ApoE4 Drives Microglial Lipid Dysregulation in Alzheimer’s Disease via Epigenetic Reprogramming of the Asxl1/LXRα–H3K4me3 Axis

Abstract

Background

The ε4 allele of apolipoprotein E gene (APOE) stands as the greatest genetic risk factor for late-onset Alzheimer’s disease (AD). Although microglia accumulating lipid droplets (LDAM) have been implicated in AD pathogenesis, the mechanistic link between ApoE4 and microglial lipid dysregulation remains elusive.

Methods

We employed a multi-omics approach, combining snRNA-seq and locus-specific epigenetic analysis, alongside microglia-specific gene manipulation in ApoE-targeted replacement (TR) mice. Primary microglia were challenged with cholesterol to simulate lipid overload conditions.

Results

In mid-life ApoE4-TR mice, microglia within the dentate gyrus developed pronounced lipid droplet accumulation, concurrent with impaired Aβ clearance and a pro-inflammatory shift. snRNA-seq unveiled a unique microglial cluster in ApoE4 mice, enriched for lipid-metabolism genes and marked by the pronounced downregulation of the hub gene Asxl1. Mechanistically, ApoE4 attenuated the Asxl1–LXRα interaction, leading to reduced H3K4me3 occupancy at promoters of lipid-efflux genes such as Abca1. Crucially, CRISPR-mediated, microglia-specific overexpression of Asxl1 restored H3K4me3 levels, normalized cholesterol efflux, and rescued Aβ phagocytic deficits in vivo.

Conclusions

Our findings define an epigenetic pathway whereby ApoE4 drives microglial dysfunction via the Asxl1–LXRα–H3K4me3 axis, fostering the LDAM phenotype. Enhancing Asxl1 function presents a promising therapeutic avenue for countering ApoE4-mediated pathogenesis in AD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03740-3.

Introduction

Alzheimer’s disease (AD) is a multifactorial neurodegenerative disorder whose incidence doubles every five years after the age of 65 [1, 2]. The ε4 allele of apolipoprotein E (APOE4 ε4) remains the strongest genetic risk factor for sporadic late-onset Alzheimer’s disease (LOAD), compared with the ε3/ε3 genotype, carrying one APOE ε4 allele increases the risk of late-onset AD by about three times; Carrying two ε4 alleles increases the risk by 8–15 times and advances the onset age by 7–9 years [3, 4]. While ApoE4 has historically been studied for its effects on amyloid-β (Aβ) aggregation and clearance, an emerging body of literature positions ApoE4 as a central regulator of microglial immunometabolism [5–8].

Microglial lipid accumulation in human AD cortical tissue is profoundly influenced by the AD-risk variant gene ApoE4 [9, 10]. In human induced pluripotent stem cell-derived microglia (iMG), this lipid-laden phenotype is elicited in an ApoE4-dependent manner, accompanied by elevated fibrillar Aβ (fAβ) levels, augmented triglyceride synthesis, robust lipid-droplet (LD) formation, pathological tau phosphorylation, and overt neurotoxicity [10]. Likewise, dysfunctional cholesterol-rich lysosomes and excessive LD accumulation within microglia have been documented in both demyelinating mouse models and patient-specific induced pluripotent stem cells (iPSC) platforms [10–13].

Microglia, the primary immune sentinels of the central nervous system, undergo dramatic transcriptional and metabolic reprogramming in response to aging, injury, or pathology [9, 14, 15].. A recently identified microglial state—termed “lipid-droplet-accumulating microglia” (LDAM) —is characterized by excessive neutral lipid storage, oxidative mitochondrial dysfunction, and blunted phagocytosis [8]. Single-cell RNA-seq of human AD cortices revealed that LDAM are enriched in ApoE4 carriers, and spatial transcriptomics showed that these cells cluster around dense-core plaques [10]. However, the molecular mechanisms by which ApoE4 instructs microglia to adopt the LDAM phenotype remain largely unknown.

Beyond lipid handling, ApoE4 is known to dampen microglial responsiveness to damage-associated molecular patterns (DAMPs) and to skew cytokine profiles toward a pro-inflammatory signature [9, 16]. Yet these functional impairments are not fully explained by altered ApoE4 lipidation or receptor binding alone. Epigenetic dysregulation has emerged as a unifying theme: ApoE4 carriers exhibit reduced chromatin accessibility at enhancers bound by the nuclear receptor LXRα [17, 18], a master transcriptional activator of cholesterol efflux genes such as Abca1 [19]. LXRα activity is tightly coupled to the integrity of the MLL/COMPASS complex, which deposits the activating mark H3K4me3 [20, 21]. Recent macrophage studies have identified Asxl1 (Additional Sex Combs-Like 1) as an essential scaffold within this complex, linking LXRα to the methyltransferase activity of MLL [22–24]. Germline Asxl1 mutations in myeloid malignancies lead to global loss of H3K4me3 and aberrant lipid accumulation [23, 25–27], but the role of Asxl1 in microglia has not been explored.

We therefore proposed a mechanistic hypothesis: ApoE4 acts to suppress Asxl1 expression. This initial insult destabilizes the Asxl1–LXRα–MLL complex, leading to diminished LXRα binding and a loss of the activating H3K4me3 mark at the Abca1 promoter. The transcriptional downregulation of Abca1 that follows disrupts cellular lipid homeostasis, directly driving microglial lipid droplet accumulation and functional impairment, including compromised Aβ clearance (as shown in the left panel of the mechanism schematic). Conversely, bolstering Asxl1 expression can restore this regulatory axis, enhance lipid efflux, and promote a functional, anti-inflammatory microglial state (right panel), thereby validating the proposed pathway. To test this model, we combined single-nucleus RNA-seq across aging time-points, lipid overload of primary microglia, and a newly generated microglia-specific Asxl1-over-expression mouse line. Our findings uncover a reversible epigenetic mechanism through which ApoE4 dictates microglial lipid metabolism and offer a potential therapeutic avenue for ApoE4-linked AD.

Results

ApoE4 Triggers Age-Dependent Lipid Accumulation in Microglia

Spatial and Cellular Specificity of Lipid Droplet Formation

To delineate the spatiotemporal distribution of lipid droplets, we systematically collected continuous coronal Sect. (30 μm thick) from the brains of female ApoE3-TR and ApoE4-TR mice at ages 5, 10, and 15 months (n = 6 per genotype and time point). Immunofluorescent imaging of the hippocampal dentate gyrus (DG) revealed a progressive, age-dependent expansion of the lipid-droplet burden in ApoE4 mice. Quantitative analysis of the cumulative Bodipy⁺ area within Iba1⁺ microglia showed a 2-fold increase in ApoE4 animals by 10 months (3.15 ± 1.2 vs. 1.56 ± 0.5, p < 0.01) and a 3.5-fold increase by 15 months (6.17 ± 1.2 vs. 2.04 ± 0.6, p < 0.001) relative to age-matched ApoE3 controls (Fig. 1A-B). Furthermore, lipid droplet accumulation was not restricted to the hippocampus but was also evident in the prefrontal cortex (PFC) of ApoE4 mice. Specifically, quantification of the lipid load within microglia (Iba1 + area) showed a significant increase in ApoE4 animals. The area occupied by large lipid droplets (> 0.4 μm) was approximately double that of ApoE3 controls at 10 months (3.54 ± 0.9 vs. 1.67 ± 0.4; p < 0.01) and escalated to nearly 3-fold by 15 months (5.97 ± 0.6 vs. 1.98 ± 0.5; p < 0.001) (S-Fig. 1A-B).

Fig. 1.

ApoE4 drives age-dependent microglial lipid accumulation and impaired Aβ uptake. A Spatial distribution of microglial lipid accumulation in the dentate gyrus (DG) of female ApoE-TR mice at 10 and 15 months of age. B-C Quantification of total lipid-droplet area and droplet-size distribution in microglia. D-E Pearson correlation between fluorescent staining with the lipophilic dye Bodipy fluorescence and cell-type markers (Iba1⁺, GFAP⁺, NeuN⁺) identifying lipid-droplet localization across microglia, astrocytes and neurons. F-G PLIN2 staining and morphometry of lipid-droplet area (F)/size(G) in middle-aged microglia. H Tail-vein Aβ-FITC488; flow cytometry of microglia (CD11b⁺) for Aβ-FITC488⁺ signal; mean fluorescence intensity (MFI). I Confocal images of Iba1⁺ microglia containing Aβ-FITC488 and quantified intracellular MFI. N = 6–10 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE3 group vs. the ApoE4 group

High-resolution three-dimensional confocal reconstructions (0.2 μm z-stacks) further resolved droplet size distributions within individual microglia. ApoE4 microglia exhibited markedly enlarged lipid droplets, with the proportion exceeding 0.4 μm in diameter rising progressively across age. Specifically, 45.8 ± 2.1% of droplets surpassed this threshold in 10-month-old ApoE4 microglia versus 15.7 ± 1.4% in ApoE3 (p < 0.001). By 15 months, the disparity intensified: 62.7 ± 2.6% of ApoE4 droplets exceeded 0.4 μm compared with 20.1 ± 1.8% in ApoE3 (p < 0.001), representing a sustained ~ 3-fold difference (Fig. 1C).

