Transcriptomic Profiling Reveals Microglia-Mediated Inflammation in the Corpus Callosum of a Transgenic Mouse Model of Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD) is a widespread neurodegenerative disorder characterized by progressive cognitive decline. While the cerebral cortex and hippocampus have been the primary focus of AD research, accumulating evidence suggests that white matter lesions in the brain, particularly in the corpus callosum, play a crucial role in the pathogenesis of the disease. In this study, we employed RNA sequencing to examine gene expression changes in the corpus callosum of 5xFAD transgenic AD mice that exhibit cognitive impairment and demyelination. Transcriptomic analysis revealed the upregulation of genes associated with microglia in the corpus callosum of AD model mice. Conversely, genes related to chaperone function were notably downregulated in the corpus callosum of AD model mice. Furthermore, we observed a notable downregulation of clock-controlled genes, indicating circadian dysfunction in this white matter region. These findings suggest that microglia-mediated neuroinflammation, disruption of chaperone function, and circadian dysfunction are implicated in the pathogenesis of white matter lesions in AD. This study provides valuable insights into the potential therapeutic targets for AD, highlighting the importance of addressing white matter pathology and circadian rhythm disturbances in AD treatment strategies.

Introduction:

Alzheimer’s disease (AD) is the most common neurodegenerative disease occurring especially in the elderly population. AD is characterized by a progressive decline in neurological functions, including memory and non-memory cognition, and accounts for an estimated 60-70% of dementia cases [1,2]. While past AD research has clarified its genetic, molecular, and cellular aspects, much remains to be understood in the pursuit of a definitive cure [2]. Highlighting AD’s impact, it presents not only financial problems, but also substantial human costs to countries, societies, families, and individuals [1]. With the global population over 65 years set to triple in the next 30 years, understanding AD pathology more deeply is urgently required to discover the effective therapeutic targets to address its socioeconomic burden [3].

Gray matter, specifically the cerebral cortex and hippocampus, has been considered the major lesions of AD pathology [4]. However, growing evidence suggests that white matter is also vulnerable in AD patients and thus a potential novel therapeutic target [5–8]. Myelin damage was clinically indicated and (micro-) structural and metabolic changes in the white matter were shown in AD patients via developing imaging technique [9–13]. In preclinical studies, an AD-causing gene mutation in oligodendrocytes (OLG) demonstrated adverse effects associated with glutamate and amyloid-β (Aβ), a contributor protein to senile plaques in AD brain [14]. Moreover, recent research demonstrated crucial evidence that tau, a protein that causes neurofibrillary tangles in human AD, could be also toxic for myelin and OLG lineage cells, and therefore might play negative roles in white matter integrity [15]. Although several insights are demonstrated, at present, there are no therapeutic approaches to overcome AD via white matter treatment, partly because of a lack of understanding of how AD-specific condition changes gene expression in white matter lesion.

Because rodents do not naturally develop the defining neuropathology linked to AD, a variety of approaches have been attempted to examine whether expressing the offending proteins also causes AD-like phenotypes [16]. To date, genetic engineering enabled to mimic AD characteristics in mice brains, and some models are developed as AD mice. Among them, a 5xFAD (harboring five familial AD mutations) transgenic mouse, which was developed in 2006, overexpresses human amyloid precursor protein with three FAD mutations and human presenilin 1 with two FAD mutations specifically in brain neurons and thus is commonly used to understand the changes in AD brain and to develop pharmacological interventions [17,18]. Therefore, we examined the gene expression changes in the corpus callosum, one of the main structures of white matter for neurocognitive functions, in 5xFAD transgenic mice through RNA sequencing (RNA-seq) analyses.

Methods:

Animals -

All experimental procedures followed NIH guidelines and were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee. Male 5xFAD transgenic mice and their wild-type mice, both with a B6SJL background, were purchased from The Jackson Laboratory and were housed in a specific pathogen-free conditioned 12 h light/dark cycle room with free access to food and water throughout the experiment. In this study, four 12-month-old male mice were used per group. At 12 months of age, as expected, 5xFAD transgenic mice exhibited cognitive impairments. Specifically, during the NORT test, unlike the wild-type mice which showed an inclination towards a novel object, the 5xFAD transgenic mice failed to do so (Supplementary Figure S1A). Yet, there was not a significant difference in the total investigation time of the objects (Supplementary Figure S1B). Biochemical analyses via Western blotting also confirmed that 5xFAD transgenic mice had reduced levels of myelin basic protein (MBP) (Supplementary Figure S1C), indicating myelin and oligodendrocyte damage.

