genotype group were used for analysis (Fig. 6H, N = 2). Future study will be needed for further validation. No discernible axonal degeneration was observed in the optic nerves and corpus callosum in all genotypes.

Deficits in synaptic plasticity have previously been reported in global and neuron-specific Bace1 knockout models or BACE1 inhibition [9, 32, 37, 41]; Zhu et al., [22]). Our sequencing results showed reduction of genes in oligodendrocytes involved in synapse organization (Fig. 4C, D). To identify potential changes in synaptic strength associated with oligodendrocyte Bace1 deletion, we examined activity-dependent synaptic plasticity in the hippocampus of 12-month-old Bace1^fl/fl;Olig2-Cre mice and Bace1^fl/fl littermate controls using field potential recordings. Specifically, we measured field excitatory postsynaptic potentials (fEPSPs) from Schaffer collateral axons projecting to CA1 apical dendrites before and after the induction of LTP (Fig. 7A). The LTP magnitude was marginally decreased in Bace1^fl/fl;Olig2-Cre mice when compared to Bace1^fl/fl littermates (Figs. 7B-C and 156.21 ± 7.26% in Bace1^fl/fl;Olig2-Cre vs.. 160.8 ± 8.09% in Bace1^fl/fl, P > 0.05). Hence, synaptic strength is unlikely affected if Bace1 is only deleted in oligodendrocytes.

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.