To determine the cell-type specificity of lipid-droplet accumulation, we immunostained microglia (Iba1), astrocytes (GFAP), and neurons (NeuN) in 10-month-old ApoE4 mice and quantified their colocalization with Bodipy. Line-profile plots showed a sharp, overlapping peak for Iba1 and Bodipy within microglia, whereas GFAP⁺ astrocytes and NeuN⁺ neurons displayed little to no signal overlap. Corroborating these data, image-correlation analysis revealed a strong positive correlation between lipid droplets and Iba1⁺ microglia (PCC = 0.575), while GFAP⁺ astrocytes (PCC = 0.205) and NeuN⁺ neurons (PCC=−0.039) exhibited progressively more negative correlations (Fig. 1D-E). Thus, age-dependent lipid-droplet accumulation in ApoE4 mice is both marked and highly selective, with microglia serving as the dominant cellular reservoir.

PLIN2 Validation Confirms Lipid Droplet Integrity

To confirm that the neutral-lipid–rich structures detected with fluorescent staining with the lipophilic dye Bodipy 493 truly represent membrane-bound lipid droplets (LDs) rather than non-specific lipid aggregates, we performed high-resolution immunostaining against Perilipin 2 (PLIN2, also known as ADRP)—a well-established lipid-droplet coat protein that is recruited to nascent LDs within minutes of neutral-lipid synthesis and remains stably associated with mature droplets [28]. When microglia accumulate lipids, they express PLIN2, which offers an orthogonal marker to identify bona fide lipid droplets.

Using systematic random sampling, we analyzed every 6th 30-µm coronal section spanning the entire rostro-caudal extent of the DG (Bregma − 1.06 mm to − 3.88 mm). Perilipin2 immunostaining confirmed that 10 month ApoE4 mice harbored a 3.1-fold denser population of Perilipin2⁺ Iba1⁺ droplet in DG (30.33 ± 4.86 vs. 11.85 ± 2.41 puncta/100 µm², p < 0.001)(Fig. 1G). Likewise, we mapped the size distribution of intracellular lipid droplets in microglia. Within PLIN2⁺ Iba1⁺ cells of ApoE4 mice, droplets were markedly enlarged: 57.82% in the DG exceeded 0.4 μm in diameter—figures nearly three-fold higher than in ApoE3-expressing microglia (Fig. 1F). The concurrent rise in PLIN2⁺Iba1⁺ puncta and droplet diameter reveals that ApoE4 not only accelerates lipid synthesis but also stabilizes nascent droplets by boosting PLIN2 recruitment—an acute compensatory response to transient lipid overload [29]. Collectively, PLIN2 immunostaining demonstrate that the Bodipy-positive structures in ApoE4 microglia are bona-fide, membrane-bound lipid droplets whose density and volume are markedly increased, providing a robust validation of LDAM formation in ApoE4-TR brain.

Functional Consequence: Impaired Aβ Phagocytosis

To evaluate the impact of ApoE genotype on microglial function in vivo, we employed a peripheral Aβ challenge model to assess phagocytic clearance. Intravenously injected FITC-conjugated Aβ( 1–42) crosses the blood-brain barrier in aged mice through receptor-mediated endocytosis pathways influenced by age and ApoE genotype [30–32]. Once within the brain parenchyma, fibrillar Aβ is predominantly cleared via phagocytosis by microglia [33, 5]. To quantify this process, we performed high-resolution confocal microscopy on perfused brain Sect. 24 h post-injection—a time frame sufficient to capture both uptake and lysosomal degradation [34]. Fluorescence intensity of internalized FITC-Aβ within Iba1⁺ microglia was measured using 3D image analysis, providing a spatially resolved assessment of phagocytic activity in the intact brain.

To determine whether the pronounced lipid droplet accumulation in ApoE4 microglia leads to functional impairment in Aβ clearance, we conducted an in vivo phagocytosis assay. Aβ1–42 was N-terminally labeled with FITC (Aβ-FITC₄₈₈) to enable tracking of peptide uptake and processing. Under brief isoflurane anesthesia (2% in O₂), mice received a single intravenous bolus of Aβ-FITC₄₈₈ (5 µg/g body weight) via the lateral tail vein. Consistent with our hypothesis, ApoE4-targeted replacement (ApoE4-TR) mice exhibited a significant deficit in Aβ handling by microglia. Fluorescence-activated cell sorting and flow cytometric analysis revealed that CD11b⁺/Aβ-FITC₄₈₈⁺ microglial events were reduced to approximately 51% of levels in ApoE3-TR controls (Fig. 1H). Moreover, the mean Aβ-FITC₄₈₈ intensity within sorted microglia was 68% lower in ApoE4 mice 24 h after injection (Fig. 1I). Together, these findings demonstrate that ApoE4-associated lipid accumulation compromises the ability of microglia to clear Aβ in vivo.

Transcriptomic Dynamics Revealed by snRNA-seq Time-Course

We performed comprehensive single-nucleus RNA sequencing analysis of hippocampal tissues from ApoE3 and ApoE4 genotype mice. Following rigorous quality control, we obtained 92,484 high-quality nuclei that passed established filtering criteria. UMAP visualization enabled the identification of 27 distinct cell clusters (clusters 0–26), representing the major cellular constituents of the hippocampus, including excitatory neurons, inhibitory neurons, oligodendrocytes, astrocytes, and microglia (Fig. 2A). Notably, based on the expression profiles of specific marker genes (such as Tmem119, P2ry12, Cx3cr1) [35], we identified clusters 9 and 21 as microglial populations, which play critical roles in immune response and lipid metabolism in the hippocampus.

Fig. 2.

Single-nucleus profiling of the mouse hippocampus uncovers cell-type clusters and age-dependent Asxl1 dysregulation in ApoE4 microglia. A UMAP projection depicting 27 distinct cell clusters (0–26) derived from single-nucleus cell RNA sequencing of hippocampal tissues from female ApoE3 and ApoE4 mice. Major cell types are distinguished by color, including excitatory neurons (e.g., Dentate Gyrus Granule Neurons), inhibitory neurons, oligodendrocytes, astrocytes, and microglia (clusters 9 and 21). B Network Venn diagram illustrating the differential expression of the Asxl1 gene in microglia across various age groups (5, 10, and 15 months) and ApoE genotypes. C A line graph illustrates the trajectory of mean Asxl1 expression in microglia over time for ApoE3 and ApoE4 genotypes. D, E, F Volcano plots illustrate the differentially expressed genes (DEGs) in hippocampal microglia from ApoE3 mice compared to ApoE4 controls at 5, 10, and 15 months of age, respectively. Each point represents an individual gene, with the x-axis indicating the log2 fold change (ApoE3/ApoE4) and the y-axis representing the -log10 (p-value). The red dot highlights Asxl1, blue dots denote other significant DEGs (p-value < 0.05), and gray dots indicate non-significant genes. G, H, I Heatmaps display the DEGs identified in the cohorts aged 5, 10, and 15 months, respectively. Each row corresponds to a gene, while each column represents an individual sample. The color bar above indicates genotype (ApoE3, red; ApoE4, blue). Gene expression Z-scores are depicted by the color scale (blue: low expression; red: high expression). Dendrograms illustrate the clustering of samples and genes

To investigate the impact of ApoE genotype on microglia, we extracted and analyzed microglial clusters 9 and 21. We systematically compared gene expression differences between ApoE3 and ApoE4 microglia across three age cohorts (5, 10, and 15 months). Network analysis revealed that Asxl1 was consistently identified as a differentially expressed gene across all time points (Fig. 2B). Analysis of its expression trajectory showed a significant age-dependent downregulation in ApoE4 microglia compared to the stable levels in ApoE3 controls (Fig. 2C). Differential expression analysis confirmed that Asxl1 emerged as a significantly downregulated gene in ApoE4 microglia at 10 and 15 months (Fig. 2D-F). This sustained dysregulation in aged ApoE4 microglia was further corroborated by heatmap visualization of the differentially expressed genes (Fig. 2G-I). However, no significant genotype-dependent differences in Asxl1 expression were detected at 5 months, suggesting that this dysregulation becomes evident with advancing age and may be associated with progressive pathological processes.

We next characterized the heterogeneity within the microglial population. Unsupervised clustering analysis revealed three distinct microglial subpopulations (Fig. 3A, B). Based on the expression profiles of canonical marker genes, we annotated them as: homeostatic microglia (HOMEOSTATIC, subcluster 1), disease-associated microglia (DAM, subcluster 2), and lipid-associated microglia (LDAM, subcluster 3) (Fig. 3C). Violin plot analysis indicated that Asxl1was expressed across all subclusters, showing a trend of decreased expression in activated states such as DAM and LDAM, although this did not reach statistical significance (Fig. 3D). To decipher the dynamic transitions between these states, we performed pseudotime trajectory analysis, which delineated a continuous activation trajectory originating from the homeostatic state (HOMEOSTATIC) (Fig. 3E). The microglial subclusters were distributed along this pseudotime axis, illustrating a progression from the homeostatic state toward the activated DAM and LDAM fates (Fig. 3F). Notably, the expression level of Asxl1exhibited a significant negative correlation with pseudotime, progressively decreasing as cells advanced along the activation trajectory (Fig. 3G). Density analysis further confirmed that cells classified as “homeostatic” and “activated” were concentrated at distinct phases of this continuum (Fig. 3H). Our analysis maps a pseudotemporal trajectory from homeostatic to activated microglia and uncovers an associated downregulation of Asxl1.

Fig. 3.