Corpus callosum sampling -

Mice were transcardially perfused with ice-cold 0.9% physiological saline followed by decapitation. Brains were removed and cooled in pre-chilled Hanks’ Balanced Salt Solution for 1 min. After the removal of the meninges and the choroid plexus, the cerebrum was sliced into 5 coronal sections using a brain matrix slicer. To minimize the inclusion of tissue outside of the corpus callosum, the thicker parts of the corpus callosum from the 2nd and 3rd slices were isolated with direct visualization using a light microscope. We quickly froze corpus callosum samples in an RNA-free tube using liquid nitrogen. Half of the corpus callosum was used for western blotting (see Supplementary Figure S1C) and the other half for RNA-seq experiments.

RNA extraction -

RNA extraction from the corpus callosum samples was performed using QIAzol^® (QIAGEN, Venlo, Netherlands) following the manufacturer’s instructions. Briefly, sonicated tissue was resuspended in 1 mL of pre-chilled QIAzol, and 0.2 mL of chloroform was added to the lysate. After mixing with a vortex mixer, the tube was centrifuged for 15 min at 12,000g. The supernatant was then transferred to a new tube, and the same amount of propanol was added. After centrifugation for 10 min at 12,000g, the supernatant was aspirated, and 1 mL of 75% ethanol was added for washing. Finally, after centrifuging the tube for 5 minutes at 7,500g, we suspended the pellet in RNase-free water. The amount and purity of purified RNA were measured by NanoDrop Spectrophotometers. The RNA sample was stored at −80°C until use.

RNA-seq analysis -

We processed RNA samples from each group for RNA-seq. Genewiz, Inc. (South Plainfield, NJ, USA) carried out both the library preparation using the rRNA depletion approach (single index) and sequencing on the Illumina HiSeq4000 platform (pair-end; 2 × 150 bp). The raw FASTQ data were aligned to the mm10 reference genome using STAR software (version: 2.7.10a). Quantification of transcripts was achieved with RSEM (version: 1.3.3), which generated count data and transcripts per kilobase million (TPM) values. While most of our analytical steps employed normalized count data, TPM values were used for graphical representations. We conducted our bioinformatics analysis in R (version: 4.3.0), primarily utilizing the DESeq2 package (version: 1.40.1) for differential expression analysis. We set a significance threshold at an adjusted p-value (padj) of less than 0.05. Our sequence data, presented as FASTQ files, have been stored under accession number PRJNA1012900. For the gene ontology (GO) evaluations, we used the Metascape platform [19].

Results:

Evaluation of Gene Expression Profiling in the Corpus Callosum from AD Model Mice using RNA-seq -

We isolated the corpus callosum from 12-month-old wild-type (WT) mice with a B6SJL background, as well as from 5x FAD transgenic (AD model) mice (Figure 1A). We ascertained the purity of the corpus callosum samples by assessing gene expression markers for oligodendrocytes and neurons across different layers, ensuring no contamination from other areas (Figure 1B). Additionally, a PCA plot revealed a clear distinction between the WT and AD model mice groups (Figure 1C).

Figure 1. Evaluation of Corpus Callosum Samples from AD Model Mice:

(A) Schematic representation of corpus callosum (CC) sampling from wild-type (WT) and 5x FAD transgenic (AD model) mice. Four samples were collected from each group. (B) Violin plots illustrating the expression of oligodendrocyte markers (Mbp and Mobp) and cortical neuron markers (Reln in layer I, Rasgrf2 in layers II/III, Pou3f2 in layers II-V, Foxp2 in layer IV). Data from both WT and AD model mice are integrated. (C) PCA plots of WT and AD model mice.