Hippocampal Microglia in ApoE4 Mice Undergo a Heterogeneous Transition along an Activation Trajectory. A Microglial key markers heatmap. B Microglial subclusters heatmap. C Microglial subclusters annotation. D Violin plots display the distribution of Asxl1 expression levels across the four microglial subclusters, with plot width representing cell density at each expression level. E Pseudotime trajectory of microglial cells. UMAP projection of microglial subclusters, with cells colored by their inferred pseudotime (scale 0 to 1). The overlaid black line indicates the principal developmental trajectory, illustrating the continuum of cellular states. F Microglial subclusters along the trajectory. G Dynamic expression of Asxl1across pseudotime. H Density distribution of microglial states. Overlapping density plots for cells classified as HOMEOSTATIC (red) and ACTIVATED (blue) states, peaking at pseudotime values of approximately 0.25 and 0.75, respectively, quantifying the distribution of these functional states along the developmental continuum

To elucidate the functional consequences of the transcriptional changes in ApoE4 microglia, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses on differentially expressed genes at each age. At 5 months, enriched pathways were predominantly associated with immune response and synaptic organization (S-Figure 2 A, D). By 10 months, the enrichment profile shifted significantly, showing pronounced involvement of pathways related to lipid metabolism, Alzheimer’s disease (AD), and MAPK signaling (S-Figure 2B, E). This trend intensified at 15 months, with strong enrichment in fatty acid biosynthesis, glycerolipid metabolism, and AD-related signaling pathways (S-Figure 2 C, F). These results indicate that gene modules containing Asxl1, which are altered in the ApoE4 genotype, undergo an age-dependent functional shift—from early roles in immunity and basic cellular functions toward later involvement in lipid metabolism regulation.

In parallel, we analyzed hippocampal astrocytes, the primary CNS producers of ApoE. In contrast to microglia, the expression of Asxl1 and Abca1 in astrocyte subclusters did not show significant genotype-dependent changes across ages (S-Figure 3 A, B). Differential expression analysis between genotypes identified distinct sets of dysregulated genes in astrocyte subclusters at 5, 10, and 15 months (S-Figure 3 C-E). Subsequent pathway enrichment analysis revealed that, similar to microglia, the transcriptional changes in ApoE4 astrocytes were age-dependent. Enriched pathways shifted from those related to synapse organization and cell adhesion at 5 months towards significant enrichment in lipid metabolism and Alzheimer’s disease signaling at 10 and 15 months (S-Figure 3 F-K). In astrocytes, ApoE4 also drives age-dependent transcriptional remodeling toward a state of dysregulated lipid metabolism and AD-related signaling. Notably, the resulting lipid droplet pathology may involve additional pathways. For instance, ApoE4 directly impairs fatty acid oxidation [36], which contributes to the observed lipid metabolic dysfunction.

To corroborate the transcriptomic profiles, we conducted quantitative RT-PCR on FACS-purified microglia from ApoE4-TR and ApoE3-TR mice at 5, 10, and 15 months of age. In agreement with the snRNA-seq results, ApoE4 mice exhibited a pronounced downregulation of key lipid-efflux genes relative to age-matched ApoE3 controls at both 10 and 15 months. Specifically, Abca1 expression was significantly reduced to 0.45 ± 0.13-fold (p < 0.01) and 0.35 ± 0.12-fold (p < 0.01), and Asxl1 levels declined to 0.44 ± 0.12-fold (p < 0.01) and 0.33 ± 0.07-fold (p < 0.001), respectively. Conversely, pro-inflammatory cytokines were markedly upregulated: IL-6 increased by 3.13 ± 0.58-fold (p < 0.01) at 10 months and 5.19 ± 0.91-fold (p < 0.001) at 15 months, while CXCL1 expression rose by 2.40 ± 0.47-fold (p < 0.01) and 3.41 ± 0.64-fold (p < 0.001) at the corresponding ages (Fig. 4A). To examine age-related changes in lipid homeostasis regulators, we performed immunoblot analysis of Asxl1 in microglia freshly isolated from ApoE-TR mice at 1, 5, 10, and 15 months of age. Asxl1 levels declined progressively in ApoE4 microglia, diverging significantly from age-matched ApoE3 controls by 10 months (− 45%, p < 0.01) and reaching a 37% reduction at 15 months (p < 0.05)(Fig. 4B). Subsequent targeted epigenetic profiling of additional lipid-efflux proteins in 10-month-old sorted microglia revealed concerted down-regulation of Asxl1 (− 47%), Abca1 (− 51%) and LXRα (− 43%) in ApoE4 versus ApoE3 mice (Fig. 4C). These molecular features align with the changes observed in Asxl1 within single-nucleus transcriptomic data, linking it to ApoE4-driven lipid metabolic dysregulation and the progressive pathology of Alzheimer’s disease.

Fig. 4.

Impaired Expression of Lipid Metabolism Proteins in Middle Aged ApoE4 Microglia. A Age-resolved RNA-seq of ApoE-TR microglia (5, 10, 15 months) showing differential expression of lipid-efflux and cytokine genes. B Immunoblots quantifying age-dependent Asxl1 loss in ApoE-TR microglia. C Quantitative immunoblotting reveals dysregulation of key lipid-metabolic proteins (Asxl1, Asxl2, LXRα, and Abca1) in middle-aged ApoE4-TR microglia. N = 4–6 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE3 group vs. the ApoE4 group

Taken together, these findings suggest a cell-type-specific pattern of dysregulation occurs in the ApoE4 hippocampus. In microglia, ApoE4 may drive an age-related decline in Asxl1, potentially underlying a pseudotemporal shift from homeostatic to activated states. This transition is accompanied by a refocusing of the transcriptome away from immune functions and toward lipid metabolism and Alzheimer’s disease pathways.

ApoE4-Driven Microglial Lipidosis Model Closely Reconstructs In Vivo Functional Deficits

Dose-Dependent Lipid Accumulation via Cholesterol Loading

To recapitulate the lipid-overload milieu observed in aged ApoE4 brains under controlled conditions, we established an in vitro cholesterol-loading paradigm using primary microglia (MG) isolated from ApoE-TR neonates. Cells were treated for 24 h with cholesterol–BSA complexes spanning a physiologically relevant range (0–50 µg/ml CHO-BSA) (Fig. 5A). Quantitative lipid droplet staining revealed a robust, dose-dependent increase in neutral-lipid droplet area specifically in ApoE4 MG (Fig. 5B). At a cholesterol dose of 10 ug/ml, the proportion of large droplets (greater than 0.4 μm) in ApoE4 microglia significantly increased, reaching 66%, which is twice as high as ApoE3 MG (38%), p < 0.01). At the highest cholesterol dose (50 µg/ml), large the proportion of droplet area reached 72%, representing a 1.8-fold elevation over ApoE3 MG (41%, p < 0.001)(Fig. 5B-C), confirming heightened lipid-accumulation sensitivity in the ApoE4 background.

Fig. 5.

Cholesterol-loaded ApoE4 microglia recapitulate in vivo dysfunction. A Primary microglia from ApoE-TR mice loaded with graded cholesterol. B-C Dose-dependent lipid droplet formation and size distribution. D-E Impaired Aβ-FITC488 uptake by ApoE4 microglia. F-G ELISA: cholesterol load dose-dependently elevates CXCL1 and IL-6 secretion. N = 6–8 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE3 group vs. the ApoE4 group

Synergistic Effect of Aβ and High Cholesterol in ApoE4 Microglia

To emulate the accumulation of ApoE4 lipid droplets under AD pathological conditions, we probed the interplay between Aβ burden and lipid accumulation by co-administering 5 µM Aβ₁₋₄₂ together with 0–50 µg/ml CHO-BSA. Under these conditions, the large droplet area of ApoE4 MG increased to 42% with the intervention of Aβ at a dose of only 5ug/ml exogenous cholesterol (S-Fig. 4A-B), while ApoE3 MG did not further increase, indicating the synergistic amplification of Aβ on lipid dysregulation in the context of ApoE4. Collectively, these data establish that ApoE4 primes microglia to accumulate lipids more readily and renders them functionally compromised—providing a robust in vitro platform for mechanistic dissection and therapeutic screening.

Coupled Pathological Outcomes: Defective Aβ Clearance and Cytokine Dysregulation

To probe the functional ramifications of these molecular alterations, we deployed a dual-readout platform that simultaneously quantifies Aβ phagocytosis and inflammatory cytokine secretion. MG were challenged with 1 µM Aβ-FITC₄₈₈ for 4 h. Live-cell imaging (IncuCyte S3) revealed that treatment with 10 µg/mL cholesterol decreased Aβ-FITC uptake by ApoE4-expressing microglia by 38% relative to the ApoE3 control (21.17 ± 3.14 µm² vs. 13.33 ± 2.03 µm², p < 0.01). Raising the cholesterol concentration to 50 µg/mL exacerbated the effect, suppressing Aβ-FITC internalization by 56% (ApoE4:7.83 ± 1.95 µm² vs. ApoE3: 17.68 ± 2.67 µm², p < 0.001)(Fig. 5D-E). We also analyzed the conditioned media by parallel ELISA. At a cholesterol concentration of 10 µg/mL, ApoE4 microglia secreted 2.1-fold more CXCL1 (962.85 ± 57.37 pg/ml vs. 446.35 ± 38.14 pg/ml in ApoE3) and 1.6-fold more IL-6 (1053.96 ± 60.48 pg/ml vs. 667.31 ± 28.09 pg/ml in ApoE3) (Fig. 5F-G, p < 0.01 for both). When the cholesterol challenge was increased to 50 µg/mL, this inflammatory disparity widened further: CXCL1 levels became 2.3-fold higher (1452.67 ± 60.28 pg/ml vs. 646.36 ± 42.06 pg/ml) and IL-6 levels became 1.8-fold higher (1498.99 ± 49.75 pg/ml vs. 871.72 ± 41.73 pg/ml) in ApoE4 microglia compared to their ApoE3 counterparts (Fig. 5F-G, p < 0.001 for both). Collectively, these results demonstrate that cholesterol exposure exacerbates both defective Aβ clearance and inflammatory cytokine secretion in ApoE4-expressing microglia in a dose-dependent manner.

Mechanistic Basis: ApoE4 Epigenetically Represses Lipid Homeostasis via the Asxl1–LXRα–H3K4me3 Axis

Molecular immunoblotting revealed that exogenous cholesterol dose-dependently repressed the expression of Asxl1, LXRα, and their common target Abca1 in ApoE4 microglia. After treatment with 5 µg/mL cholesterol, the expression levels of Asxl1, LXRα, and Abca1 in the ApoE4 microglia dropped to 47.69 ± 3.66%, 43.59 ± 4.72%, and 46.91 ± 5.84%, respectively, of those in the ApoE3 control group subjected to the same treatment (p < 0.05 or p < 0.01 for each). The suppression was sustained at 10 µg/mL (Asxl1: 56.75 ± 7.38%, p < 0.01; LXRα: 49.26 ± 6.75%, p < 0.05; Abca1: 48.05 ± 5.79%, p < 0.05) and at 50 µg/mL (Asxl1: 48.59 ± 8.17%, p < 0.05; LXRα: 43.92 ± 9.32%, p < 0.05; Abca1: 47.89 ± 6.55%, p < 0.05)(Fig. 6A-B). However, Asxl2 levels did not differ significantly among any of the groups. Co-immunoprecipitation and Western blotting confirmed a robust, endogenous interaction between Asxl1 and LXRα in ApoE3 microglia. Upon treatment with 10 µg/mL cholesterol, this association was markedly weakened in ApoE4 cells: compared to IgG, LXRα immunoprecipitation enriched Asxl1 61-fold in ApoE3, but only 16-fold in ApoE4, a 70.5% reduction. Reciprocally, Asxl1 immunoprecipitation enriched LXRα 78-fold in ApoE3, yet only 19-fold in ApoE4, corresponding to a 75.6% decrease (Fig. 6E).

Fig. 6.

ApoE4 represses lipid-homeostasis genes via the Asxl1–LXRα–H3K4me3 epigenetic axis. A–B Cholesterol dose-dependently down-regulates Asxl1, LXRα and Abca1 in ApoE4 microglia. C–D Cholesterol load reduces H3K4me3 but not H3K9me2, H3K27me2. E Co-IP confirms Asxl1–LXRα interaction under high cholesterol (10 µg mL⁻¹ CHO-BSA). F ChIP-qPCR shows cholesterol-rich ApoE4 microglia lose H3K4me3 and LXRα occupancy at the Abca1 gene promoter. N = 3–6 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE3 group vs. the ApoE4 group

To elucidate how Asxl1 interacts with LXRα to epigenetically regulate lipid-metabolism genes in microglia, we isolated nuclear proteins from a high-fat ApoE cell model and analyzed the histone methylation status. In ApoE4 microglia, the abundance of H3K4me3 decreased by 49.17% at a cholesterol concentration of 10 µg/ml and by 56.74% at 50 µg/ml (p < 0.01 and p < 0.001, respectively; Fig. 6C–D). To confirm H3K4me3-mediated regulation of the lipid-efflux gene Abca1, we performed ChIP-qPCR across the Abca1 promoter (−1 kb to + 200 bp relative to the TSS). In ApoE4 microglia exposed to a 10 µg/mL cholesterol-BSA complex (lipid-enriched model), H3K4me3 enrichment was 52.7% lower than in ApoE3 controls, whereas ChIP with LXRα antibodies showed a 41.16% decrease in LXRα binding at the same region (Fig. 6F). Collectively, these data indicate that ApoE4 destabilizes the Asxl1–LXRα complex, diminishing H3K4me3 deposition at lipid-efflux gene promoters and enforcing transcriptional silencing.

Therapeutic Proof-of-Concept: Microglia-Targeted Asxl1 Overexpression Reverses Pathology

Genetic Strategy and Validation of Conditional Transgenics

A CAG-loxP-STOP-loxP-Asxl1-3×FLAG knock-in cassette was inserted into the Rosa26 safe-harbor locus and crossed to CX3CR1-Cre mice, followed by > 10 generations of back-crossing onto the ApoE4-TR background (S-Fig. 5 A). Western blot of sorted microglia demonstrated a 3.1 ± 0.3-fold increase in Asxl1-3×Flag protein relative to ApoE4/Asxl1ᶠˡᵒˣ littermates (p < 0.001) (S-Fig. 5B).

In Vivo Reduction of Lipid Droplets

To assess the impact of Asxl1 overexpression on lipid-droplet accumulation in microglia of middle-aged ApoE4 mice, we conducted unbiased stereological analyses on three cohorts of 10-month-old mice. Quantitative analysis demonstrated that Asxl1 overexpression markedly rescued the lipid-droplet pathology in ApoE4 MG. It reduced the area of lipid-droplet accumulation in the dentate gyrus by 51.6% (Fig. 7A). Furthermore, three-dimensional volumetric reconstructions revealed a specific depletion of large lipid droplets, decreasing their proportion from 58.96 ± 2.71% in ApoE4/Asxl1ᶠˡᵒˣ controls to 27.52 ± 2.08% in ApoE4/Asxl1 + + mice (p < 0.001), thereby reversing the droplet size distribution to a level indistinguishable from ApoE3 controls (Fig. 7A). Collectively, these data indicate that Asxl1 overexpression normalizes pathological lipid accumulation in ApoE4 microglia.

Fig. 7.

Asxl1 overexpression rescues the functional signature of middle-aged ApoE4 microglia. A High-resolution spatial mapping of microglial lipid deposition in the dentate gyrus of 10-month-old female ApoE-TR mice with high Asxl1 expression. Comprehensive quantification reveals restored total lipid-droplet burden and normalized droplet-size distribution within microglia. B Microglial Aβ clearance is restored by Asxl1 overexpression. Representative confocal images of Iba1⁺ microglia engulfing Aβ-FITC₄₈₈ and quantitative analysis of intracellular Aβ-FITC₄₈₈ fluorescence intensity. C Tail-vein Aβ-FITC₄₈₈ administration followed by microglial flow-cytometry (CD11b⁺) shows a marked increase in Aβ-FITC₄₈₈⁺ events and mean fluorescence intensity (MFI) upon Asxl1 overexpression. D Asxl1 overexpression rebalances ApoE4 microglia from a pro-inflammatory state, as evidenced by reduced levels of IL-6 and CXCL1. N = 3–6 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE3 group vs. the ApoE4 group

Restoration of Aβ Clearance and Resolution of Neuroinflammation

Twenty-four hours after intravenous Aβ-FITC₄₈₈ (5 µg/g), microglia were isolated by flow cytometry and their fluorescence quantified. ApoE4/Asxl1⁺⁺ microglia exhibited 2.3-fold higher Aβ-FITC₄₈₈ fluorescence than ApoE4/Asxl1ᶠˡᵒˣ controls (Fig. 7B). Gating on CD11b⁺/Aβ-FITC₄₈₈⁺ events revealed a 56% increase in the phagocytic index in Asxl1-overexpressing mice (Fig. 7C), restoring clearance efficiency to levels indistinguishable from ApoE3 controls. Thus, Asxl1 overexpression fully rescues Aβ uptake by ApoE4 microglia.

Consistent with restored Aβ clearance, sorted microglia from Asxl1-overexpressing ApoE4 mice displayed a markedly attenuated inflammatory profile. IL-6 levels fell by 45.6% and CXCL1 by 38.7% relative to flox controls (Fig. 7D), indicating that Asxl1 overexpression reverts ApoE4 microglia from a pro-inflammatory to a homeostatic state.

Reversal of Cellular Cholesterol Overload by Asxl1-Mediated lipid Efflux In Vitro

Primary microglia isolated from ApoE3 mice, ApoE4 mice with microglial-specific Asxl1 overexpression (ApoE4/Asxl1⁺⁺), and ApoE4/Asxl1ᶠˡᵒˣ control littermates were treated with 10 µg/ml cholesterol–BSA complex for 24 h. Immunofluorescence analysis showed that Asxl1 overexpression reduced the area occupied by large lipid droplets (> 0.4 μm) by 44.8% compared to ApoE4/Asxl1ᶠˡᵒˣ controls (Fig. 8A). This clearance of lipid droplets was accompanied by a 54.5% increase in the phagocytosis of Aβ-FITC₄₈₈ (20.50 ± 1.52 μm² vs. 9.33 ± 0.56 μm² in controls; p < 0.001; Fig. 8B) and a significant reduction in the secretion of pro-inflammatory cytokines. Specifically, CXCL1 secretion decreased by approximately 52% (753.23 ± 62.49 pg/mL vs. 1,448.99 ± 113.13 pg/mL in controls), and IL-6 secretion decreased by approximately 47% (581.72 ± 24.94 pg/mL vs. 1,170.99 ± 89.30 pg/mL in controls) (p < 0.01 and p < 0.001, respectively; Fig. 8C).

Fig. 8.