Transcriptome Profiling in the Corpus Callosum of AD Model Mice -

To identify AD-affected genes, we filtered the differentially expressed genes (DEGs) with the criteria: |Fold change| > 1.5, padj < 0.05, and mean Base > 50. Notably, 575 genes exhibited increased expression in the corpus callosum of the AD model mice, while 75 genes showed reduced expression in AD model mice (Figure 2A). Upon examining the top 10 genes with upregulated and downregulated expression (Figure 2B), the list of the most upregulated genes included Cst7 (log2FoldChange = 7.83942, padj = 1.917793e-70), Itgax (log2FoldChange = 7.112489, padj = 1.347471e-69), and Clec7a (log2FoldChange = 5.08063, padj = 1.017015e-55), all of which are associated with disease-associated microglia [20]. Additionally, genes relevant to neurologic disorders, including AD, were identified. Specifically, Ccl3 (a chemokine; log2FoldChange = 6.255937, padj = 1.246517e-31) [21], Ctse (also known as Cathepsin E; log2FoldChange = 5.08467, padj = 1.27059e-48) [22], Lilrb4a (log2FoldChange = 4.947678, padj = 7.636763e-38) [23], Gpnmb (log2FoldChange = 4.862764, padj = 1.12133e-08) [24], Il4i1 (log2FoldChange = 4.646696, padj = 3.452024e-18) [25], and Hcar2 (Gpr109; log2FoldChange = 4.500396, padj = 1.205566e-23) [26] were identified, and these genes have been reported to play roles in immune responses. Among the top 10 downregulated genes were those involved in chaperone function, including the heat shock protein 70kDa (HSP70) family members Hspa1a (log2FoldChange = −2.048041, padj = 2.439622e-06) and Hspa1b (log2FoldChange = −2.322014, padj = 1.48472e-06), as well as neural activity-dependent genes like Arc (log2FoldChange = −1.457951, padj = 0.04504809). Interestingly, the circadian rhythm-associated gene Per2 was also among the downregulated genes (log2FoldChange = −1.242973, padj = 8.11109e-10) (Figure 2B).

Figure 2. Transcriptome Profiling in the Corpus Callosum of AD Model Mice:

(A) Volcano plot of DEGs between WT and AD model mice. The DEGs cutoff was set at padj < 0.05 and | Fold Change | >1.5. (B) Bar graph of the top 10 genes upregulated and downregulated in AD model mice compared to WT.

Microglial Gene Expression Alterations in the Corpus Callosum of AD Model Mice -

GO analysis of the DEGs revealed that the top 10 upregulated genes predominantly fell under immune response-associated GO categories. Notably, these categories included the Tyrobp causal network in microglia (WP3625: logP = −37.20067), cell activation (GO:0001775: logP = −35.82189), leukocyte activation (GO:0045321: logP = −34.81682), Neutrophil degranulation (R-MMU-6798695: logP = −34.48122), positive regulation of immune response (GO:0050778: logP = −33.89490), immune effector process (GO:0002252: logP = −33.12380), innate immune response (GO:0045087: logP = −31.87365), regulation of immune effector process (GO:0002697: logP = −30.29901), and the inflammatory response (GO:0006954: logP = −29.49761) (Figure 3A).

Figure 3. Elevated Microglial Gene Profiling in the Corpus Callosum of AD Model Mice:

(A) Top 10 GO list with DEGs. (B) Classification of DEGs using cellular markers from previously published RNA-seq data [27,28]. (C) Violin plots of microglial markers in WT and AD model mice. (D) MA plots in WT and AD model mice; genes associated with GWAS are shown in blue, DEGs in red, and DEGs highly expressed in microglia and associated with GWAS are shown in green. (E) Bar plot (log2Fold Change) of DEGs highly expressed in microglia and associated with GWAS. (F) Classification of gene lists using public single-cell RNA-seq data [29]. The abbreviations used in this figure are as follows: In the oligodendrocyte lineage, OPC stands for oligodendrocyte precursor cells, OLG for oligodendrocytes, and OEG for olfactory ensheathing glia; in the astrocyte lineage, NSC represents neural stem cells, ARP for astrocyte-restricted precursors, and ASC for astrocytes; the neuronal lineage includes NRP as neuronal-restricted precursors, NEUR_immature for immature neurons, NEUR_mature for mature neurons, and NendC for neuroendocrine cells; ependymal cells have EPC as ependymocytes, HypEPC for hypendymal cells, TNC for tanycytes, and CPC for choroid plexus epithelial cells; vasculature cells comprise EC for endothelial cells, PC for pericytes, Hb-VC for hemoglobin-expressing vascular cells, VSMC for vascular smooth muscle cells, VLMC for vascular and leptomeningeal cells, and ABC for arachnoid barrier cells; and for immune cells, MG indicates microglia, MNC for monocytes, MAC for macrophages, DC for dendritic cells, and NEUT for neutrophils.

When analyzing genes with notable expression in central nervous system cells based on public RNA-seq datasets [27,28], the majority were derived from microglia (Figure 3B). A direct examination using microglia markers further revealed that most of them had increased expression (Figure 3C). By cross-referencing our DEGs with genes implicated in AD from GWAS studies (MONDO_0004975; representing 1125 genes in mice), we identified multiple genes that showed enhanced expression in the corpus callosum of the AD model mice. These genes included Itgax (log2FoldChange = 7.112489, padj = 1.347471e-69), Trem2 (log2FoldChange = 2.917976, padj = 4.742697e-86), Bcl3 (log2FoldChange = 2.416525, padj = 3.107311e-10), Fcer1g (log2FoldChange = 2.32248, padj = 4.488103e-33), Tbxas1 (log2FoldChange = 2.081317, padj = 1.447661e-21), Havcr2 (log2FoldChange = 1.968202, padj = 6.243716e-31), Cd33 (log2FoldChange = 1.639613, padj = 5.156761e-18), Grn (log2FoldChange = 1.437208, padj = 8.186139e-53), Inpp5d (log2FoldChange = 1.247693, padj = 4.597314e-30), Plcg2 (log2FoldChange = 1.209738, padj = 6.237314e-14), and Abi3 (log2FoldChange = 1.05795, padj = 1.817239e-15) (Figure 3E, Supplementary Figure S2). In further detailed analysis using publicly available single-cell RNA-seq data of the adult mouse brain [29], the majority of these genes were enriched in microglia and other immune cells such as monocytes, macrophages, and neutrophils (Figure 3F).

Potential Decrease in Chaperone Function in the Corpus Callosum of AD Model Mice -

In the corpus callosum of AD, downregulated genes included members of the HSP70 family, such as Hspa1a (log2FoldChange = −2.048041, padj = 2.439622e-06) and Hspa1b (log2FoldChange = −2.322014, padj = 1.48472e-06) (Figure 2B). In the GO analysis, top-ranked categories predominantly pertained to chaperone function, such as Regulation of HSF1-mediated heat shock response (R-MMU-3371453: logP = −5.768248), chaperone-mediated protein folding (GO:0061077: logP = −5.698024), chaperone cofactor-dependent protein refolding (GO:0051085: logP = −5.446003), Cellular response to heat stress (R-MMU-3371556: logP = −5.209638), ‘de novo’ post-translational protein folding (GO:0051084: logP = −5.187357), ‘de novo’ protein folding (GO:0006458: logP = −5.140127), protein refolding (GO:0042026: logP = −4.734230), and HSF1-dependent transactivation (R-MMU-3371571: logP = −4.301036) (Figure 3A). Within these GO categories, apart from Hspa1a and Hspa1b, genes such as Hspa5 (log2FoldChange = −0.6078, padj = 5.542026e-09), Hspb1 (log2FoldChange = −0.5929298, padj = 0.005106073), Dnajb1 (log2FoldChange = −0.7465945, padj = 1.05478e-06), and Nup43 (log2FoldChange = −0.73581, padj = 0.02015798) also demonstrated decreased expression in AD model mice (Figure 4).

Figure 4. Downregulation of Genes Related to Chaperone Function in AD Model Mice:

Bar plots (with dot plots) comparing gene expression between WT and AD model mice. Error bars indicate SEM.

Potential Disruption of Circadian Rhythm Functions in the Corpus Callosum of AD Model Mice -

Notably, circadian rhythm-involved genes like Per2 (log2FoldChange = −1.242973, padj = 8.11109e-10) were among the top downregulated genes (Figure 2B). Some of these downregulated DEGs (Cry1, Per1, Per2, Vwf) are regulated by genes such as Clock and Arntl (Bmal1) (Figure 5A). GO analysis revealed top-ranked categories associated with the entrainment of the circadian clock, such as by photoperiod (GO:0043153: logP = −4.129897) (Figure 3A). Other categories included entrainment of the circadian clock (GO:0009649: logP = −4.027543) and Circadian rhythm – Mus musculus (house mouse) (mmu04710: logP = −3.723830) in GSEA plots in AD (Figure 5B). Upon examining genes associated with circadian rhythms in the GO, we found genes such as Cry1 (log2FoldChange = −0.6972396, padj = 0.0005563706), Per1 (log2FoldChange = −0.7284916, padj = 0.003697868), Per2 (log2FoldChange = −1.242973, padj = 8.11109e-10), and Vwf (log2FoldChange = −0.6916326, padj = 4.964108e-08) included (Figure 5C). However, genes such as Dbp (log2FoldChange = −0.5295548, padj = 0.03860281), Cry2 (log2FoldChange = −0.6972396, padj = 0.0005563706), Per3 (log2FoldChange = −0.4177657, padj = 0.006854384), and Nr1d1 (log2FoldChange = −0.3691834, padj = 0.08547361) showed a reduced or declining expression pattern, other circadian rhythm-associated genes such as Arntl (log2FoldChange = 0.183808, padj = 0.46584), Clock (log2FoldChange = 0.1064053, padj = 0.4669249), and Rorb (log2FoldChange = −0.09287985, padj = 0.7034156) displayed minimal changes or only slight reductions in expression (Supplementary Figure S3).