Reversal of Cellular Cholesterol Overload by Asxl1-Mediated lipid Efflux In Vitro. A Restores lipid-droplet load and size distribution. B Restores Aβ-FITC uptake and intracellular fluorescence intensity. C Normalizes IL-6 and CXCL1 secretion. D [³H]-cholesterol efflux assay confirms restored lipid export. N = 3–6 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE4/Asxl1⁺⁺ group vs. the ApoE4/Asxl1^flox group

The functional recovery mediated by Asxl1 was further supported by [³H]-cholesterol efflux assays. Asxl1 overexpression significantly rescued the impaired cholesterol export in ApoE4 microglia, increasing HDL-mediated efflux to 84.4% and ApoA-1-mediated efflux to 80.3% of the levels observed in ApoE3 microglia. In contrast, ApoE4/Asxl1ᶠˡᵒˣ controls exhibited only 59.5% (HDL) and 49.8% (ApoA-1) of the ApoE3 efflux capacity (HDL: 20.54 ± 1.45% vs. 14.49 ± 1.28%, p < 0.001; ApoA-1: 12.68 ± 1.15% vs. 7.86 ± 0.75%, p < 0.01; Fig. 8D). These results confirm that Asxl1 restoration effectively re-establishes robust cholesterol efflux functionality.

Epigenetic Reprogramming of the Lipid Efflux Pathway

To investigate how Asxl1 overexpression reshapes lipid‑metabolism networks and the underlying epigenetic landscape in ApoE4 microglia, we performed Western blot, co‑immunoprecipitation (Co‑IP), and chromatin immunoprecipitation followed by qPCR (ChIP‑qPCR) on purified microglia from 10‑month‑old mice. Compared with ApoE4/Asxl1ᶠˡᵒˣ controls, ApoE4/Asxl1⁺⁺ microglia showed significant upregulation of key proteins: Asxl1 (2.3‑fold), Abca1 (2.2‑fold), and LXRα (1.8‑fold) (Fig. 9A). Notably, Asxl2 expression remained unchanged across groups. Co‑IP assays revealed a markedly enhanced Asxl1–LXRα interaction, with a 64.4% increase when Asxl1 was immunoprecipitated and a 63.2% increase when LXRα was immunoprecipitated, confirming strengthened functional coupling (Fig. 9B). At the epigenetic level, the active histone mark H3K4me3 increased 2.6‑fold (Fig. 9C). ChIP‑qPCR further demonstrated a 2.1‑fold enrichment of H3K4me3 at the Abca1 promoter region (− 1 kb to + 200 bp; p < 0.001) (Fig. 9D). Collectively, these findings demonstrate that Asxl1 overexpression restores lipid homeostasis in ApoE4 microglia by reactivating Abca1 transcription, as evidenced by restored global H3K4me3 levels (immunoblot) and increased H3K4me3 enrichment at the Abca1 promoter (ChIP–qPCR).

Fig. 9.

Epigenetic Reprogramming of the Lipid Efflux Pathway. A Asxl1 overexpression up-regulates Asxl1, LXRα, and Abca1 protein levels in ApoE4 microglia. B In vivo co-immunoprecipitation confirms that Asxl1 overexpression re-establishes the Asxl1–LXRα complex. C Asxl1 elevates H3K4me3 in ApoE4 microglia without altering global H3K9me2 or H3K27me2 levels. D ChIP-qPCR demonstrates that Asxl1 overexpression restores H3K4me3 occupancy at the Abca1 promoter in ApoE4 microglia. N = 3–5 per group, expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, for the ApoE4/Asxl1⁺⁺ group vs. the ApoE4/Asxl1^flox group

Discussion

ApoE4 programs microglia into a lipid-laden, dysfunctional state

We present multiple lines of evidence establishing ApoE4 as a driver and predisposing factor for LDAM formation in an in vivo context. LDAM constitute a transcriptionally distinct subpopulation that partially overlaps with “disease-associated microglia” (DAM) [8, 37] and Trem2-negative microglia [5], yet exhibits a unique metabolic signature characterized by ACSL1-mediated cholesterol esterification and PPARγ-independent lipid synthesis. Spatially resolved transcriptomic profiling has mapped LDAM to peri-plaque niches, where their local abundance positively correlates with soluble Aβ levels [10]. Our data extend these observations by demonstrating that ApoE4 potently accelerates LDAM emergence. First, mid-life ApoE4-TR mice already exhibit robust lipid droplet accumulation in the dentate gyrus and prefrontal cortex—regions that are among the earliest to accumulate Aβ in human AD. Second, snRNA-seq trajectory analysis places ApoE4 microglia along a continuous path toward a transcriptional state that closely resembles human LDAM. Third, cholesterol loading of ApoE4 primary microglia fully recapitulates the LDAM signature, underscoring a cell-autonomous mechanism. Crucially, this phenotype precedes significant plaque burden, suggesting that ApoE4-induced metabolic reprogramming is an early, potentially causal event rather than a late consequence of pathology. Extending these observations, the present study demonstrates that ApoE4 accelerates LDAM emergence even in the absence of overt amyloid plaques, indicating that ApoE4-driven metabolic dysfunction can precede and potentially exacerbate amyloid pathology. This temporal sequence aligns with clinical evidence showing that ApoE4 carriers display elevated CSF Aβ42 decades prior to cognitive decline [38], implicating LDAM as a mechanistic link between early ApoE4-mediated metabolic stress and prodromal AD pathogenesis.

Asxl1–LXRα–H3K4me3: an unrecognized ApoE4 effector axis

While the precise molecular link between the ApoE4 protein and reduced Asxl1 levels remains to be fully mapped (and is a key future direction), our data position Asxl1 as a critical regulatory node. The coordinated downregulation of Asxl1 and its downstream target Abca1 in ApoE4 microglia precedes or coincides with the emergence of a pronounced LDAM signature. This temporal and correlative evidence suggests that Asxl1 deficiency could be permissive or instructive for the lipid-accumulating cellular program.

ApoE4 may influence the epigenetic regulator Asxl1 through several converging mechanisms. The ApoE ε4 allele frequently forms a risk haplotype with neighboring genes such as TOMM40 and APOC1 [39], potentially altering three-dimensional chromatin structure and enabling long-range enhancer-promoter interactions that directly regulate Asxl1 transcription. Concurrently, ApoE4 drives microglia into a “pre-activated” state [40], creating a chronic inflammatory and stressed environment that can further modulate Asxl1 expression or activity. In human carriers of ApoE ε4, genome-wide methylation differences are observed and are notably enriched in lipid metabolism genes such as SREBF2 and ABCG1 [41]. Asxl1, as a key epigenetic regulator, may act as a critical effector in establishing or maintaining this ApoE4-specific methylation pattern, particularly within lipid metabolic pathways. We have present Asxl1 not simply as a differentially expressed gene, but as a potential epigenetic regulator whose loss may lower the threshold for LDAM commitment in the context of ApoE4-induced lipid stress.

Previous studies have attributed ApoE4-mediated dysregulation of microglial lipid metabolism to reduced LXR ligand (the nuclear receptors NR1H2/3) availability or impaired LXR nuclear translocation [42]. Our work shifts the paradigm by demonstrating that ApoE4 acts upstream, at the level of chromatin, to dismantle the Asxl1–LXRα transcriptional complex. Co-IP revealed that ApoE4 LDAM decreases the physical interaction between Asxl1 and LXRα by ~ 50%. Asxl1 is a pivotal chromatin regulator that orchestrates gene transcription by modulating histone methylation marks, thereby influencing cell fate decisions between differentiation and proliferation. It achieves this by engaging the MLL3/4 histone methyltransferase complex—via a C-terminal interaction—to LXRα-bound enhancers [43]. In the absence of Asxl1, this recruitment is lost, resulting in a pronounced reduction of H3K4me3 at the promoters of key lipid-export genes such as Abca1 [44]. ChIP analyses further indicate that these epigenetic changes selectively occur at the same Abca1 promoter site where LXR α binds, and synergistically with H3K4me3, leading to downregulation of cholesterol efflux gene expression Abca1. Asxl1 overexpression is sufficient to restore H3K4me3 occupancy, thereby confirming the specificity of this mechanism. This restoration encompasses both global H3K4me3 levels (immunoblot) and targeted enrichment at the Abca1 promoter (ChIP–qPCR). Collectively, the discovery of the Asxl1–LXRα–H3K4me3 axis elucidates a novel pathogenic mechanism, initiated by ApoE4 and operating at the chromatin level, which provides a critical framework for understanding how ApoE4 reshapes microglial fate and function.

Therapeutic implications of epigenetic rescue

The present study demonstrates that microglia-restricted Asxl1 over-expression reverses ApoE4-driven lipid-droplet accumulation, restores Aβ phagocytosis, and attenuates neuroinflammation (Fig. 10). These findings align with recent reports that Asxl1 is an obligate scaffold linking LXRα to the MLL/COMPASS methyltransferase complex, enabling H3K4me3 deposition at cholesterol-efflux loci [24]. Consistent with this mechanism, we observed that Asxl1 rescue reopened 56% of previously silenced LXRα-bound enhancers (Fig. 9D). Importantly, the therapeutic window appears broad: even 10-month-old ApoE4-KI mice showed rapid LDAM reprogramming, indicating that the state remains plastic across advanced aging. This contrasts with germline Asxl1 mutations in myeloid malignancies, where H3K4me3 loss leads to irreversible lipid accumulation [27, 45]. Therapeutically, O-GlcNAc transferase activators that stabilize endogenous Asxl1 [27], or LXRα agonists that by pass Asxl1 haplo-insufficiency, may offer druggable alternatives to gene therapy. Together, our genetic rescue data argue that the development of the LDAM state is potentially preventable. They identify the Asxl1/LXRα/H3K4me3 axis as a candidate regulatory node, highlighting a potential avenue for early intervention in ApoE4-driven microglial pathogenesis.

Fig. 10.

Schematic Mechanism: We therefore proposed a mechanistic hypothesis: ApoE4 acts to suppress Asxl1 expression. This initial insult destabilizes the Asxl1–LXRα–MLL complex, leading to diminished LXRα binding and a loss of the activating H3K4me3 mark at the Abca1 promoter. The transcriptional downregulation of Abca1 that follows disrupts cellular lipid homeostasis, directly driving microglial lipid droplet accumulation and functional impairment, including compromised Aβ clearance (as shown in the left panel of the mechanism schematic). Conversely, bolstering Asxl1 expression can restore this regulatory axis, enhance lipid efflux, and promote a functional, anti-inflammatory microglial state (right panel), thereby validating the proposed pathway

Limitations and future directions

Our study establishes that ApoE4 rewires the microglial epigenome through an Asxl1–LXRα–H3K4me3 axis to generate LDAM, and identifies Asxl1 up-regulation as a therapeutic entry point for ApoE4-driven LDAM in Alzheimer’s disease. Several caveats merit attention. First, although our knock-in mouse model faithfully expresses human ApoE4, it does not fully recapitulate the human lipidome or the tempo of brain aging. Second, All experiments only used age-matched female ApoE3-TR and ApoE4-TR mice. Males were excluded because ApoE4 confers a higher Alzheimer’s disease risk and more severe pathology in females. Focusing on one sex reduces biological variation, allowing a more controlled analysis of core ApoE4 effects and establishing a baseline for future comparisons. A limitation is the lack of age-matched male controls, which prevents definitive conclusions on sex differences and restricts generalization beyond females. This study details ApoE4 effects in this higher-risk group, providing a necessary foundation for future work including both sexes to fully elucidate ApoE4-related sex differences in pathophysiology. Third, a key limitation is that we did not employ established Alzheimer’s disease mouse models (which encompass both tau and Aβ pathology). Consequently, we did not interrogate the consequences of Asxl1 modulation on tau or Aβ pathology—processes known to be strongly shaped by ApoE4 [46]. Therefore, future studies should utilize triple-transgenic ApoE4 models to address this gap. Promising next steps include (i) structure-guided design of small-molecule stabilizers that fortify the Asxl1–LXRα interface, and (ii) lipid-nanoparticle delivery of Asxl1 mRNA to bypass germline manipulation. Finally, while our epigenetic rescue paradigm is encouraging, rigorous long-term safety profiling and precise dose titration will be prerequisites for clinical translation.

Conclusions

Our findings establish a direct link between ApoE4 and microglial lipid dysregulation via the novel Asxl1–LXRα–H3K4me3 axis, moving beyond association to define a causal epigenetic mechanism in AD pathogenesis. The demonstration that CRISPR-mediated Asxl1 restoration can reverse key pathological features—impaired cholesterol efflux and Aβ phagocytosis—in an ApoE4 context transforms our understanding of this genetic risk factor from a static vulnerability to a therapeutically modifiable driver. Enhancing Asxl1 function thus emerges as a compelling strategic avenue for precision medicine in ApoE4-associated Alzheimer’s disease.

Methods

Animals

ApoE3-TR and ApoE4-TR mice (Jackson Laboratory, #005861, #005862) were maintained on a C57BL/6J background. Conditional Asxl1 over-expression mice (Rosa26^CAG-lox-STOP-lox-Asxl1-3xFLAG) were crossed to CX3CR1-Cre mice (Cyagen Biosciences) and back crossed to ApoE4-TR for > 10 generations. All experiments utilized age-matched female mice. For the aging study, ApoE3-TR and ApoE4-TR mice were analyzed at 1, 5, 10, and 15 months of age. For the overexpression study, 10-month-old ApoE4-TR mice overexpressing Asxl1 were compared with their age-matched floxed littermate controls. Currently, this strain of mice has been successfully bred and cultivated in our experimental animal center. They are fed in an SPF grade environment, with a temperature of 21 ± 2 ℃ and a light exposure time controlled at 12:12 h (with a light on time of 6 AM and a light off time of 6 PM), and sufficient water and food. All experimental protocols and procedures have been approved by the Institutional Animal Care and Use Committee of Fuzhou University Affiliated Provincial Hospital (Animal Ethics Approval Number: IACUC-FPH-SL-20240228 [0125]).

Single-nucleus RNA Sequencing of Mouse Hippocampal Tissue

Hippocampi were dissected from female ApoE3-TR and ApoE4-TR mice at 5, 10, and 15 months of age (n = 4 per age and genotype). Following CO₂ euthanasia, brains were rapidly extracted. Under a stereomicroscope, both hippocampi were carefully isolated, cleared of meninges and choroid plexus, and either immediately snap-frozen in liquid nitrogen or placed in ice-cold nuclei isolation buffer for immediate processing. For nuclei isolation, hippocampal tissues were minced on a chilled plate and transferred into 2 mL of ice-cold nuclei isolation buffer (0.25 M sucrose, 25 mM KCl, 5 mM MgCl₂, 10 mM Tris-HCl pH 7.4, 0.1% Triton X-100, 1× RNase inhibitor, 0.1% BSA). Tissues were homogenized using a Dounce homogenizer with 10–15 strokes of a loose pestle. The homogenate was sequentially filtered through 70 μm and 40 μm cell strainers. To remove myelin and debris, the filtered homogenate was underlaid with a sucrose cushion (1.8 M sucrose in nuclei isolation buffer without Triton X-100) and centrifuged at 1,300 ×g for 30 min at 4 °C. The supernatant was carefully discarded, and the purified nuclei pellet was gently resuspended in 1 mL of wash buffer (PBS containing 1% BSA and 1× RNase inhibitor). After centrifugation at 500 ×g for 5 min at 4 °C, the pellet was resuspended in a small volume of wash buffer and passed through a 30 μm flow cytometry-compatible strainer twice. Nuclei were kept on ice for quality assessment, counting, and subsequent use in single-nucleus RNA sequencing library preparation.

snRNA-seq data processing and analysis

This study analyzed single-nucleus RNA sequencing (snRNA-seq) data generated from hippocampal tissues of ApoE3 and ApoE4 knock-in mice at 5, 10, and 15 months of age. The data are available in the Gene Expression Omnibus (GEO) under accession number [GSE319246].

The Seurat R package was employed with default settings to process single-nucleus feature counts and digital gene expression matrices, which were subsequently transformed into Seurat objects. Expression matrices from various datasets were integrated utilizing the Merge function in R. For the analysis, genes identified in a minimum of three cells and cells expressing at least 200 genes were retained. Low-quality or non-viable cells were excluded by filtering out cells in the top or bottom 1% of total gene expression levels and those exhibiting mitochondrial gene expression exceeding 8%, using the Seurat package. Normalization of single-cell unique molecular identifier (UMI) data was achieved through regularized negative binomial regression. To mitigate batch effects across datasets, the Harmony method was applied, effectively reducing significant cell-lineage batch effects within the principal component analysis space. Dimensionality reduction and clustering were performed using the Seurat package, and the outcomes were visualized via the uniform manifold approximation and projection (UMAP) algorithm.

To reconstruct the dynamic transitions of microglial subpopulations, pseudotime trajectory analysis was performed using the Slingshot package. The integrated and harmony‑corrected microglial subset (clusters 9 and 21) was used as input. The UMAP embedding was adopted as the reduced‑dimensional space for trajectory inference. Starting from the cluster annotated as homeostatic microglia (HOMEOSTATIC), Slingshot was run to construct smooth branching trajectories toward the disease‑associated microglia (DAM) and lipid‑associated microglia (LDAM) clusters. Pseudotime values were assigned to each cell along the inferred paths, representing the relative progression from the homeostatic state to activated states.

To identify differentially expressed genes (DEGs), we employed the Seurat FindMarkers function, utilizing the Wilcoxon likelihood-ratio test with default parameters. Genes were classified as DEGs if they were expressed in over 10% of the cells within a cluster and exhibited an average log fold change greater than 0.25. For cell type annotation of each cluster, we integrated the expression profiles of canonical markers identified among the DEGs with existing literature knowledge. Doublet cells, characterized by the expression of markers from multiple cell types, were manually excluded from the analysis. The cell type identity of each cluster was determined by cross-referencing the expression of canonical markers found in the DEGs with the CellMarker database.

To explore the potential functions of differentially expressed genes (DEGs), we conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses utilizing the “clusterProfiler” R package version 3.16.1. Pathways with an adjusted p-value (p_adj) of less than 0.05 were considered significantly enriched. The GO gene sets employed as references encompassed the categories of molecular function (MF), biological process (BP), and cellular component (CC). For the Gene Set Variation Analysis (GSVA) pathway enrichment analysis, the mean gene expression of each cell type was utilized as input data, employing the GSVA package.

Immunohistochemistry and lipid droplet imaging

Brain tissues were fixed with 4% PFA and cryosectioned at 30 μm thickness for analysis. Sections were incubated with the following primary antibodies: anti-Iba1 (Wako, 1:500) and anti-Perilipin2 (Proteintech, 1:200). Additionally, neutral lipids were visualized using the lipophilic fluorescent dye BODIPY. Images were acquired on a Zeiss LSM980 confocal microscope; droplet volume and number were quantified in 3D using Imaris (Bitplane).