Figure 5. Altered Expression of Circadian Rhythm-Related Genes in AD Model Mice:

(A) Target Regulatory Relationship (TRR) plot using the reduced DEGs. Clock and Arntl play a central role in circadian rhythms. (B) GSEA plot of gene sets related to circadian rhythm. (C) Bar graph of genes related to circadian rhythm in DEGs in the brain of AD model mice. Error bars indicate SEM.

Discussion:

In this study, we analyzed the transcriptomic changes in the corpus callosum of the AD model mice using RNA-seq. Our findings revealed that (i) pathway analyses demonstrated an upregulation of genes associated with inflammatory response and regulation of cytokine production, especially those derived from microglia, in AD model mice (Figure 3), and (ii) genes related to chaperone function were downregulated (Figure 4), and (iii) notably, clock-controlled genes exhibited a marked downregulation (Figure 5). The upregulation of genes linked to microglia-mediated inflammatory responses, as observed in our study, aligns with previous research that emphasizes the central role of demyelination and neuronal damage driven by microglia-mediated neuroinflammation in white matter injury [30,31]. Furthermore, our data also support the idea that disruption of chaperone function and circadian dysfunction may contribute to AD pathologies.

One of the most significant findings of this study is the notable inflammatory response related to microglial activation in the corpus callosum of the AD brain. Microglia, the principal immune cells constituting 10% of all brain cells, play pivotal roles in both healthy and pathological states [31,32]. Accumulating evidence suggests that in the aging brain and in AD, microglia-mediated chronic neuroinflammation might be the central mechanism [33]. In AD patients, activated microglia have been observed surrounding neurotic plaques composed of Aβ peptide, highlighting an association between microglia and the accumulation of Aβ in the pathology [34]. Recent studies have elucidated that myelin dysfunction and demyelination injuries in the corpus callosum are strongly implicated as potent promoting factors of amyloid deposition in AD model mice [35]. Re-analysis of the bulk RNA-seq of microglia derived from the brains of 6-month-old 5xFAD transgenic mice (GSE178296) [35] revealed that among the DEGs obtained in the corpus callosum in this study (a total of 650 genes), 134 genes, including Itgax, Trem2, Fcer1g, and Grn, were similarly altered DEGs in microglia (|log2Fold Change| > 0.58, padj < 0.05) (Supplementary Figure S4). This suggests strongly that activated microglia in the corpus callosum of the 5xFAD transgenic mice are deeply involved in their pathology. An in-vitro study shows that microglia can phagocytize Aβ, leading to the significant accumulation of surface molecules associated with type I and II major histocompatibility complex [36–38]. Moreover, activated microglia can stimulate neurons to overproduce Aβ, causing synaptic damage and other neurodegenerative changes in AD [39]. Genome-wide association studies have highlighted genes related to the immune response (CR1, CD33, MS4A, CLU, ABCA7, EPHA1, and HLA-DRB5-HLA-DRB1) that are highly expressed in microglia and correlate with an increased risk of sporadic and late-onset AD [40]. These findings, spanning preclinical and clinical studies as well as in-vitro and in-vivo studies, emphasize the multifaceted role of microglia in AD pathology. Although the exact role of microglia and neuroinflammation in the corpus callosum of AD brain remains unclear, our results underscore the significance of microglia-mediated neuroinflammation in its pathology.