Image processing & 3-D quantification

Images were acquired on a Zeiss LSM 980 inverted confocal microscope equipped with a 63×/1.4 NA Plan-Apochromat oil-immersion objective, Airyscan 2 detector, and ZEN 3.5 software. Z-stacks: 0.3 μm optical sections, 20–25 planes per stack, covering the full 30 μm thickness. Pixel dwell time 1.58 µs, 2× line averaging, 1,024 × 1,024 frame size (0.09 μm × 0.09 μm × 0.3 μm voxel). Imaris 9.9 (Bitplane) was used for: 3-D surface rendering of microglia (Iba1 channel) to define cell volumes. Spot detection (Bodipy) to identify individual lipid droplets: Seed threshold = 0.4 μm diameter (3 voxels); Quality filter = ≥ 0.3 μm³ volume; Batch analysis wizard applied identical settings across all brains. Outputs: Total droplet number per microglial cell. Total droplet volume per microglial cell (µm³). Droplet size distribution (diameter: <0.2, 0.2–0.4, > 0.4 μm).

In vivo Aβ clearance

Preparation of Aβ peptides human Aβ ( 1–42) hydrochloride salt (Bachem, Buben dorf, Switzerland, H-6466) was dissolved to 1 mM in hexafluoroisopropanol (HFIP; Sigma-Aldrich, 10522–8), aliquoted and stored at − 20 °C after HFIP evaporation. For oligomeric assembly, peptides were resuspended to 5 mM in DMSO by sonication, then diluted to 100 µM in phenol red-free DMEM and incubated at 4 °C for 24 h.

FITC-Aβ( 1–42) preparation for in vivo injection : Lyophilized Aβ( 1–42)-FITC was treated identically (HFIP film → DMSO → 100 µM DMEM) to ensure comparable assembly state. Final injectate was passed through a 0.22 μm PVDF filter into a sterile glass vial.

10 month-old ApoE3-TR and ApoE4-TR mice (n = 8 per group) were fasted 4 h prior to injection to reduce plasma lipoprotein interference. Tail-vein catheterization: animals were warmed under a heat lamp for 5 min to dilate the vein; a 26 G catheter was inserted and flushed with 50 µl sterile saline. 5 µg Aβ( 1–42)-FITC per gram body weight (≈ 100 µl volume for 25 g mouse) was delivered as a slow bolus over 30 s. Post-injection, the catheter was flushed with 50 µl saline to ensure complete delivery; At 24 h post-injection, animals were re-anesthetized with isoflurane and transcardially perfused with 30 ml ice-cold PBS followed by 30 ml 4% PFA. Brains were removed, post-fixed overnight at 4 °C, cryoprotected in 30% sucrose (48 h), and sectioned oronally at 30 μm using a Leica CM3050S cryostat. Sections were counterstained with DAPI (1 µg/ml, 10 min) and mounted with ProLong Gold. Low-magnification tiled images (20×, 0.8 NA) were acquired to reconstruct the entire hippocampus and cortex. High-resolution z-stacks (63×/1.4 NA oil, 0.3 μm z-step, 20 planes) were captured in dentate gyrus. FITC-Aβ fluorescence intensity was measured in Imaris 9.9: 3-D surface masks of microglia (Iba1 staining) were generated. Mean Fluorescence Intensity (MFI) of FITC-Aβ within each mask, normalized to background (non-cellular parenchyma), was calculated.

FACS-based quantification of microglial uptake

Anesthetize 10 month old mice with isoflurane (2%) and inject 5 µg fluorescent Aβ (≈ 0.25 mg kg⁻¹) in 100 µL sterile PBS via the lateral tail vein. Deeply anesthetize animals, transcardially perfuse with 20 mL ice-cold PBS, and rapidly dissect cortices/hippocampi on ice.

Mince tissue into 1 mm³ pieces and enzymatically digest (papain 20 U mL⁻¹, DNase I 50 µg mL⁻¹, 30 min, 37 °C). Triturate gently, filter through 70 μm strainer, and pellet cells (300 × g, 10 min). Resuspend in 37% Percoll, underlay with 70% Percoll, and centrifuge (800 × g, 20 min, brake off). Collect the interface layer (microglia-enriched), wash twice with FACS buffer (2% FBS, 2 mM EDTA in PBS). Incubate 1 × 10⁶ cells with Fc block (anti-CD16/32, 1:100, 10 min, 4 °C), then stain for 30 min on ice with the following fluorochrome-conjugated antibodies: CD11b-PE/Cy7 (clone M1/70, 1:200), and Fixable Viability Dye eFluor 780 (1:1000). Wash twice and resuspend in 300 µL FACS buffer. Acquire on a 5-laser BD FACS Aria III; compensate with single-color controls. Gate sequentially: (i) singlets (FSC-H vs. FSC-A), (ii) live cells (Viability Dye-negative), (iii) microglia (CD11b⁺), and (iv) Aβ-positive events (HiLyte488⁺). Record mean fluorescence intensity (MFI) and percentage of HiLyte488⁺ microglia. Export data for further statistical analysis.

Primary microglia culture

Mixed glia from P0–P2 pups were cultured in DMEM/F12 + 10% FBS. Microglia were shaken off (200 rpm, 2 h) and seeded at 2 × 10⁵ cells per well. Cholesterol loading with cholesterol–BSA complexes : A 20 mM cholesterol stock was prepared in absolute ethanol. 100 µl of this stock was rapidly injected into 9.9 ml 37 °C KRB (Krebs–Ringer bicarbonate buffer) containing 5% essentially fatty-acid-free BSA (Sigma A6003) under constant vortexing. The mixture was sonicated on ice for 30 s (30% amplitude, 1 s pulse/1 s rest) and sterile-filtered (0.22 μm). Final cholesterol concentrations: 0, 5, 10, 50 µg/ml. Cholesterol–BSA complexes (or BSA-only vehicle) were added in triplicate wells and incubated for 24 h at 37 °C, 5% CO₂. Medium was replenished after 24 h to maintain cholesterol levels.

Aβ-FITC uptake assay (IncuCyte S3 live-cell imaging)

After 24 h cholesterol loading, medium was removed, cells were washed once with warm PBS, and fresh serum-free medium containing 1 µM Aβ( 1–42)-FITC (AnaSpec AS-23525-01, oligomeric form prepared as previously described) was added for 4 h. Plates were transferred to an IncuCyte S3 live-cell analysis system maintained at 37 °C, 5% CO₂. Images were acquired using a 20× objective under the FITC channel. After background subtraction, the integrated fluorescent intensity (IFI) of each well was quantified with IncuCyte software. Data are expressed as relative fluorescent units (RFU) and normalized to cell confluence determined by phase-contrast masking.

Cholesterol efflux assay with [³H]-cholesterol

After 24 h cholesterol loading, cells were washed and incubated overnight in serum-free DMEM containing 1 µCi ml⁻¹ [1,2-³H(N)]-cholesterol (PerkinElmer NET139001MC) complexed to 0.2% BSA. Following labelling, cells were washed 3× with warm PBS to remove unincorporated isotope. Fresh serum-free medium ± 10 µg ml⁻¹ human ApoA1 (Sigma SRP4695) and human HDL(Solarbio L1567) were added (triplicate wells). After 4 h at 37 °C, 100 µl medium was collected and centrifuged (500 g, 5 min) to remove cell debris. Cells were lysed in 100 µl 0.1 M NaOH/0.1% Triton X-100 for 30 min at RT. 50 µl aliquots of medium and lysate were mixed with 3 ml Ultima Gold scintillation cocktail and counted in a Tri-Carb 2910 TR liquid-scintillation counter (PerkinElmer). Efflux (%) = [cpm_medium/(cpm_medium + cpm_lysate)] × 100.

Chromatin immunoprecipitation (ChIP)

ChIP on brain tissue was performed according to previously published. Briefly, hippocampal tissues were homogenized in ice cold PBS to create chunks with 0.5 mm³ or smaller, then subjected to cross-linking with formaldehyde (0.75% final) for 8 min at room temperature, followed by glycine (150 mM final) quenching for an additional 10 min at room temperature. Cross-linked tissues were harvested by centrifugation at 2,000 g for 10 min at 4^。C, resuspended and homogenized in ice-cold lysis buffer (50 mM Tris pH 8.0, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100). Nuclei were collected by centrifugation at 2,000 g for 5 min at 4^。C, and suspended in ice-cold ChIP-seq nuclear lysis buffer (10 mM Tris pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 0.2% SDS). Chromatin was sheared to an average size of 200-1,000 bp using the Bioruptor UCD-200 Sonicator. 10% of the sonicated chromatin was saved as an input. Sheared chromatin was incubated with the indicated antibodies, that were conjugated to protein A/G beads (MCE) overnight in IP buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA pH 8.0, 1% NP-40, 0.5% Sodium Deoxycholate, 0.1% SDS, Protease Inhibitors (MCE) at 4^。C. The next day, beads were washed one time with low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA pH 8.0, 150 mM NaCl, 20 mM Tris-HCl pH 8.0), one time with high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA pH 8.0, 500 mM NaCl, 20 mM Tris-HCl pH 8.0), one time with LiCl buffer (0.25 M LiCl, 1% NP-40, 1% Sodium Deoxycholate, 1 mM EDTA, 10 mM Tris-HCl pH 8.0), and then resuspended in elution buffer (1% SDS, 100 mM NaHCO3). DNA was treated with RNase (Sigma), proteinase K (Sangon), cross-link reversal and purified with PCR purification kit (Sangon) according to the manufacturer’s instructions. Purified DNA was subjected to sequencing or qPCR analyses.

Co-immunoprecipitation

Primary microglia (MG) were isolated from post-natal day 0–2 (P0–P2) mice as previously described. Mixed glial cultures were prepared in DMEM/F12 supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (P/S). Microglia were purified by gentle shaking (200 rpm, 2 h) and seeded at a density of 5 × 10⁵ cells per well in 6-well plates.After 48 h of culture, cells were treated with or without cholesterol–BSA complexes (0–50 µg/ml) for an additional 24 h. Cells were then washed twice with ice-cold PBS and lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitors (Roche) and phosphatase inhibitors (Roche). Lysates were sonicated for 10 s (30% amplitude) to ensure complete lysis and then centrifuged at 14,000 xg for 15 min at 4 °C to remove insoluble debris. Protein concentration was determined using the BCA assay (Pierce). Anti-Asxl1 antibody (Abcam, ab198247) was diluted to 2 µg/ml in PBS and coupled to Protein G Dynabeads (Invitrogen) according to the manufacturer’s instructions. Beads were then washed three times with PBS to remove unbound antibody. 500 µg of total protein lysate was diluted to 500 µl with lysis buffer and incubated with the antibody-coupled Dynabeads overnight at 4 °C with gentle rotation. After incubation, the beads were washed five times with RIPA buffer to remove non-specific binding. Each wash step involved 1-minute incubation with gentle rotation followed by magnetic separation.Bound proteins were eluted from the beads by adding 50 µl of 2× Laemmli sample buffer (Bio-Rad) containing 5% β-mercaptoethanol and boiling for 5 min at 95 °C. The eluted proteins were then separated by SDS-PAGE on a 10% Bis-Tris gel (Invitrogen). After electrophoresis on 10% SDS-PAGE, immunoprecipitated products were analyzed by Western blotting (WB) using anti-LXRα monoclonal antibody (Abcam; 1:1000).

Magnetic Bead Separation Protocol for Isolating Microglia from Mouse Brain

Anesthetize the mice and perfuse the brain tissue with ice-cold PBS to remove blood. Under a dissecting microscope, carefully dissect the target brain region (hippocampus) and remove the meninges. Mince the brain tissue into small pieces (approximately 1 mm³). Add pre-warmed digestion solution (Papain digestion solution) and incubate at 37 °C for 15–20 min. Terminate the digestion by adding serum-containing culture medium. Pass the digested tissue through a 70 μm cell strainer to collect a single-cell suspension.Wash the cells with MACS Buffer and centrifuge at 300×g for 3 min, discarding the supernatant. Resuspend the cell pellet in MACS Buffer and add an appropriate amount of CD11b+ magnetic beads (typically 10–20 µL per 10⁷ cells). Incubate at 4 °C for 15 min, gently mixing every 5 min. Filter the incubated cell suspension through a 30 μm nylon mesh or 40 μm filter. Load the filtered cell suspension onto a MACS separation column (LS column) placed in the magnetic field. Collect the flow-through (unlabeled cells), while microglia will be retained on the column.Wash the column with MACS Buffer, repeating twice (each time adding 3 mL of buffer). Remove the separation column from the magnetic field, add 5 mL of MACS Buffer, and flush the retained cells off the column using a plunger; these are the microglia. By following the above steps, highly pure microglia can be efficiently isolated from mouse brain tissue, suitable for gene expression analysis, Western Blot Analysis.

Western Blot Analysis

After isolating microglia using magnetic bead separation, we perform Western blot analysis to study protein expression. Briefly, After sorting, microglia were lysed using RIPA buffer to release intracellular proteins. Protein concentration was determined using the BCA assay. Subsequently, protein separation was performed via 4–12% SDS-PAGE electrophoresis. The separated proteins were then transferred to a PVDF membrane using electroblotting. After transmembrane, the tissues were blocked in 5% bovine serum albumin (BSA), incubated with primary antibodies at 4℃ overnight, and further incubated with secondary antibodies for one hour (LI-COR, USA). Finally, the membranes were detected with the Odyssey Sa color infrared laser imaging system (LI-COR, USA) andanalysis with Image J software. The primary antibodies were as follows: anti-GAPDH (mouse1:2000, cell signaling technology CST), anti-Abca1 (1:1000, Abcam), anti-Asxl1 (mouse 1:1000, Abcam), anti-Asxl2 (rabbit 1:1000, Abcam), anti-H3 (mouse 1:500, CST), anti-H3K4me3 (rabbit 1:1000, Abcam), anti-H3K27me2 (rabbit 1:1000, Abcam), anti-H3K9me 2(rabbit 1:1000, Abcam).

ELISA Protocol for Quantification of IL-6 and CXCL1

Supernatant levels of IL-6 and CXCL1 were measured with commercial DuoSet ELISA kits (R&D, DY406 and DY453) following the manufacturer’s instructions. Briefly, 96-well plates were coated overnight with capture antibody (2 µg mL⁻¹), blocked, and incubated with standards or samples (100 µL, 2 h). After biotinylated detection antibody and streptavidin-HRP steps, TMB substrate was added, absorbance read at 450 nm, and concentrations calculated from 4-parameter logistic curves.

Real-time polymerase chain reaction

Total RNA was extracted from the isolated microglia using TriPure isolation reagent (Roche), and reverse transcription was performed using cDNA synthesis kit. Then perform polymerase chain reaction (PCR) and measure fluorescence using the Step One Plus real-time PCR system (Life Technologies Applied Biosystems). The primers used are as follows:

Gene	Locus	Forward/Reverse Primer
Abca1	NM_013454	sense 5-GCTTGTTGGCCTCAGTTAAGG-3; anti-sense 5-GTAGCTCAGGCGTACAGAGAT−3
Asxl1	NM_001039939	sense 5-CTACTCAGATGCTCCAATGACAC-3; antisense 5-TGAAAAGACTAATGCGGCCAG-3
Asxl2	NM_172421	sense 5-TGTCCCAGTAGTTCCTCAGTC-3; anti-sense 5-TGGGTTTCATGGTGATAAGCTC-3
CxcL1	NM_008176	sense 5-CTGGGATTCACCTCAAGAACATC-3; anti-sense 5-CAGGGTCAAGGCAAGCCTC-3
IL-6	NM_031168	sense 5-CCAAGAGGTGAGTGCTTCCC-3 anti-sense 5-CTGTTGTTCAGACTCTCTCCCT-3
Actin	NM_007393	sense 5-GGCTGTATTCCCCTCCATCG − 3; anti-sense 5- CCAGTTGGTAACAATGCCATGT − 3

Data statistical analysis

Statistical analysis was carried out using GraphPad Prism 9.0. For most data, the Gaussian distribution of the data was assessed using the D’Agostino and Pearson normality and ShapiroWilk normality tests. If the data passed the Gaussian distribution test, parametric unpaired two-tailed t-tests were used for two groups and one-way analyses of variance(ANOVA), followed by Tukey’s multiple comparisons tests, for three or more groups. Otherwise, nonparametric unpaired Mann–Whitney tests were used for two groups. Two-way ANOVAs with Bonferroni’s multiple comparisons test was used for experiments containing two independent variables.

Abbreviations

ApoE4

Apolipoprotein E4

AD

Alzheimer’s disease

LDAM

Lipid-droplet-accumulating microglia

TR

Targeted replacement

DG

Dentate gyrus

LOAD

Late-onset Alzheimer’s disease

Aβ

Amyloid-β

iMG

induced pluripotent stem cell-derived microglia

Aβ

fibrillar Aβ

LD

Lipid-droplet

iPSC

Induced pluripotent stem cells

DAMPs

Damage-associated molecular patterns

Asxl1

Additional Sex Combs-Like 1

MG

Microglia

PLIN2

Perilipin 2

CHO

Cholesterol

ApoE4/Asxl1⁺⁺

ApoE4/Asxl1-overexpressing

Co-IP

Co-immunoprecipitation

MFI

Mean fluorescence intensity

ANOVA

Analysis of variance

PCR

Polymerase chain reaction

mRNA

messenger-RNA

me

Methylation

Authors’ contributions

Lanyan Lin: Conceptualization, Data curation and analysis, Investigation, Project administration, Writing original draft, Reviewing & editing. Zhen Pan: Data curation and analysis, animal behavior test, Methodology, Writing original draft. Zhen Wei: Data curation and analysis, animal behavior test, Methodology, Writing original draft. Xiulong Jiang: Methodology, behavior test, Data analysis. Yongxing Lai: Methodology, animal behaviortest. Mingfeng Chen: Methodology, animal behavior test. FanLin: Conceptualization, Funding acquisition, Project administration, Resources, Supervision.

Funding

These studies were funded by National Natural Science Foundation of China (No.82401643), Fujian Provincial Science and Technology Innovation Joint Funding Project (No.2022J01413, No.2023Y9289, No.2025J01073), and the National and Fujian Province’s Key Clinical Specialty Discipline Construction Programs of P.R.C.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The animal study protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of Fuzhou University Affiliated Provincial Hospital (Protocol Number: IACUC-FPH-SL-20240228[0125]). All methods were carried out in compliance with the ARRIVE guidelines and the NIH Guide for the Care and Use of Laboratory Animals.

Consent for publication

All participants provided written informed consent for the publication of their anonymized data.

Competing interests

The authors declare no competing interests.