On the other hand, downregulated genes highlighted a suppression of chaperone function, indicating potential disruptions in protein folding in the corpus callosum of AD model mice. Given that protein misfolding is a key player in AD pathogenesis, this decrease suggests compromised cellular machinery to manage misfolded proteins in the corpus callosum. Members of the HSP70 family, such as Hspa1a and Hspa1b, which were found to be downregulated in our study of AD mouse models, are believed to co-localize with Aβ plaques and play a role in neuroprotective responses that inhibit Aβ aggregation [41–43]. Since Aβ plaques are observed in the corpus callosum of 12-month-old 5xFAD transgenic mice used in this study [18], disruption of chaperone function might contribute to the formation of these β-amyloid plaques. We also showed that the expression levels of several genes regulated by circadian rhythms (Cry1, Per1, Per2, and Vwf) were decreased in a mouse model of AD (Figure 5C), while the mRNA levels of Arntl (known as Bmal1) and Clock, which play a central role in circadian rhythms, were unchanged (Supplementary Figure S3). Although confirmation of the expression of these proteins is needed in the future, the changes in genes regulated by circadian rhythms that we found in this study may be involved in AD pathogenesis. Circadian rhythms are modulated by various pathways, such as the master oscillator in the suprachiasmatic nuclei located in the hypothalamus and the other oscillators in all organs throughout the body, for robust timekeeping. Recent research suggests that circadian rhythms play critical roles in various mechanisms associated with CNS diseases, including AD and stroke [44–48]. Pathologically, a study using transgenic AD mice showed a significantly reduced number of the neuropeptides, arginine vasopressin, and vasoactive intestinal polypeptide, secreted from neurons in suprachiasmatic nuclei [49]. In the clinical setting, circadian rhythm disruption is prevalent in the early stage of AD patients, and circadian alignment therapy is beneficial for treating sundowning syndrome and other cognitive symptoms in advanced AD patients [50–54]. Preclinical studies have also reported circadian dysfunction in mouse models of AD, although the results are somewhat variable and inconsistent across models, ages, and conditions [55]. These findings suggest that circadian dysfunction and non-typical lesions of AD, other than the cerebral cortex and hippocampus, are involved in AD pathology. Our finding exhibiting a downregulation of clock-controlled genes in the white matter of the AD brain also supports these findings and suggests that white matter can be a novel therapeutic target of circadian dysfunction in AD patients.

However, our current study has several limitations. First, our study examined only 5xFAD transgenic mice. Although this transgenic AD model mice have contributed to the understanding of AD mechanism, the full spectrum of AD effect has not been recapitulated in a mouse model. This transgenic mouse model remains limited in AD phenotypes compared with AD humans. Further studies with other AD models, e.g. EFAD mice (5xFAD expressing a human apoE isoform), may be required to confirm the present findings [56]. Second, the use of bulk samples of corpus callosum in this study contains the possibility that significant changes in gene expression in some cell types may have been canceled. It will be useful for future studies to examine gene expression profiles with single-cell RNA sequencing to further our understanding of transcriptomic profiles of the corpus callosum in the AD brain. Third, while we focused on the changes in gene expression of AD corpus callosum, we did not examine the cerebral cortex and hippocampus, which are central to AD pathology. Future studies should investigate whether these changes in gene expression are region-specific or not. Fourth, our tissue sampling was performed in only a one-time point, ZT3-5, which is the inactive phase for rodents. Since emergent finding suggests that the influence of circadian rhythm must be considered for translational study in the central nervous system, further study with multiple time points may be needed to confirm our findings [44]. Lastly, male animal bias should be considered because we included only males in the present study.

In conclusion, our study sheds light on the previously understudied white matter pathology in Alzheimer’s disease, particularly in the corpus callosum. We have identified key transcriptomic changes, including microglia-mediated inflammation, synaptic dysfunction, and circadian disruption, in this critical brain region. These findings broaden our understanding of AD beyond the cortex and hippocampus, offering new avenues for therapeutic interventions.

Supplementary Material

Supplementary material related to this article can be found in the “Supplementary information”.

Abbreviations:

Aβ

amyloid-β

AD

Alzheimer’s disease

CC

corpus callosum

DEGs

differentially expressed genes

GO

gene ontology

HSP70

heat shock protein 70kDa

MBP

myelin basic protein

NORT

novel object recognition test

OLG

oligodendrocytes

OPC

oligodendrocyte precursor cell

padj

adjusted p-value

RNA-seq

RNA sequencing

SEM

standard error of the mean

Data Availability:

The RNA-seq data have been deposited in the public repository under the accession code listed in the Materials and Methods.